AuroraMiddleware/skia2
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity
85f0e63ca4
skia2/tests/PathOpsOpTest.cpp

6552 lines
266 KiB
C++
Raw Normal View History

Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
/*
* Copyright 2012 Google Inc.
*
* Use of this source code is governed by a BSD-style license that can be
* found in the LICENSE file.
*/
#include "PathOpsExtendedTest.h"
Fix last pathops skp bug This fixes the last bug discovered by iterating through the 800K skp corpus representing the top 1M websites. For every clip on the stack, the paths are replaced with the pathop intersection. The resulting draw is compared with the original draw for pixel errors. At least two prominent bugs remain. In one, the winding value is confused by a cubic with an inflection. In the other, a quad/cubic pair, nearly coincident, fails to find an intersection. These minor changes include ignoring very tiny self-intersections of cubics, and processing degenerate edges that don't connect to anything else. R=reed@android.com TBR=reed Author: caryclark@google.com Review URL: https://codereview.chromium.org/340103002
2014-06-18 14:24:19 +00:00
#include "PathOpsTestCommon.h"
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
Path ops formerly found the topmost unprocessed edge and determined its angle sort order to initialize the winding. This never worked correctly with cubics and was flaky with paths consisting mostly of vertical edges. This replacement shoots axis-aligned rays through all intersecting edges to find the outermost one either horizontally or vertically. The resulting code is smaller and twice as fast. To support this, most of the horizontal / vertical intersection code was rewritten and standardized, and old code supporting the top-directed winding was deleted. Contours were pointed to by an SkTDArray. Instead, put them in a linked list, and designate the list head with its own class to ensure that methods that take lists of contours start at the top. This change removed a large percentage of memory allocations used by path ops. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1111333002
2015-05-11 14:21:27 +00:00
class PathTest_Private {
public:
PathTest_Private(SkPath* path)
: fPath(path) {}
void setPt(int index, SkScalar x, SkScalar y) {
fPath->setPt(index, x, y);
}
SkPath* fPath;
};
static void path_edit(const SkPoint& from, const SkPoint& to, SkPath* path) {
PathTest_Private testPath(path);
for (int index = 0; index < path->countPoints(); ++index) {
if (SkDPoint::ApproximatelyEqual(path->getPoint(index), from)) {
testPath.setPt(index, to.fX, to.fY);
return;
}
}
}
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp1d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(0,2, 1,0, 1,0);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,1);
pathB.cubicTo(0,1, 1,0, 2,0);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp2d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,2);
path.cubicTo(0,1, 1,0, 1,0);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,1);
pathB.cubicTo(0,1, 2,0, 1,0);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp3d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(2,3, 1,0, 1,0);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,1);
pathB.cubicTo(0,1, 1,0, 3,2);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp5d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(0,2, 1,0, 2,0);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,1);
pathB.cubicTo(0,2, 1,0, 2,0);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp6d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(0,6, 1,0, 3,0);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,1);
pathB.cubicTo(0,3, 1,0, 6,0);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp7d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(3,4, 1,0, 3,0);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,1);
pathB.cubicTo(0,3, 1,0, 4,3);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp8d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(0,5, 1,0, 4,0);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,1);
pathB.cubicTo(0,4, 1,0, 5,0);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp9d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(1,6, 1,0, 2,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,1);
pathB.cubicTo(1,2, 1,0, 6,1);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void quadOp9d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.quadTo(1,6, 1.5f,1);
path.quadTo(1.5f,0.5f, 2,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,1);
pathB.quadTo(1,2, 1.4f,1);
pathB.quadTo(3,0.4f, 6,1);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void lineOp9d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.lineTo(1,6);
path.lineTo(1.5f,1);
path.lineTo(1.8f,0.8f);
path.lineTo(2,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,1);
pathB.lineTo(1,2);
pathB.lineTo(1.4f,1);
pathB.lineTo(3,0.4f);
pathB.lineTo(6,1);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp1i(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(1,2, 1,0, 2,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,1);
pathB.cubicTo(1,2, 1,0, 2,1);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp10d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(1,3, 1,0, 4,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,1);
pathB.cubicTo(1,4, 1,0, 3,1);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp11d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(3,4, 1,0, 5,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,1);
pathB.cubicTo(1,5, 1,0, 4,3);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp12d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(1,6, 1,0, 1,0);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,1);
pathB.cubicTo(0,1, 1,0, 6,1);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp13d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(4,5, 1,0, 5,3);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,1);
pathB.cubicTo(3,5, 1,0, 5,4);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp14d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(0,2, 2,0, 2,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,2);
pathB.cubicTo(1,2, 1,0, 2,0);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp15d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(3,6, 2,0, 2,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,2);
pathB.cubicTo(1,2, 1,0, 6,3);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp16d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,2);
path.cubicTo(0,1, 3,0, 1,0);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,3);
pathB.cubicTo(0,1, 2,0, 1,0);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp17d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,2);
path.cubicTo(0,2, 4,0, 2,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,4);
pathB.cubicTo(1,2, 2,0, 2,0);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp18d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(3,5, 2,0, 2,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,2);
pathB.cubicTo(1,2, 1,0, 5,3);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp19i(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,2);
path.cubicTo(0,1, 2,1, 6,2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1,2);
pathB.cubicTo(2,6, 2,0, 1,0);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp20d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(0,1, 6,0, 2,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,6);
pathB.cubicTo(1,2, 1,0, 1,0);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp21d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(0,1, 2,1, 6,5);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1,2);
pathB.cubicTo(5,6, 1,0, 1,0);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp22d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(2,3, 3,0, 2,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,3);
pathB.cubicTo(1,2, 1,0, 3,2);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp23d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(1,2, 4,0, 2,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,4);
pathB.cubicTo(1,2, 1,0, 2,1);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp24d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(1,2, 2,0, 3,2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,2);
pathB.cubicTo(2,3, 1,0, 2,1);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void testIntersect1(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath one, two;
one.addRect(0, 0, 6, 6, SkPath::kCW_Direction);
two.addRect(3, 3, 9, 9, SkPath::kCW_Direction);
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, one, two, kIntersect_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void testUnion1(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath one, two;
one.addRect(0, 0, 6, 6, SkPath::kCW_Direction);
two.addRect(3, 3, 9, 9, SkPath::kCW_Direction);
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, one, two, kUnion_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void testDiff1(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath one, two;
one.addRect(0, 0, 6, 6, SkPath::kCW_Direction);
two.addRect(3, 3, 9, 9, SkPath::kCW_Direction);
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, one, two, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void testXor1(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath one, two;
one.addRect(0, 0, 6, 6, SkPath::kCW_Direction);
two.addRect(3, 3, 9, 9, SkPath::kCW_Direction);
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, one, two, kXOR_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void testIntersect2(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath one, two;
one.addRect(0, 0, 6, 6, SkPath::kCW_Direction);
two.addRect(0, 3, 9, 9, SkPath::kCW_Direction);
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, one, two, kIntersect_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void testUnion2(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath one, two;
one.addRect(0, 0, 6, 6, SkPath::kCW_Direction);
two.addRect(0, 3, 9, 9, SkPath::kCW_Direction);
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, one, two, kUnion_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void testDiff2(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath one, two;
one.addRect(0, 0, 6, 6, SkPath::kCW_Direction);
two.addRect(0, 3, 9, 9, SkPath::kCW_Direction);
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, one, two, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void testXor2(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath one, two;
one.addRect(0, 0, 6, 6, SkPath::kCW_Direction);
two.addRect(0, 3, 9, 9, SkPath::kCW_Direction);
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, one, two, kXOR_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void testOp1d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.addRect(0, 0, 1, 1, SkPath::kCW_Direction);
path.addRect(0, 0, 2, 2, SkPath::kCW_Direction);
pathB.setFillType(SkPath::kWinding_FillType);
pathB.addRect(0, 0, 1, 1, SkPath::kCW_Direction);
pathB.addRect(0, 0, 1, 1, SkPath::kCW_Direction);
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void testOp2d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.addRect(0, 0, 1, 1, SkPath::kCW_Direction);
path.addRect(0, 0, 2, 2, SkPath::kCW_Direction);
pathB.setFillType(SkPath::kEvenOdd_FillType);
pathB.addRect(0, 0, 1, 1, SkPath::kCW_Direction);
pathB.addRect(0, 0, 1, 1, SkPath::kCW_Direction);
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void testOp3d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.addRect(0, 0, 1, 1, SkPath::kCW_Direction);
path.addRect(1, 1, 2, 2, SkPath::kCW_Direction);
pathB.setFillType(SkPath::kWinding_FillType);
pathB.addRect(0, 0, 1, 1, SkPath::kCW_Direction);
pathB.addRect(0, 0, 1, 1, SkPath::kCW_Direction);
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void testOp1u(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.addRect(0, 0, 1, 1, SkPath::kCW_Direction);
path.addRect(0, 0, 3, 3, SkPath::kCW_Direction);
pathB.setFillType(SkPath::kWinding_FillType);
pathB.addRect(0, 0, 1, 1, SkPath::kCW_Direction);
pathB.addRect(0, 0, 1, 1, SkPath::kCW_Direction);
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kUnion_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void testOp4d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.addRect(0, 0, 1, 1, SkPath::kCW_Direction);
path.addRect(2, 2, 4, 4, SkPath::kCW_Direction);
pathB.setFillType(SkPath::kWinding_FillType);
pathB.addRect(0, 0, 1, 1, SkPath::kCW_Direction);
pathB.addRect(0, 0, 1, 1, SkPath::kCW_Direction);
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void testOp5d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kEvenOdd_FillType);
path.addRect(0, 0, 2, 2, SkPath::kCW_Direction);
path.addRect(0, 0, 3, 3, SkPath::kCW_Direction);
pathB.setFillType(SkPath::kEvenOdd_FillType);
pathB.addRect(0, 0, 1, 1, SkPath::kCW_Direction);
pathB.addRect(0, 0, 1, 1, SkPath::kCW_Direction);
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void testOp6d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kEvenOdd_FillType);
path.addRect(0, 0, 1, 1, SkPath::kCW_Direction);
path.addRect(0, 0, 3, 3, SkPath::kCW_Direction);
pathB.setFillType(SkPath::kWinding_FillType);
pathB.addRect(0, 0, 1, 1, SkPath::kCW_Direction);
pathB.addRect(0, 0, 1, 1, SkPath::kCW_Direction);
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void testOp7d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kEvenOdd_FillType);
path.addRect(0, 0, 2, 2, SkPath::kCW_Direction);
path.addRect(0, 0, 1, 1, SkPath::kCW_Direction);
pathB.setFillType(SkPath::kEvenOdd_FillType);
pathB.addRect(0, 0, 1, 1, SkPath::kCW_Direction);
pathB.addRect(0, 0, 1, 1, SkPath::kCW_Direction);
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void testOp2u(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kEvenOdd_FillType);
path.addRect(0, 0, 2, 2, SkPath::kCW_Direction);
path.addRect(0, 0, 2, 2, SkPath::kCW_Direction);
pathB.setFillType(SkPath::kWinding_FillType);
pathB.addRect(0, 0, 3, 3, SkPath::kCW_Direction);
pathB.addRect(1, 1, 2, 2, SkPath::kCW_Direction);
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kUnion_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void testOp8d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.addRect(0, 0, 640, 480);
pathB.moveTo(577330, 1971.72f);
pathB.cubicTo(10.7082f, -116.596f, 262.057f, 45.6468f, 294.694f, 1.96237f);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp25i(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(2,4, 5,0, 3,2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,5);
pathB.cubicTo(2,3, 1,0, 4,2);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp26d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(3,4, 4,0, 3,2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,4);
pathB.cubicTo(2,3, 1,0, 4,3);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp27d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(3,6, 1,0, 5,2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,1);
pathB.cubicTo(2,5, 1,0, 6,3);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp28u(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(1,4, 6,0, 3,2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,6);
pathB.cubicTo(2,3, 1,0, 4,1);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kUnion_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp29d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(2,5, 6,0, 4,2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,6);
pathB.cubicTo(2,4, 1,0, 5,2);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp30d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(2,5, 6,0, 5,3);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,6);
pathB.cubicTo(3,5, 1,0, 5,2);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp31d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,2);
path.cubicTo(0,3, 2,1, 4,0);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1,2);
pathB.cubicTo(0,4, 2,0, 3,0);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp31u(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,2);
path.cubicTo(0,3, 2,1, 4,0);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1,2);
pathB.cubicTo(0,4, 2,0, 3,0);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kUnion_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp31x(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,2);
path.cubicTo(0,3, 2,1, 4,0);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1,2);
pathB.cubicTo(0,4, 2,0, 3,0);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kXOR_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp32d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(1,2, 6,0, 3,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,6);
pathB.cubicTo(1,3, 1,0, 2,1);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp33i(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(1,2, 6,0, 3,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,6);
pathB.cubicTo(1,3, 1,0, 2,1);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp34d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(3,5, 2,1, 3,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1,2);
pathB.cubicTo(1,3, 1,0, 5,3);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp35d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(1,5, 2,1, 4,0);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1,2);
pathB.cubicTo(0,4, 1,0, 5,1);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp36u(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(1,6, 2,0, 5,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,2);
pathB.cubicTo(1,5, 1,0, 6,1);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kUnion_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp37d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(2,6, 6,1, 4,3);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1,6);
pathB.cubicTo(3,4, 1,0, 6,2);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp38d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(0,6, 3,2, 4,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(2,3);
pathB.cubicTo(1,4, 1,0, 6,0);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp39d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(2,3, 5,1, 4,3);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1,5);
pathB.cubicTo(3,4, 1,0, 3,2);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp40d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(1,5, 3,2, 4,2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(2,3);
pathB.cubicTo(2,4, 1,0, 5,1);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp41i(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(2,6, 4,3, 6,4);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(3,4);
pathB.cubicTo(4,6, 1,0, 6,2);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp42d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(1,2, 6,5, 5,4);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(5,6);
pathB.cubicTo(4,5, 1,0, 2,1);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp43d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,2);
path.cubicTo(1,2, 4,0, 3,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,4);
pathB.cubicTo(1,3, 2,0, 2,1);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp44d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,2);
path.cubicTo(3,6, 4,0, 3,2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,4);
pathB.cubicTo(2,3, 2,0, 6,3);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp45d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,2);
path.cubicTo(2,4, 4,0, 3,2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,4);
pathB.cubicTo(2,3, 2,0, 4,2);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp46d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,2);
path.cubicTo(3,5, 5,0, 4,2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,5);
pathB.cubicTo(2,4, 2,0, 5,3);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp47d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(1,6, 6,2, 5,4);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(2,6);
pathB.cubicTo(4,5, 1,0, 6,1);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp48d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,2);
path.cubicTo(2,3, 5,1, 3,2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1,5);
pathB.cubicTo(2,3, 2,0, 3,2);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp49d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,2);
path.cubicTo(1,5, 3,2, 4,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(2,3);
pathB.cubicTo(1,4, 2,0, 5,1);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp50d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,3);
path.cubicTo(1,6, 5,0, 5,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,5);
pathB.cubicTo(1,5, 3,0, 6,1);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp51d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,3);
path.cubicTo(1,2, 4,1, 6,0);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1,4);
pathB.cubicTo(0,6, 3,0, 2,1);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp52d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,2);
path.cubicTo(1,2, 5,4, 4,3);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(4,5);
pathB.cubicTo(3,4, 2,0, 2,1);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp53d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,3);
path.cubicTo(1,2, 5,3, 2,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(3,5);
pathB.cubicTo(1,2, 3,0, 2,1);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp54d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,4);
path.cubicTo(1,3, 5,4, 4,2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(4,5);
pathB.cubicTo(2,4, 4,0, 3,1);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp55d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,5);
path.cubicTo(1,3, 3,2, 5,0);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(2,3);
pathB.cubicTo(0,5, 5,0, 3,1);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp56d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(2,6, 5,0, 2,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,5);
pathB.cubicTo(1,2, 1,0, 6,2);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp57d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,5);
path.cubicTo(0,5, 5,4, 6,4);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(4,5);
pathB.cubicTo(4,6, 5,0, 5,0);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp58d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,5);
path.cubicTo(3,4, 6,5, 5,3);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(5,6);
pathB.cubicTo(3,5, 5,0, 4,3);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp59d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(5,6, 4,0, 4,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,4);
pathB.cubicTo(1,4, 1,0, 6,5);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp60d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,2);
path.cubicTo(4,6, 6,0, 5,2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,6);
pathB.cubicTo(2,5, 2,0, 6,4);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp61d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(1,2);
path.cubicTo(0,5, 3,2, 6,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(2,3);
pathB.cubicTo(1,6, 2,1, 5,0);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp62d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(1,3);
path.cubicTo(5,6, 5,3, 5,4);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(3,5);
pathB.cubicTo(4,5, 3,1, 6,5);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp63d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(2,3);
path.cubicTo(0,4, 3,2, 5,3);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(2,3);
pathB.cubicTo(3,5, 3,2, 4,0);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp64d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.moveTo(0,1);
path.cubicTo(0,1, 1,0, 3,0);
path.lineTo(0,1);
path.close();
pathB.moveTo(0,1);
pathB.cubicTo(0,3, 1,0, 1,0);
pathB.lineTo(0,1);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp65d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.moveTo(0,1);
path.cubicTo(1,5, 1,0, 1,0);
path.lineTo(0,1);
path.close();
pathB.moveTo(0,1);
pathB.cubicTo(0,1, 1,0, 5,1);
pathB.lineTo(0,1);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void rectOp1d(skiatest::Reporter* reporter, const char* filename) {
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
SkPath path, pathB;
path.moveTo(0,1);
path.cubicTo(0,1, 1,0, 3,0);
path.lineTo(0,1);
path.close();
pathB.moveTo(0,1);
pathB.cubicTo(0,3, 1,0, 1,0);
pathB.lineTo(0,1);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp66u(skiatest::Reporter* reporter, const char* filename) {
shape ops work in progress Try to fix the 32 bit build by making some math decisions more robust. Rewrite the cubic intersection special case that detects if only end points are shared. Rewrite the angle sort setup that computes whether a cubic bends to the left or right. git-svn-id: http://skia.googlecode.com/svn/trunk@8726 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-17 15:49:16 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(2,6, 4,2, 5,3);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(2,4);
pathB.cubicTo(3,5, 1,0, 6,2);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kUnion_SkPathOp, filename);
shape ops work in progress Try to fix the 32 bit build by making some math decisions more robust. Rewrite the cubic intersection special case that detects if only end points are shared. Rewrite the angle sort setup that computes whether a cubic bends to the left or right. git-svn-id: http://skia.googlecode.com/svn/trunk@8726 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-17 15:49:16 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp67u(skiatest::Reporter* reporter, const char* filename) {
shape ops work in progress Try to fix the 32 bit build by making some math decisions more robust. Rewrite the cubic intersection special case that detects if only end points are shared. Rewrite the angle sort setup that computes whether a cubic bends to the left or right. git-svn-id: http://skia.googlecode.com/svn/trunk@8726 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-17 15:49:16 +00:00
SkPath path, pathB;
path.moveTo(3,5);
path.cubicTo(1,6, 5,0, 3,1);
path.lineTo(3,5);
path.close();
pathB.moveTo(0,5);
pathB.cubicTo(1,3, 5,3, 6,1);
pathB.lineTo(0,5);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kUnion_SkPathOp, filename);
shape ops work in progress Try to fix the 32 bit build by making some math decisions more robust. Rewrite the cubic intersection special case that detects if only end points are shared. Rewrite the angle sort setup that computes whether a cubic bends to the left or right. git-svn-id: http://skia.googlecode.com/svn/trunk@8726 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-17 15:49:16 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp68u(skiatest::Reporter* reporter, const char* filename) {
path ops work in progress fix bugs in tests on 32 bit release Most changes revolve around pinning computed t values very close to zero and one. git-svn-id: http://skia.googlecode.com/svn/trunk@8745 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-18 15:58:21 +00:00
SkPath path, pathB;
path.moveTo(0,5);
path.cubicTo(4,5, 4,1, 5,0);
path.close();
pathB.moveTo(1,4);
pathB.cubicTo(0,5, 5,0, 5,4);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kUnion_SkPathOp, filename);
path ops work in progress fix bugs in tests on 32 bit release Most changes revolve around pinning computed t values very close to zero and one. git-svn-id: http://skia.googlecode.com/svn/trunk@8745 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-18 15:58:21 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp69d(skiatest::Reporter* reporter, const char* filename) {
path ops work in progress fix bugs in tests on 32 bit release Most changes revolve around pinning computed t values very close to zero and one. git-svn-id: http://skia.googlecode.com/svn/trunk@8745 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-18 15:58:21 +00:00
SkPath path, pathB;
path.moveTo(1,3);
path.cubicTo(0,1, 3,1, 2,0);
path.close();
pathB.moveTo(1,3);
pathB.cubicTo(0,2, 3,1, 1,0);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
path ops work in progress fix bugs in tests on 32 bit release Most changes revolve around pinning computed t values very close to zero and one. git-svn-id: http://skia.googlecode.com/svn/trunk@8745 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-18 15:58:21 +00:00
}
path ops : fix empty-diff bug, op-in-place add some debugging around reverse diff, inverse Review URL: https://codereview.chromium.org/13851015 git-svn-id: http://skia.googlecode.com/svn/trunk@8852 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-25 11:51:54 +00:00
SkPathOp ops[] = {
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
kUnion_SkPathOp,
kXOR_SkPathOp,
kReverseDifference_SkPathOp,
kXOR_SkPathOp,
kReverseDifference_SkPathOp,
path ops : fix empty-diff bug, op-in-place add some debugging around reverse diff, inverse Review URL: https://codereview.chromium.org/13851015 git-svn-id: http://skia.googlecode.com/svn/trunk@8852 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-25 11:51:54 +00:00
};
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void rRect1(skiatest::Reporter* reporter, const char* filename) {
remove SkFloatToScalar macro BUG= R=reed@google.com, djsollen@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/85463005 git-svn-id: http://skia.googlecode.com/svn/trunk@12385 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-11-25 19:44:07 +00:00
SkScalar xA = 0.65f;
SkScalar xB = 10.65f;
SkScalar xC = 20.65f;
SkScalar xD = 30.65f;
SkScalar xE = 40.65f;
SkScalar xF = 50.65f;
SkScalar yA = 0.65f;
SkScalar yB = 10.65f;
SkScalar yC = 20.65f;
SkScalar yD = 30.65f;
SkScalar yE = 40.65f;
SkScalar yF = 50.65f;
path ops -- fix skp bugs This fixes a series of bugs discovered by running the small set of Skia skp files through pathops to flatten the clips. Review URL: https://codereview.chromium.org/14798004 git-svn-id: http://skia.googlecode.com/svn/trunk@9042 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-07 18:51:31 +00:00
SkPath paths[5];
SkRect rects[5];
rects[0].set(xB, yB, xE, yE);
paths[0].addRoundRect(rects[0], SkIntToScalar(5), SkIntToScalar(5)); // red
rects[1].set(xA, yA, xD, yD);
paths[1].addRoundRect(rects[1], SkIntToScalar(5), SkIntToScalar(5)); // green
rects[2].set(xC, yA, xF, yD);
paths[2].addRoundRect(rects[2], SkIntToScalar(5), SkIntToScalar(5)); // blue
rects[3].set(xA, yC, xD, yF);
paths[3].addRoundRect(rects[3], SkIntToScalar(5), SkIntToScalar(5)); // yellow
rects[4].set(xC, yC, xF, yF);
paths[4].addRoundRect(rects[4], SkIntToScalar(5), SkIntToScalar(5)); // cyan
SkPath path;
path.setFillType(SkPath::kInverseEvenOdd_FillType);
for (int index = 0; index < 5; ++index) {
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
testPathOp(reporter, path, paths[index], ops[index], filename);
pathops coincidence and security rewrite Most changes stem from working on an examples bracketed by #if DEBUG_UNDER_DEVELOPMENT // tiger These exposed many problems with coincident curves, as well as errors throughout the code. Fixing these errors also fixed a number of fuzzer-inspired bug reports. * Line/Curve Intersections Check to see if the end of the line nearly intersects the curve. This was a FIXME in the old code. * Performance Use a central chunk allocator. Plumb the allocator into the global variable state so that it can be shared. (Note that 'SkGlobalState' is allocated on the stack and is visible to children functions but not other threads.) * Refactor Let SkOpAngle grow up from a structure to a class. Let SkCoincidentSpans grow up from a structure to a class. Rename enum Alias to AliasMatch. * Coincidence Rewrite Add more debugging to coincidence detection. Parallel debugging routines have read-only logic to report the current coincidence state so that steps through the logic can expose whether things got better or worse. More functions can error-out and cause the pathops engine to non-destructively exit. * Accuracy Remove code that adjusted point locations. Instead, offset the curve part so that sorted curves all use the same origin. Reduce the size (and influence) of magic numbers. * Testing The debug suite with verify and the full release suite ./out/Debug/pathops_unittest -v -V ./out/Release/pathops_unittest -v -V -x expose one error. That error is captured as cubics_d3. This error exists in the checked in code as well. BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 Review-Url: https://codereview.chromium.org/2128633003
2016-07-18 17:01:36 +00:00
REPORTER_ASSERT(reporter, Op(path, paths[index], ops[index], &path));
path ops -- fix skp bugs This fixes a series of bugs discovered by running the small set of Skia skp files through pathops to flatten the clips. Review URL: https://codereview.chromium.org/14798004 git-svn-id: http://skia.googlecode.com/svn/trunk@9042 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-07 18:51:31 +00:00
}
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void skp1(skiatest::Reporter* reporter, const char* filename) {
path ops -- fix skp bugs This fixes a series of bugs discovered by running the small set of Skia skp files through pathops to flatten the clips. Review URL: https://codereview.chromium.org/14798004 git-svn-id: http://skia.googlecode.com/svn/trunk@9042 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-07 18:51:31 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(189,7);
path.cubicTo(189,5.34314585f, 190.34314f,4, 192,4);
path.lineTo(243,4);
path.cubicTo(244.65686f,4, 246,5.34314585f, 246,7);
path.lineTo(246,21);
path.cubicTo(246,22.6568546f, 244.65686f,24, 243,24);
path.lineTo(192,24);
path.cubicTo(190.34314f,24, 189,22.6568546f, 189,21);
path.lineTo(189,7);
path.close();
path.moveTo(191,8);
path.cubicTo(191,6.89543009f, 191.895432f,6, 193,6);
path.lineTo(242,6);
path.cubicTo(243.104568f,6, 244,6.89543009f, 244,8);
path.lineTo(244,20);
path.cubicTo(244,21.1045704f, 243.104568f,22, 242,22);
path.lineTo(193,22);
path.cubicTo(191.895432f,22, 191,21.1045704f, 191,20);
path.lineTo(191,8);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(189,4);
pathB.lineTo(199,14);
pathB.lineTo(236,14);
pathB.lineTo(246,4);
pathB.lineTo(189,4);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
path ops -- fix skp bugs This fixes a series of bugs discovered by running the small set of Skia skp files through pathops to flatten the clips. Review URL: https://codereview.chromium.org/14798004 git-svn-id: http://skia.googlecode.com/svn/trunk@9042 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-07 18:51:31 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void skp2(skiatest::Reporter* reporter, const char* filename) {
path ops -- fix skp bugs This fixes a series of bugs discovered by running the small set of Skia skp files through pathops to flatten the clips. Review URL: https://codereview.chromium.org/14798004 git-svn-id: http://skia.googlecode.com/svn/trunk@9042 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-07 18:51:31 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(253.000000f, 11757.0000f);
path.lineTo(253.000000f, 222.000000f);
path.lineTo(823.000000f, 222.000000f);
path.lineTo(823.000000f, 11757.0000f);
path.lineTo(253.000000f, 11757.0000f);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(258.000000f, 1028.00000f);
pathB.lineTo(258.000000f, 1027.00000f);
pathB.lineTo(823.000000f, 1027.00000f);
pathB.lineTo(823.000000f, 1028.00000f);
pathB.lineTo(258.000000f, 1028.00000f);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
path ops -- fix skp bugs This fixes a series of bugs discovered by running the small set of Skia skp files through pathops to flatten the clips. Review URL: https://codereview.chromium.org/14798004 git-svn-id: http://skia.googlecode.com/svn/trunk@9042 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-07 18:51:31 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void skp3(skiatest::Reporter* reporter, const char* filename) {
path ops -- fix skp bugs This fixes a series of bugs discovered by running the small set of Skia skp files through pathops to flatten the clips. Review URL: https://codereview.chromium.org/14798004 git-svn-id: http://skia.googlecode.com/svn/trunk@9042 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-07 18:51:31 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(717.000000f, 507.000000f);
path.lineTo(717.000000f, 425.000000f);
path.lineTo(973.000000f, 425.000000f);
path.lineTo(973.000000f, 507.000000f);
path.quadTo(973.000000f, 508.242645f, 972.121582f, 509.121613f);
path.quadTo(971.242615f, 510.000000f, 970.000000f, 510.000000f);
path.lineTo(720.000000f, 510.000000f);
path.quadTo(718.757385f, 510.000000f, 717.878418f, 509.121613f);
path.quadTo(717.000000f, 508.242645f, 717.000000f, 507.000000f);
path.close();
path.moveTo(719.000000f, 426.000000f);
path.lineTo(971.000000f, 426.000000f);
path.lineTo(971.000000f, 506.000000f);
path.cubicTo(971.000000f, 507.104584f, 970.104553f, 508.000000f, 969.000000f, 508.000000f);
path.lineTo(721.000000f, 508.000000f);
path.cubicTo(719.895447f, 508.000000f, 719.000000f, 507.104584f, 719.000000f, 506.000000f);
path.lineTo(719.000000f, 426.000000f);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(717.000000f, 510.000000f);
pathB.lineTo(760.000000f, 467.000000f);
pathB.lineTo(930.000000f, 467.000000f);
pathB.lineTo(973.000000f, 510.000000f);
pathB.lineTo(717.000000f, 510.000000f);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
path ops -- fix skp bugs This fixes a series of bugs discovered by running the small set of Skia skp files through pathops to flatten the clips. Review URL: https://codereview.chromium.org/14798004 git-svn-id: http://skia.googlecode.com/svn/trunk@9042 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-07 18:51:31 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void skp4(skiatest::Reporter* reporter, const char* filename) {
path ops -- fix skp bugs This fixes a series of bugs discovered by running the small set of Skia skp files through pathops to flatten the clips. Review URL: https://codereview.chromium.org/14798004 git-svn-id: http://skia.googlecode.com/svn/trunk@9042 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-07 18:51:31 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(230.756805f, 591.756775f);
path.quadTo(232.514725f, 590.000000f, 235.000000f, 590.000000f);
path.lineTo(300.000000f, 590.000000f);
path.quadTo(302.485291f, 590.000000f, 304.243195f, 591.756775f);
path.quadTo(306.000000f, 593.514709f, 306.000000f, 596.000000f);
path.lineTo(306.000000f, 617.000000f);
path.lineTo(229.000000f, 617.000000f);
path.lineTo(229.000000f, 596.000000f);
path.quadTo(229.000000f, 593.514709f, 230.756805f, 591.756775f);
path.close();
path.moveTo(231.000000f, 597.000000f);
path.cubicTo(231.000000f, 594.238586f, 233.238571f, 592.000000f, 236.000000f, 592.000000f);
path.lineTo(299.000000f, 592.000000f);
path.cubicTo(301.761414f, 592.000000f, 304.000000f, 594.238586f, 304.000000f, 597.000000f);
path.lineTo(304.000000f, 616.000000f);
path.lineTo(231.000000f, 616.000000f);
path.lineTo(231.000000f, 597.000000f);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(306.000000f, 590.000000f);
pathB.lineTo(292.000000f, 604.000000f);
pathB.lineTo(305.000000f, 617.000000f);
pathB.lineTo(306.000000f, 617.000000f);
pathB.lineTo(306.000000f, 590.000000f);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
path ops -- fix skp bugs This fixes a series of bugs discovered by running the small set of Skia skp files through pathops to flatten the clips. Review URL: https://codereview.chromium.org/14798004 git-svn-id: http://skia.googlecode.com/svn/trunk@9042 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-07 18:51:31 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void skp5(skiatest::Reporter* reporter, const char* filename) {
path ops -- fix skp bugs This fixes a series of bugs discovered by running the small set of Skia skp files through pathops to flatten the clips. Review URL: https://codereview.chromium.org/14798004 git-svn-id: http://skia.googlecode.com/svn/trunk@9042 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-07 18:51:31 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(18.0000000f, 226.000000f);
path.quadTo(14.6862917f, 226.000000f, 12.3423996f, 228.342407f);
path.quadTo(10.0000000f, 230.686295f, 10.0000000f, 234.000000f);
path.lineTo(10.0000000f, 253.000000f);
path.lineTo(1247.00000f, 253.000000f);
path.lineTo(1247.00000f, 234.000000f);
path.quadTo(1247.00000f, 230.686295f, 1244.65759f, 228.342407f);
path.quadTo(1242.31372f, 226.000000f, 1239.00000f, 226.000000f);
path.lineTo(18.0000000f, 226.000000f);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kInverseWinding_FillType);
pathB.moveTo(18.0000000f, 226.000000f);
pathB.lineTo(1239.00000f, 226.000000f);
pathB.cubicTo(1243.41833f, 226.000000f, 1247.00000f, 229.581726f, 1247.00000f, 234.000000f);
pathB.lineTo(1247.00000f, 252.000000f);
pathB.lineTo(10.0000000f, 252.000000f);
pathB.lineTo(10.0000000f, 234.000000f);
pathB.cubicTo(10.0000000f, 229.581726f, 13.5817204f, 226.000000f, 18.0000000f, 226.000000f);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
path ops -- fix skp bugs This fixes a series of bugs discovered by running the small set of Skia skp files through pathops to flatten the clips. Review URL: https://codereview.chromium.org/14798004 git-svn-id: http://skia.googlecode.com/svn/trunk@9042 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-07 18:51:31 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp70d(skiatest::Reporter* reporter, const char* filename) {
path ops -- fix skp bugs This fixes a series of bugs discovered by running the small set of Skia skp files through pathops to flatten the clips. Review URL: https://codereview.chromium.org/14798004 git-svn-id: http://skia.googlecode.com/svn/trunk@9042 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-07 18:51:31 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(0,5, 4,0, 5,0);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,4);
pathB.cubicTo(0,5, 1,0, 5,0);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
path ops : fix empty-diff bug, op-in-place add some debugging around reverse diff, inverse Review URL: https://codereview.chromium.org/13851015 git-svn-id: http://skia.googlecode.com/svn/trunk@8852 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-25 11:51:54 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp71d(skiatest::Reporter* reporter, const char* filename) {
bump picture version since SkPath has changed (conics) enable conics in SkPath git-svn-id: http://skia.googlecode.com/svn/trunk@9370 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-31 15:17:50 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(0,5, 4,1, 6,4);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1,4);
pathB.cubicTo(4,6, 1,0, 5,0);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
bump picture version since SkPath has changed (conics) enable conics in SkPath git-svn-id: http://skia.googlecode.com/svn/trunk@9370 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-31 15:17:50 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp72i(skiatest::Reporter* reporter, const char* filename) {
bump picture version since SkPath has changed (conics) enable conics in SkPath git-svn-id: http://skia.googlecode.com/svn/trunk@9370 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-31 15:17:50 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(0,5, 5,2, 5,4);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(2,5);
pathB.cubicTo(4,5, 1,0, 5,0);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
bump picture version since SkPath has changed (conics) enable conics in SkPath git-svn-id: http://skia.googlecode.com/svn/trunk@9370 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-31 15:17:50 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp73d(skiatest::Reporter* reporter, const char* filename) {
bump picture version since SkPath has changed (conics) enable conics in SkPath git-svn-id: http://skia.googlecode.com/svn/trunk@9370 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-31 15:17:50 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(3,4, 4,0, 6,4);
path.lineTo(0,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,4);
pathB.cubicTo(4,6, 1,0, 4,3);
pathB.lineTo(0,4);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
bump picture version since SkPath has changed (conics) enable conics in SkPath git-svn-id: http://skia.googlecode.com/svn/trunk@9370 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-31 15:17:50 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp74d(skiatest::Reporter* reporter, const char* filename) {
path ops -- rewrite angle sort This is a major change resulting from a minor tweak. In the old code, the intersection point of two curves was shared between them, but the intersection points and end points of sorted edges was computed directly from the intersection T value. In this CL, both intersection points and sorted points are the same, and intermediate control points are computed to preserve their slope. The sort itself has been completely rewritten to be more robust and remove 'magic' checks, conditions that empirically worked but couldn't be rationalized. This CL was triggered by errors generated computing the clips of SKP files. At this point, all 73M standard tests work and at least the first troublesome SKPs work. Review URL: https://codereview.chromium.org/15338003 git-svn-id: http://skia.googlecode.com/svn/trunk@9432 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-06-04 17:59:42 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(1,5, 5,1, 5,1);
path.lineTo(0,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1,5);
pathB.cubicTo(1,5, 1,0, 5,1);
pathB.lineTo(1,5);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
path ops -- rewrite angle sort This is a major change resulting from a minor tweak. In the old code, the intersection point of two curves was shared between them, but the intersection points and end points of sorted edges was computed directly from the intersection T value. In this CL, both intersection points and sorted points are the same, and intermediate control points are computed to preserve their slope. The sort itself has been completely rewritten to be more robust and remove 'magic' checks, conditions that empirically worked but couldn't be rationalized. This CL was triggered by errors generated computing the clips of SKP files. At this point, all 73M standard tests work and at least the first troublesome SKPs work. Review URL: https://codereview.chromium.org/15338003 git-svn-id: http://skia.googlecode.com/svn/trunk@9432 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-06-04 17:59:42 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp75d(skiatest::Reporter* reporter, const char* filename) {
bump picture version since SkPath has changed (conics) enable conics in SkPath git-svn-id: http://skia.googlecode.com/svn/trunk@9370 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-31 15:17:50 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(0,4, 5,1, 6,4);
path.lineTo(0,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1,5);
pathB.cubicTo(4,6, 1,0, 4,0);
pathB.lineTo(1,5);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
bump picture version since SkPath has changed (conics) enable conics in SkPath git-svn-id: http://skia.googlecode.com/svn/trunk@9370 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-31 15:17:50 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp76u(skiatest::Reporter* reporter, const char* filename) {
path ops -- rewrite angle sort This is a major change resulting from a minor tweak. In the old code, the intersection point of two curves was shared between them, but the intersection points and end points of sorted edges was computed directly from the intersection T value. In this CL, both intersection points and sorted points are the same, and intermediate control points are computed to preserve their slope. The sort itself has been completely rewritten to be more robust and remove 'magic' checks, conditions that empirically worked but couldn't be rationalized. This CL was triggered by errors generated computing the clips of SKP files. At this point, all 73M standard tests work and at least the first troublesome SKPs work. Review URL: https://codereview.chromium.org/15338003 git-svn-id: http://skia.googlecode.com/svn/trunk@9432 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-06-04 17:59:42 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(0,2, 2,0, 5,3);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,2);
pathB.cubicTo(3,5, 1,0, 2,0);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kUnion_SkPathOp, filename);
path ops -- rewrite angle sort This is a major change resulting from a minor tweak. In the old code, the intersection point of two curves was shared between them, but the intersection points and end points of sorted edges was computed directly from the intersection T value. In this CL, both intersection points and sorted points are the same, and intermediate control points are computed to preserve their slope. The sort itself has been completely rewritten to be more robust and remove 'magic' checks, conditions that empirically worked but couldn't be rationalized. This CL was triggered by errors generated computing the clips of SKP files. At this point, all 73M standard tests work and at least the first troublesome SKPs work. Review URL: https://codereview.chromium.org/15338003 git-svn-id: http://skia.googlecode.com/svn/trunk@9432 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-06-04 17:59:42 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp77i(skiatest::Reporter* reporter, const char* filename) {
bump picture version since SkPath has changed (conics) enable conics in SkPath git-svn-id: http://skia.googlecode.com/svn/trunk@9370 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-31 15:17:50 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(0,1);
path.cubicTo(1,3, 2,0, 3,2);
path.lineTo(0,1);
path.close();
pathB.setFillType(SkPath::kEvenOdd_FillType);
pathB.moveTo(0,2);
pathB.cubicTo(2,3, 1,0, 3,1);
pathB.lineTo(0,2);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
bump picture version since SkPath has changed (conics) enable conics in SkPath git-svn-id: http://skia.googlecode.com/svn/trunk@9370 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-31 15:17:50 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp78u(skiatest::Reporter* reporter, const char* filename) {
bump picture version since SkPath has changed (conics) enable conics in SkPath git-svn-id: http://skia.googlecode.com/svn/trunk@9370 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-31 15:17:50 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(1,6);
path.cubicTo(1,6, 5,0, 6,1);
path.lineTo(1,6);
path.close();
pathB.setFillType(SkPath::kEvenOdd_FillType);
pathB.moveTo(0,5);
pathB.cubicTo(1,6, 6,1, 6,1);
pathB.lineTo(0,5);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kUnion_SkPathOp, filename);
bump picture version since SkPath has changed (conics) enable conics in SkPath git-svn-id: http://skia.googlecode.com/svn/trunk@9370 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-31 15:17:50 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp79u(skiatest::Reporter* reporter, const char* filename) {
bump picture version since SkPath has changed (conics) enable conics in SkPath git-svn-id: http://skia.googlecode.com/svn/trunk@9370 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-31 15:17:50 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(1,3, 1,0, 6,4);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,1);
pathB.cubicTo(4,6, 1,0, 3,1);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
bump picture version since SkPath has changed (conics) enable conics in SkPath git-svn-id: http://skia.googlecode.com/svn/trunk@9370 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-31 15:17:50 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp80i(skiatest::Reporter* reporter, const char* filename) {
bump picture version since SkPath has changed (conics) enable conics in SkPath git-svn-id: http://skia.googlecode.com/svn/trunk@9370 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-31 15:17:50 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(2,3, 2,1, 4,3);
path.lineTo(0,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1,2);
pathB.cubicTo(3,4, 1,0, 3,2);
pathB.lineTo(1,2);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
bump picture version since SkPath has changed (conics) enable conics in SkPath git-svn-id: http://skia.googlecode.com/svn/trunk@9370 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-31 15:17:50 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp81d(skiatest::Reporter* reporter, const char* filename) {
path ops -- rewrite angle sort This is a major change resulting from a minor tweak. In the old code, the intersection point of two curves was shared between them, but the intersection points and end points of sorted edges was computed directly from the intersection T value. In this CL, both intersection points and sorted points are the same, and intermediate control points are computed to preserve their slope. The sort itself has been completely rewritten to be more robust and remove 'magic' checks, conditions that empirically worked but couldn't be rationalized. This CL was triggered by errors generated computing the clips of SKP files. At this point, all 73M standard tests work and at least the first troublesome SKPs work. Review URL: https://codereview.chromium.org/15338003 git-svn-id: http://skia.googlecode.com/svn/trunk@9432 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-06-04 17:59:42 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(4,6, 4,3, 5,4);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(3,4);
pathB.cubicTo(4,5, 1,0, 6,4);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
path ops -- rewrite angle sort This is a major change resulting from a minor tweak. In the old code, the intersection point of two curves was shared between them, but the intersection points and end points of sorted edges was computed directly from the intersection T value. In this CL, both intersection points and sorted points are the same, and intermediate control points are computed to preserve their slope. The sort itself has been completely rewritten to be more robust and remove 'magic' checks, conditions that empirically worked but couldn't be rationalized. This CL was triggered by errors generated computing the clips of SKP files. At this point, all 73M standard tests work and at least the first troublesome SKPs work. Review URL: https://codereview.chromium.org/15338003 git-svn-id: http://skia.googlecode.com/svn/trunk@9432 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-06-04 17:59:42 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp82i(skiatest::Reporter* reporter, const char* filename) {
bump picture version since SkPath has changed (conics) enable conics in SkPath git-svn-id: http://skia.googlecode.com/svn/trunk@9370 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-31 15:17:50 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(0,1);
path.cubicTo(2,3, 5,2, 3,0);
path.lineTo(0,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(2,5);
pathB.cubicTo(0,3, 1,0, 3,2);
pathB.lineTo(2,5);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
bump picture version since SkPath has changed (conics) enable conics in SkPath git-svn-id: http://skia.googlecode.com/svn/trunk@9370 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-31 15:17:50 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp83i(skiatest::Reporter* reporter, const char* filename) {
bump picture version since SkPath has changed (conics) enable conics in SkPath git-svn-id: http://skia.googlecode.com/svn/trunk@9370 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-31 15:17:50 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path ops -- rewrite angle sort This is a major change resulting from a minor tweak. In the old code, the intersection point of two curves was shared between them, but the intersection points and end points of sorted edges was computed directly from the intersection T value. In this CL, both intersection points and sorted points are the same, and intermediate control points are computed to preserve their slope. The sort itself has been completely rewritten to be more robust and remove 'magic' checks, conditions that empirically worked but couldn't be rationalized. This CL was triggered by errors generated computing the clips of SKP files. At this point, all 73M standard tests work and at least the first troublesome SKPs work. Review URL: https://codereview.chromium.org/15338003 git-svn-id: http://skia.googlecode.com/svn/trunk@9432 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-06-04 17:59:42 +00:00
path.cubicTo(0,3, 2,1, 4,1);
path.lineTo(0,1);
bump picture version since SkPath has changed (conics) enable conics in SkPath git-svn-id: http://skia.googlecode.com/svn/trunk@9370 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-31 15:17:50 +00:00
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
path ops -- rewrite angle sort This is a major change resulting from a minor tweak. In the old code, the intersection point of two curves was shared between them, but the intersection points and end points of sorted edges was computed directly from the intersection T value. In this CL, both intersection points and sorted points are the same, and intermediate control points are computed to preserve their slope. The sort itself has been completely rewritten to be more robust and remove 'magic' checks, conditions that empirically worked but couldn't be rationalized. This CL was triggered by errors generated computing the clips of SKP files. At this point, all 73M standard tests work and at least the first troublesome SKPs work. Review URL: https://codereview.chromium.org/15338003 git-svn-id: http://skia.googlecode.com/svn/trunk@9432 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-06-04 17:59:42 +00:00
pathB.moveTo(1,2);
pathB.cubicTo(1,4, 1,0, 3,0);
pathB.lineTo(1,2);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
path ops -- rewrite angle sort This is a major change resulting from a minor tweak. In the old code, the intersection point of two curves was shared between them, but the intersection points and end points of sorted edges was computed directly from the intersection T value. In this CL, both intersection points and sorted points are the same, and intermediate control points are computed to preserve their slope. The sort itself has been completely rewritten to be more robust and remove 'magic' checks, conditions that empirically worked but couldn't be rationalized. This CL was triggered by errors generated computing the clips of SKP files. At this point, all 73M standard tests work and at least the first troublesome SKPs work. Review URL: https://codereview.chromium.org/15338003 git-svn-id: http://skia.googlecode.com/svn/trunk@9432 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-06-04 17:59:42 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp84d(skiatest::Reporter* reporter, const char* filename) {
path ops -- rewrite angle sort This is a major change resulting from a minor tweak. In the old code, the intersection point of two curves was shared between them, but the intersection points and end points of sorted edges was computed directly from the intersection T value. In this CL, both intersection points and sorted points are the same, and intermediate control points are computed to preserve their slope. The sort itself has been completely rewritten to be more robust and remove 'magic' checks, conditions that empirically worked but couldn't be rationalized. This CL was triggered by errors generated computing the clips of SKP files. At this point, all 73M standard tests work and at least the first troublesome SKPs work. Review URL: https://codereview.chromium.org/15338003 git-svn-id: http://skia.googlecode.com/svn/trunk@9432 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-06-04 17:59:42 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,4);
path.cubicTo(2,3, 6,3, 3,2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(3,6);
pathB.cubicTo(2,3, 4,0, 3,2);
bump picture version since SkPath has changed (conics) enable conics in SkPath git-svn-id: http://skia.googlecode.com/svn/trunk@9370 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-31 15:17:50 +00:00
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
bump picture version since SkPath has changed (conics) enable conics in SkPath git-svn-id: http://skia.googlecode.com/svn/trunk@9370 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-31 15:17:50 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void skpClip1(skiatest::Reporter* reporter, const char* filename) {
path ops -- rewrite angle sort This is a major change resulting from a minor tweak. In the old code, the intersection point of two curves was shared between them, but the intersection points and end points of sorted edges was computed directly from the intersection T value. In this CL, both intersection points and sorted points are the same, and intermediate control points are computed to preserve their slope. The sort itself has been completely rewritten to be more robust and remove 'magic' checks, conditions that empirically worked but couldn't be rationalized. This CL was triggered by errors generated computing the clips of SKP files. At this point, all 73M standard tests work and at least the first troublesome SKPs work. Review URL: https://codereview.chromium.org/15338003 git-svn-id: http://skia.googlecode.com/svn/trunk@9432 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-06-04 17:59:42 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(1126.17114f, 877.171204f);
path.quadTo(1127.34314f, 876.000000f, 1129.00000f, 876.000000f);
path.lineTo(1243.00000f, 876.000000f);
path.quadTo(1244.65686f, 876.000000f, 1245.82886f, 877.171204f);
path.quadTo(1247.00000f, 878.343140f, 1247.00000f, 880.000000f);
path.lineTo(1247.00000f, 907.000000f);
path.lineTo(1246.00000f, 907.000000f);
path.lineTo(1246.00000f, 880.000000f);
path.cubicTo(1246.00000f, 878.343140f, 1244.65686f, 877.000000f, 1243.00000f, 877.000000f);
path.lineTo(1129.00000f, 877.000000f);
path.cubicTo(1127.34314f, 877.000000f, 1126.00000f, 878.343140f, 1126.00000f, 880.000000f);
path.lineTo(1126.00000f, 907.000000f);
path.lineTo(1125.00000f, 907.000000f);
path.lineTo(1125.00000f, 880.000000f);
path.quadTo(1125.00000f, 878.343140f, 1126.17114f, 877.171204f);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1247.00000f, 876.000000f);
pathB.lineTo(1231.00000f, 892.000000f);
pathB.lineTo(1246.00000f, 907.000000f);
pathB.lineTo(1247.00000f, 907.000000f);
pathB.lineTo(1247.00000f, 876.000000f);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
path ops -- rewrite angle sort This is a major change resulting from a minor tweak. In the old code, the intersection point of two curves was shared between them, but the intersection points and end points of sorted edges was computed directly from the intersection T value. In this CL, both intersection points and sorted points are the same, and intermediate control points are computed to preserve their slope. The sort itself has been completely rewritten to be more robust and remove 'magic' checks, conditions that empirically worked but couldn't be rationalized. This CL was triggered by errors generated computing the clips of SKP files. At this point, all 73M standard tests work and at least the first troublesome SKPs work. Review URL: https://codereview.chromium.org/15338003 git-svn-id: http://skia.googlecode.com/svn/trunk@9432 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-06-04 17:59:42 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void skpClip2(skiatest::Reporter* reporter, const char* filename) {
path ops -- rewrite angle sort This is a major change resulting from a minor tweak. In the old code, the intersection point of two curves was shared between them, but the intersection points and end points of sorted edges was computed directly from the intersection T value. In this CL, both intersection points and sorted points are the same, and intermediate control points are computed to preserve their slope. The sort itself has been completely rewritten to be more robust and remove 'magic' checks, conditions that empirically worked but couldn't be rationalized. This CL was triggered by errors generated computing the clips of SKP files. At this point, all 73M standard tests work and at least the first troublesome SKPs work. Review URL: https://codereview.chromium.org/15338003 git-svn-id: http://skia.googlecode.com/svn/trunk@9432 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-06-04 17:59:42 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(134.000000f, 11414.0000f);
path.cubicTo(131.990234f, 11414.0000f, 130.326660f, 11415.4824f, 130.042755f, 11417.4131f);
path.cubicTo(130.233124f, 11418.3193f, 131.037079f, 11419.0000f, 132.000000f, 11419.0000f);
path.lineTo(806.000000f, 11419.0000f);
path.cubicTo(806.962891f, 11419.0000f, 807.766907f, 11418.3193f, 807.957275f, 11417.4131f);
path.cubicTo(807.673401f, 11415.4824f, 806.009766f, 11414.0000f, 804.000000f, 11414.0000f);
path.lineTo(134.000000f, 11414.0000f);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kInverseWinding_FillType);
pathB.moveTo(132.000000f, 11415.0000f);
pathB.lineTo(806.000000f, 11415.0000f);
pathB.cubicTo(807.104553f, 11415.0000f, 808.000000f, 11415.4473f, 808.000000f, 11416.0000f);
pathB.lineTo(808.000000f, 11417.0000f);
pathB.cubicTo(808.000000f, 11418.1045f, 807.104553f, 11419.0000f, 806.000000f, 11419.0000f);
pathB.lineTo(132.000000f, 11419.0000f);
pathB.cubicTo(130.895432f, 11419.0000f, 130.000000f, 11418.1045f, 130.000000f, 11417.0000f);
pathB.lineTo(130.000000f, 11416.0000f);
pathB.cubicTo(130.000000f, 11415.4473f, 130.895432f, 11415.0000f, 132.000000f, 11415.0000f);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
path ops -- rewrite angle sort This is a major change resulting from a minor tweak. In the old code, the intersection point of two curves was shared between them, but the intersection points and end points of sorted edges was computed directly from the intersection T value. In this CL, both intersection points and sorted points are the same, and intermediate control points are computed to preserve their slope. The sort itself has been completely rewritten to be more robust and remove 'magic' checks, conditions that empirically worked but couldn't be rationalized. This CL was triggered by errors generated computing the clips of SKP files. At this point, all 73M standard tests work and at least the first troublesome SKPs work. Review URL: https://codereview.chromium.org/15338003 git-svn-id: http://skia.googlecode.com/svn/trunk@9432 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-06-04 17:59:42 +00:00
}
path ops work in progress BUG= Review URL: https://codereview.chromium.org/18058007 git-svn-id: http://skia.googlecode.com/svn/trunk@9908 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-07-08 17:17:02 +00:00
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void skp96prezzi1(skiatest::Reporter* reporter, const char* filename) {
path ops work in progress BUG= Review URL: https://codereview.chromium.org/18058007 git-svn-id: http://skia.googlecode.com/svn/trunk@9908 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-07-08 17:17:02 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(157.464005f, 670.463989f);
path.quadTo(158.928925f, 669.000000f, 161.000000f, 669.000000f);
path.lineTo(248.000000f, 669.000000f);
path.quadTo(250.071075f, 669.000000f, 251.535995f, 670.463989f);
path.quadTo(253.000000f, 671.928955f, 253.000000f, 674.000000f);
path.lineTo(253.000000f, 706.000000f);
path.lineTo(251.000000f, 706.000000f);
path.lineTo(251.000000f, 675.000000f);
path.cubicTo(251.000000f, 672.790833f, 249.209137f, 671.000000f, 247.000000f, 671.000000f);
path.lineTo(162.000000f, 671.000000f);
path.cubicTo(159.790863f, 671.000000f, 158.000000f, 672.790833f, 158.000000f, 675.000000f);
path.lineTo(158.000000f, 706.000000f);
path.lineTo(156.000000f, 706.000000f);
path.lineTo(156.000000f, 674.000000f);
path.quadTo(156.000000f, 671.928955f, 157.464005f, 670.463989f);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(156.000000f, 669.000000f);
pathB.lineTo(178.500000f, 691.500000f);
pathB.lineTo(230.500000f, 691.500000f);
pathB.lineTo(253.000000f, 669.000000f);
pathB.lineTo(156.000000f, 669.000000f);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
path ops work in progress BUG= Review URL: https://codereview.chromium.org/18058007 git-svn-id: http://skia.googlecode.com/svn/trunk@9908 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-07-08 17:17:02 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void skpancestry_com1(skiatest::Reporter* reporter, const char* filename) {
path ops work in progress BUG= Review URL: https://codereview.chromium.org/18058007 git-svn-id: http://skia.googlecode.com/svn/trunk@9908 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-07-08 17:17:02 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(161.000000f, 925.000000f);
path.cubicTo(159.874390f, 925.000000f, 158.835663f, 925.371948f, 158.000000f, 925.999634f);
path.lineTo(158.000000f, 926.000000f);
path.lineTo(1108.00000f, 926.000000f);
path.lineTo(1108.00000f, 925.999634f);
path.cubicTo(1107.16443f, 925.371948f, 1106.12561f, 925.000000f, 1105.00000f, 925.000000f);
path.lineTo(161.000000f, 925.000000f);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kEvenOdd_FillType);
pathB.moveTo(161.000000f, 926.000000f);
pathB.lineTo(1105.00000f, 926.000000f);
pathB.cubicTo(1107.20911f, 926.000000f, 1109.00000f, 927.790833f, 1109.00000f, 930.000000f);
pathB.lineTo(1109.00000f, 956.000000f);
pathB.cubicTo(1109.00000f, 958.209167f, 1107.20911f, 960.000000f, 1105.00000f, 960.000000f);
pathB.lineTo(161.000000f, 960.000000f);
pathB.cubicTo(158.790863f, 960.000000f, 157.000000f, 958.209167f, 157.000000f, 956.000000f);
pathB.lineTo(157.000000f, 930.000000f);
pathB.cubicTo(157.000000f, 927.790833f, 158.790863f, 926.000000f, 161.000000f, 926.000000f);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
path ops work in progress BUG= Review URL: https://codereview.chromium.org/18058007 git-svn-id: http://skia.googlecode.com/svn/trunk@9908 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-07-08 17:17:02 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void skpeldorado_com_ua1(skiatest::Reporter* reporter, const char* filename) {
path ops work in progress BUG= Review URL: https://codereview.chromium.org/18058007 git-svn-id: http://skia.googlecode.com/svn/trunk@9908 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-07-08 17:17:02 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(286.695129f, 291.000000f);
path.lineTo(229.304855f, 561.000000f);
path.lineTo(979.304871f, 561.000000f);
path.lineTo(1036.69507f, 291.000000f);
path.lineTo(286.695129f, 291.000000f);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1006.69513f, 291.000000f);
pathB.cubicTo(1023.26367f, 291.000000f, 1033.84021f, 304.431458f, 1030.31836f, 321.000000f);
pathB.lineTo(985.681519f, 531.000000f);
pathB.cubicTo(982.159790f, 547.568542f, 965.873413f, 561.000000f, 949.304871f, 561.000000f);
pathB.lineTo(259.304871f, 561.000000f);
pathB.cubicTo(242.736313f, 561.000000f, 232.159805f, 547.568542f, 235.681549f, 531.000000f);
pathB.lineTo(280.318420f, 321.000000f);
pathB.cubicTo(283.840179f, 304.431458f, 300.126587f, 291.000000f, 316.695129f, 291.000000f);
pathB.lineTo(1006.69513f, 291.000000f);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
path ops work in progress BUG= Review URL: https://codereview.chromium.org/18058007 git-svn-id: http://skia.googlecode.com/svn/trunk@9908 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-07-08 17:17:02 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void skpbyte_com1(skiatest::Reporter* reporter, const char* filename) {
path ops work in progress BUG= Review URL: https://codereview.chromium.org/18058007 git-svn-id: http://skia.googlecode.com/svn/trunk@9908 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-07-08 17:17:02 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(968.000000f, 14.0000000f);
path.cubicTo(965.238586f, 14.0000000f, 963.000000f, 16.2385769f, 963.000000f, 19.0000000f);
path.lineTo(963.000000f, 32.0000000f);
path.cubicTo(963.000000f, 34.7614250f, 965.238586f, 37.0000000f, 968.000000f, 37.0000000f);
path.lineTo(1034.00000f, 37.0000000f);
path.cubicTo(1036.76147f, 37.0000000f, 1039.00000f, 34.7614250f, 1039.00000f, 32.0000000f);
path.lineTo(1039.00000f, 19.0000000f);
path.cubicTo(1039.00000f, 16.2385769f, 1036.76147f, 14.0000000f, 1034.00000f, 14.0000000f);
path.lineTo(968.000000f, 14.0000000f);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kInverseWinding_FillType);
pathB.moveTo(968.000000f, 14.0000000f);
pathB.lineTo(1034.00000f, 14.0000000f);
pathB.cubicTo(1036.76147f, 14.0000000f, 1039.00000f, 16.2385750f, 1039.00000f, 19.0000000f);
pathB.lineTo(1039.00000f, 32.0000000f);
pathB.cubicTo(1039.00000f, 34.2091408f, 1036.76147f, 36.0000000f, 1034.00000f, 36.0000000f);
pathB.lineTo(968.000000f, 36.0000000f);
pathB.cubicTo(965.238586f, 36.0000000f, 963.000000f, 34.2091408f, 963.000000f, 32.0000000f);
pathB.lineTo(963.000000f, 19.0000000f);
pathB.cubicTo(963.000000f, 16.2385750f, 965.238586f, 14.0000000f, 968.000000f, 14.0000000f);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
path ops work in progress BUG= Review URL: https://codereview.chromium.org/18058007 git-svn-id: http://skia.googlecode.com/svn/trunk@9908 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-07-08 17:17:02 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void skphealth_com76(skiatest::Reporter* reporter, const char* filename) {
path ops work in progress BUG= Review URL: https://codereview.chromium.org/18058007 git-svn-id: http://skia.googlecode.com/svn/trunk@9908 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-07-08 17:17:02 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(708.099182f, 7.09919119f);
path.lineTo(708.099182f, 7.09920025f);
path.quadTo(704.000000f, 11.2010098f, 704.000000f, 17.0000000f);
path.lineTo(704.000000f, 33.0000000f);
path.lineTo(705.000000f, 33.0000000f);
path.lineTo(705.000000f, 17.0000000f);
path.cubicTo(705.000000f, 13.4101496f, 706.455078f, 10.1601505f, 708.807617f, 7.80761385f);
path.lineTo(708.099182f, 7.09919119f);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(704.000000f, 3.00000000f);
pathB.lineTo(704.000000f, 33.0000000f);
pathB.lineTo(705.000000f, 33.0000000f);
pathB.lineTo(719.500000f, 3.00000000f);
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
path ops work in progress BUG= Review URL: https://codereview.chromium.org/18058007 git-svn-id: http://skia.googlecode.com/svn/trunk@9908 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-07-08 17:17:02 +00:00
}
path ops -- rewrite angle sort This is a major change resulting from a minor tweak. In the old code, the intersection point of two curves was shared between them, but the intersection points and end points of sorted edges was computed directly from the intersection T value. In this CL, both intersection points and sorted points are the same, and intermediate control points are computed to preserve their slope. The sort itself has been completely rewritten to be more robust and remove 'magic' checks, conditions that empirically worked but couldn't be rationalized. This CL was triggered by errors generated computing the clips of SKP files. At this point, all 73M standard tests work and at least the first troublesome SKPs work. Review URL: https://codereview.chromium.org/15338003 git-svn-id: http://skia.googlecode.com/svn/trunk@9432 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-06-04 17:59:42 +00:00
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void skpahrefs_com88(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@10096 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-07-16 07:00:56 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(1099.82886f, 7.17117119f);
path.lineTo(1099.12134f, 7.87867832f);
path.cubicTo(1099.66418f, 8.42157173f, 1100.00000f, 9.17157173f, 1100.00000f, 10.0000000f);
path.lineTo(1100.00000f, 28.0000000f);
path.cubicTo(1100.00000f, 29.6568546f, 1098.65686f, 31.0000000f, 1097.00000f, 31.0000000f);
path.lineTo(1088.00000f, 31.0000000f);
path.lineTo(1088.00000f, 32.0000000f);
path.lineTo(1097.00000f, 32.0000000f);
path.quadTo(1098.65686f, 32.0000000f, 1099.82886f, 30.8288002f);
path.quadTo(1101.00000f, 29.6568546f, 1101.00000f, 28.0000000f);
path.lineTo(1101.00000f, 10.0000000f);
path.quadTo(1101.00000f, 8.34314537f, 1099.82886f, 7.17119980f);
path.lineTo(1099.82886f, 7.17117119f);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1101.00000f, 6.00000000f);
pathB.lineTo(1088.00000f, 6.00000000f);
pathB.lineTo(1088.00000f, 19.0000000f);
pathB.lineTo(1101.00000f, 32.0000000f);
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@10096 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-07-16 07:00:56 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void skpahrefs_com29(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@10096 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-07-16 07:00:56 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(1037.17114f, 7.17119980f);
path.quadTo(1038.34314f, 6.00000000f, 1040.00000f, 6.00000000f);
path.lineTo(1074.00000f, 6.00000000f);
path.lineTo(1074.00000f, 32.0000000f);
path.lineTo(1040.00000f, 32.0000000f);
path.quadTo(1038.34314f, 32.0000000f, 1037.17114f, 30.8288002f);
path.quadTo(1036.00000f, 29.6568546f, 1036.00000f, 28.0000000f);
path.lineTo(1036.00000f, 10.0000000f);
path.quadTo(1036.00000f, 8.34314537f, 1037.17114f, 7.17119980f);
path.close();
path.moveTo(1037.00000f, 10.0000000f);
path.cubicTo(1037.00000f, 8.34314537f, 1038.34314f, 7.00000000f, 1040.00000f, 7.00000000f);
path.lineTo(1073.00000f, 7.00000000f);
path.lineTo(1073.00000f, 31.0000000f);
path.lineTo(1040.00000f, 31.0000000f);
path.cubicTo(1038.34314f, 31.0000000f, 1037.00000f, 29.6568546f, 1037.00000f, 28.0000000f);
path.lineTo(1037.00000f, 10.0000000f);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1036.00000f, 32.0000000f);
pathB.lineTo(1049.00000f, 19.0000000f);
pathB.lineTo(1073.00000f, 31.0000000f);
pathB.lineTo(1074.00000f, 32.0000000f);
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@10096 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-07-16 07:00:56 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp85d(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@10096 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-07-16 07:00:56 +00:00
SkPath path;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(1,6, 1,0, 6,2);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,1);
pathB.cubicTo(2,6, 1,0, 6,1);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
path ops near exact Modify line intersections to first - match exact ends - compute intersections - match near ends where the exact ends are preferred, then near matches, then computed matches. This pulls matches towards existing end points when possible, and keeps intersection distances consistent with different line/line line/quad and line/cubic computations. BUG= Review URL: https://codereview.chromium.org/19183003 git-svn-id: http://skia.googlecode.com/svn/trunk@10073 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-07-15 13:29:13 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void skpkkiste_to98(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@10096 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-07-16 07:00:56 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(96, 122);
path.cubicTo(94.6192932f, 122, 93.3692932f, 122.559647f, 92.4644699f, 123.46447f);
path.lineTo(94.1715698f, 125.17157f);
path.cubicTo(94.8954315f, 124.447708f, 95.8954315f, 124, 97, 124);
path.lineTo(257, 124);
path.cubicTo(258.104553f, 124, 259.104584f, 124.447708f, 259.82843f, 125.17157f);
path.lineTo(261.535522f, 123.46447f);
path.cubicTo(260.630707f, 122.559647f, 259.380707f, 122, 258, 122);
path.lineTo(96, 122);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(258, 122);
pathB.cubicTo(260.761414f, 122, 263, 124.238579f, 263, 127);
pathB.lineTo(263, 284);
pathB.cubicTo(263, 286.761414f, 260.761414f, 289, 258, 289);
pathB.lineTo(96, 289);
pathB.cubicTo(93.2385788f, 289, 91, 286.761414f, 91, 284);
pathB.lineTo(91, 127);
pathB.cubicTo(91, 124.238579f, 93.2385788f, 122, 96, 122);
pathB.lineTo(258, 122);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@10096 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-07-16 07:00:56 +00:00
}
path ops near exact Modify line intersections to first - match exact ends - compute intersections - match near ends where the exact ends are preferred, then near matches, then computed matches. This pulls matches towards existing end points when possible, and keeps intersection distances consistent with different line/line line/quad and line/cubic computations. BUG= Review URL: https://codereview.chromium.org/19183003 git-svn-id: http://skia.googlecode.com/svn/trunk@10073 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-07-15 13:29:13 +00:00
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void issue1417(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@10299 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-07-24 07:01:12 +00:00
SkPath path1;
path1.moveTo(122.58908843994140625f, 82.2836456298828125f);
path1.quadTo(129.8215789794921875f, 80, 138, 80);
path1.quadTo(147.15692138671875f, 80, 155.1280364990234375f, 82.86279296875f);
path1.lineTo(161.1764678955078125f, 100);
path1.lineTo(161.1764678955078125f, 100);
path1.lineTo(115.29412078857421875f, 100);
path1.lineTo(115.29412078857421875f, 100);
path1.lineTo(122.58908843994140625f, 82.2836456298828125f);
path1.lineTo(122.58908843994140625f, 82.2836456298828125f);
path1.close();
path1.moveTo(98.68194580078125f, 140.343841552734375f);
path1.lineTo(115.29412078857421875f, 100);
path1.lineTo(115.29412078857421875f, 100);
path1.lineTo(97.9337615966796875f, 100);
path1.lineTo(97.9337615966796875f, 100);
path1.quadTo(88, 112.94264984130859375f, 88, 130);
path1.quadTo(88, 131.544830322265625f, 88.08148956298828125f, 133.0560302734375f);
path1.lineTo(98.68194580078125f, 140.343841552734375f);
path1.lineTo(98.68194580078125f, 140.343841552734375f);
path1.close();
path1.moveTo(136.969696044921875f, 166.6666717529296875f);
path1.lineTo(98.68194580078125f, 140.343841552734375f);
path1.lineTo(98.68194580078125f, 140.343841552734375f);
path1.lineTo(93.45894622802734375f, 153.02825927734375f);
path1.lineTo(93.45894622802734375f, 153.02825927734375f);
path1.quadTo(96.94116973876953125f, 159.65185546875f, 102.64466094970703125f, 165.3553466796875f);
path1.quadTo(110.7924652099609375f, 173.503143310546875f, 120.8179779052734375f, 177.1177825927734375f);
path1.lineTo(136.969696044921875f, 166.6666717529296875f);
path1.lineTo(136.969696044921875f, 166.6666717529296875f);
path1.close();
path1.moveTo(175.8309783935546875f, 141.5211334228515625f);
path1.lineTo(136.969696044921875f, 166.6666717529296875f);
path1.lineTo(136.969696044921875f, 166.6666717529296875f);
path1.lineTo(153.15728759765625f, 177.7956390380859375f);
path1.lineTo(153.15728759765625f, 177.7956390380859375f);
path1.quadTo(164.392425537109375f, 174.318267822265625f, 173.3553466796875f, 165.3553466796875f);
path1.quadTo(177.805816650390625f, 160.9048614501953125f, 180.90380859375f, 155.8941650390625f);
path1.lineTo(175.8309783935546875f, 141.5211334228515625f);
path1.lineTo(175.8309783935546875f, 141.5211334228515625f);
path1.close();
path1.moveTo(175.8309783935546875f, 141.5211334228515625f);
path1.lineTo(187.8782806396484375f, 133.7258148193359375f);
path1.lineTo(187.8782806396484375f, 133.7258148193359375f);
path1.quadTo(188, 131.8880615234375f, 188, 130);
path1.quadTo(188, 112.942657470703125f, 178.0662384033203125f, 100);
path1.lineTo(161.1764678955078125f, 100);
path1.lineTo(161.1764678955078125f, 100);
path1.lineTo(175.8309783935546875f, 141.5211334228515625f);
path1.lineTo(175.8309783935546875f, 141.5211334228515625f);
path1.close();
SkPath path2;
path2.moveTo(174.117645263671875f, 100);
path2.lineTo(161.1764678955078125f, 100);
path2.lineTo(161.1764678955078125f, 100);
path2.lineTo(155.1280364990234375f, 82.86279296875f);
path2.lineTo(155.1280364990234375f, 82.86279296875f);
path2.quadTo(153.14971923828125f, 82.15229034423828125f, 151.098419189453125f, 81.618133544921875f);
path2.lineTo(143.5294189453125f, 100);
path2.lineTo(143.5294189453125f, 100);
path2.lineTo(161.1764678955078125f, 100);
path2.lineTo(161.1764678955078125f, 100);
path2.lineTo(168.23529052734375f, 120);
path2.lineTo(168.23529052734375f, 120);
path2.lineTo(181.1764678955078125f, 120);
path2.lineTo(181.1764678955078125f, 120);
path2.lineTo(186.3661956787109375f, 134.7042236328125f);
path2.lineTo(186.3661956787109375f, 134.7042236328125f);
path2.lineTo(187.8782806396484375f, 133.7258148193359375f);
path2.lineTo(187.8782806396484375f, 133.7258148193359375f);
path2.quadTo(188, 131.8880615234375f, 188, 130);
path2.quadTo(188, 124.80947113037109375f, 187.080169677734375f, 120);
path2.lineTo(181.1764678955078125f, 120);
path2.lineTo(181.1764678955078125f, 120);
path2.lineTo(174.117645263671875f, 100);
path2.lineTo(174.117645263671875f, 100);
path2.close();
path2.moveTo(88.91983795166015625f, 120);
path2.lineTo(107.0588226318359375f, 120);
path2.lineTo(107.0588226318359375f, 120);
path2.lineTo(98.68194580078125f, 140.343841552734375f);
path2.lineTo(98.68194580078125f, 140.343841552734375f);
path2.lineTo(88.08148956298828125f, 133.0560302734375f);
path2.lineTo(88.08148956298828125f, 133.0560302734375f);
path2.quadTo(88, 131.544830322265625f, 88, 130);
path2.quadTo(88, 124.80951690673828125f, 88.91983795166015625f, 120);
path2.close();
path2.moveTo(96.67621612548828125f, 145.21490478515625f);
path2.lineTo(98.68194580078125f, 140.343841552734375f);
path2.lineTo(98.68194580078125f, 140.343841552734375f);
path2.lineTo(120.68767547607421875f, 155.4727783203125f);
path2.lineTo(120.68767547607421875f, 155.4727783203125f);
path2.lineTo(118.68194580078125f, 160.343841552734375f);
path2.lineTo(118.68194580078125f, 160.343841552734375f);
path2.lineTo(96.67621612548828125f, 145.21490478515625f);
path2.lineTo(96.67621612548828125f, 145.21490478515625f);
path2.close();
path2.moveTo(113.232177734375f, 173.5789947509765625f);
path2.quadTo(116.8802642822265625f, 175.69805908203125f, 120.8179779052734375f, 177.1177825927734375f);
path2.lineTo(132.2864990234375f, 169.6969757080078125f);
path2.lineTo(132.2864990234375f, 169.6969757080078125f);
path2.lineTo(118.68194580078125f, 160.343841552734375f);
path2.lineTo(118.68194580078125f, 160.343841552734375f);
path2.lineTo(113.232177734375f, 173.5789947509765625f);
path2.lineTo(113.232177734375f, 173.5789947509765625f);
path2.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
// FIXME : difficult data, circle back later
testPathOp(reporter, path1, path2, kUnion_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@10096 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-07-16 07:00:56 +00:00
}
add florin tests hopefully fix build BUG= Review URL: https://codereview.chromium.org/19183005 git-svn-id: http://skia.googlecode.com/svn/trunk@10079 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-07-15 15:00:58 +00:00
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void issue1418(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@10299 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-07-24 07:01:12 +00:00
SkPath path1;
path1.moveTo(0, 0);
path1.lineTo(1, 0);
path1.lineTo(1, 0);
path1.lineTo(1, 1);
path1.lineTo(1, 1);
path1.lineTo(0, 1);
path1.lineTo(0, 1);
path1.lineTo(0, 0);
path1.lineTo(0, 0);
path1.close();
SkPath path2;
path2.moveTo(0.64644664525985717773f, -0.35355341434478759766f);
path2.quadTo(0.79289329051971435547f, -0.50000005960464477539f, 1.0000001192092895508f, -0.50000005960464477539f);
path2.quadTo(1.2071068286895751953f, -0.50000005960464477539f, 1.3535535335540771484f, -0.35355341434478759766f);
path2.quadTo(1.5000001192092895508f, -0.20710679888725280762f, 1.5000001192092895508f, 0);
path2.quadTo(1.5000001192092895508f, 0.20710679888725280762f, 1.3535535335540771484f, 0.35355341434478759766f);
path2.quadTo(1.2071068286895751953f, 0.50000005960464477539f, 1.0000001192092895508f, 0.50000005960464477539f);
path2.quadTo(0.79289329051971435547f, 0.50000005960464477539f, 0.64644664525985717773f, 0.35355341434478759766f);
path2.quadTo(0.50000005960464477539f, 0.20710679888725280762f, 0.50000005960464477539f, 0);
turn off debugging printfs fix pathops issues 1417, 1418 be more rigorous about pulling intersections of lines to end points rewrite cubic/line and quad/line intersections to share style BUG= Review URL: https://codereview.chromium.org/19543005 git-svn-id: http://skia.googlecode.com/svn/trunk@10270 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-07-23 15:27:41 +00:00
path2.quadTo(0.50000005960464477539f, -0.20710679888725280762f, 0.64644664525985717773f, -0.35355341434478759766f);
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path1, path2, kIntersect_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@10096 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-07-16 07:00:56 +00:00
}
turn off debugging printfs fix pathops issues 1417, 1418 be more rigorous about pulling intersections of lines to end points rewrite cubic/line and quad/line intersections to share style BUG= Review URL: https://codereview.chromium.org/19543005 git-svn-id: http://skia.googlecode.com/svn/trunk@10270 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-07-23 15:27:41 +00:00
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp85i(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@10299 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-07-24 07:01:12 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(3, 4);
path.cubicTo(1, 5, 4, 3, 6, 4);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(3, 4);
pathB.cubicTo(4, 6, 4, 3, 5, 1);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
turn off debugging printfs fix pathops issues 1417, 1418 be more rigorous about pulling intersections of lines to end points rewrite cubic/line and quad/line intersections to share style BUG= Review URL: https://codereview.chromium.org/19543005 git-svn-id: http://skia.googlecode.com/svn/trunk@10270 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-07-23 15:27:41 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void issue1418b(skiatest::Reporter* reporter, const char* filename) {
path ops work in progress path ops work in progress BUG= Review URL: https://codereview.chromium.org/21359002 git-svn-id: http://skia.googlecode.com/svn/trunk@11291 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-09-16 15:55:01 +00:00
SkPath path1;
path1.moveTo(0, 0);
path1.lineTo(1, 0);
path1.lineTo(1, 1);
path1.lineTo(0, 1);
path1.lineTo(0, 0);
path1.close();
path1.setFillType(SkPath::kWinding_FillType);
SkPath path2;
path2.moveTo(0.646446645f, -0.353553414f);
path2.quadTo(0.792893291f, -0.50000006f, 1.00000012f, -0.50000006f);
path2.quadTo(1.20710683f, -0.50000006f, 1.35355353f, -0.353553414f);
path2.quadTo(1.50000012f, -0.207106799f, 1.50000012f, 0);
path2.quadTo(1.50000012f, 0.207106799f, 1.35355353f, 0.353553414f);
path2.quadTo(1.20710683f, 0.50000006f, 1.00000012f, 0.50000006f);
path2.quadTo(0.792893291f, 0.50000006f, 0.646446645f, 0.353553414f);
path2.quadTo(0.50000006f, 0.207106799f, 0.50000006f, 0);
path2.quadTo(0.50000006f, -0.207106799f, 0.646446645f, -0.353553414f);
path2.close();
path2.moveTo(1.00000012f, 0.50000006f);
path2.lineTo(1.00000012f, 1.00000012f);
path2.lineTo(0.50000006f, 1.00000012f);
path2.quadTo(0.50000006f, 0.792893291f, 0.646446645f, 0.646446645f);
path2.quadTo(0.792893291f, 0.50000006f, 1.00000012f, 0.50000006f);
path2.close();
path2.setFillType(SkPath::kEvenOdd_FillType);
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path1, path2, kIntersect_SkPathOp, filename);
path ops work in progress path ops work in progress BUG= Review URL: https://codereview.chromium.org/21359002 git-svn-id: http://skia.googlecode.com/svn/trunk@11291 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-09-16 15:55:01 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void rectOp1i(skiatest::Reporter* reporter, const char* filename) {
path ops work in progress path ops work in progress BUG= Review URL: https://codereview.chromium.org/21359002 git-svn-id: http://skia.googlecode.com/svn/trunk@11291 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-09-16 15:55:01 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.addRect(0, 0, 1, 1, SkPath::kCW_Direction);
path.addRect(2, 2, 4, 4, SkPath::kCW_Direction);
pathB.setFillType(SkPath::kWinding_FillType);
pathB.addRect(0, 0, 1, 1, SkPath::kCW_Direction);
pathB.addRect(0, 0, 2, 2, SkPath::kCW_Direction);
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
path ops work in progress path ops work in progress BUG= Review URL: https://codereview.chromium.org/21359002 git-svn-id: http://skia.googlecode.com/svn/trunk@11291 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-09-16 15:55:01 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void rectOp2i(skiatest::Reporter* reporter, const char* filename) {
path ops work in progress path ops work in progress BUG= Review URL: https://codereview.chromium.org/21359002 git-svn-id: http://skia.googlecode.com/svn/trunk@11291 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-09-16 15:55:01 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kEvenOdd_FillType);
path.addRect(0, 0, 1, 1, SkPath::kCW_Direction);
path.addRect(0, 0, 3, 3, SkPath::kCW_Direction);
pathB.setFillType(SkPath::kWinding_FillType);
pathB.addRect(0, 0, 2, 2, SkPath::kCW_Direction);
pathB.addRect(0, 0, 2, 2, SkPath::kCW_Direction);
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
path ops work in progress path ops work in progress BUG= Review URL: https://codereview.chromium.org/21359002 git-svn-id: http://skia.googlecode.com/svn/trunk@11291 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-09-16 15:55:01 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void rectOp3x(skiatest::Reporter* reporter, const char* filename) {
path ops work in progress path ops work in progress BUG= Review URL: https://codereview.chromium.org/21359002 git-svn-id: http://skia.googlecode.com/svn/trunk@11291 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-09-16 15:55:01 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(0, 0);
path.lineTo(3, 0);
path.lineTo(3, 3);
path.lineTo(0, 3);
path.close();
path.moveTo(2, 2);
path.lineTo(3, 2);
path.lineTo(3, 3);
path.lineTo(2, 3);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1, 1);
pathB.lineTo(3, 1);
pathB.lineTo(3, 3);
pathB.lineTo(1, 3);
pathB.close();
pathB.moveTo(2, 2);
pathB.lineTo(3, 2);
pathB.lineTo(3, 3);
pathB.lineTo(2, 3);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kXOR_SkPathOp, filename);
path ops work in progress path ops work in progress BUG= Review URL: https://codereview.chromium.org/21359002 git-svn-id: http://skia.googlecode.com/svn/trunk@11291 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-09-16 15:55:01 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void issue1435(skiatest::Reporter* reporter, const char* filename) {
path ops work in progress path ops work in progress BUG= Review URL: https://codereview.chromium.org/21359002 git-svn-id: http://skia.googlecode.com/svn/trunk@11291 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-09-16 15:55:01 +00:00
SkPath path1;
path1.moveTo(160, 60);
path1.lineTo(220, 230);
path1.lineTo(60, 120);
path1.lineTo(260, 120);
path1.lineTo(90, 230);
path1.lineTo(160, 60);
path1.close();
path1.setFillType(SkPath::kEvenOdd_FillType);
SkPath path2;
path2.moveTo(142.589081f, 102.283646f);
path2.quadTo(149.821579f, 100, 158, 100);
path2.quadTo(167.156921f, 100, 175.128036f, 102.862793f);
path2.lineTo(181.176468f, 120);
path2.lineTo(135.294128f, 120);
path2.lineTo(142.589081f, 102.283646f);
path2.close();
path2.moveTo(118.681946f, 160.343842f);
path2.lineTo(135.294128f, 120);
path2.lineTo(117.933762f, 120);
path2.quadTo(108, 132.942657f, 108, 150);
path2.quadTo(108, 151.54483f, 108.08149f, 153.05603f);
path2.lineTo(118.681946f, 160.343842f);
path2.close();
path2.moveTo(156.969696f, 186.666672f);
path2.lineTo(118.681946f, 160.343842f);
path2.lineTo(113.458946f, 173.028259f);
path2.quadTo(116.94117f, 179.651855f, 122.644661f, 185.355347f);
path2.quadTo(130.792465f, 193.503143f, 140.817978f, 197.117783f);
path2.lineTo(156.969696f, 186.666672f);
path2.close();
path2.moveTo(195.830978f, 161.521133f);
path2.lineTo(156.969696f, 186.666672f);
path2.lineTo(173.157288f, 197.795639f);
path2.quadTo(184.392426f, 194.318268f, 193.355347f, 185.355347f);
path2.quadTo(197.805817f, 180.904861f, 200.903809f, 175.894165f);
path2.lineTo(195.830978f, 161.521133f);
path2.close();
path2.moveTo(195.830978f, 161.521133f);
path2.lineTo(207.878281f, 153.725815f);
path2.quadTo(208, 151.888062f, 208, 150);
path2.quadTo(208, 132.942657f, 198.066238f, 120);
path2.lineTo(181.176468f, 120);
path2.lineTo(195.830978f, 161.521133f);
path2.close();
path2.setFillType(SkPath::kEvenOdd_FillType);
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path1, path2, kIntersect_SkPathOp, filename);
path ops work in progress path ops work in progress BUG= Review URL: https://codereview.chromium.org/21359002 git-svn-id: http://skia.googlecode.com/svn/trunk@11291 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-09-16 15:55:01 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void skpkkiste_to716(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@10299 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-07-24 07:01:12 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(1173, 284);
path.cubicTo(1173, 285.125824f, 1173.37207f, 286.164734f, 1174, 287.000488f);
path.lineTo(1174, 123.999496f);
path.cubicTo(1173.37207f, 124.835243f, 1173, 125.874168f, 1173, 127);
path.lineTo(1173, 284);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1340, 122);
pathB.cubicTo(1342.76147f, 122, 1345, 124.238579f, 1345, 127);
pathB.lineTo(1345, 284);
pathB.cubicTo(1345, 286.761414f, 1342.76147f, 289, 1340, 289);
pathB.lineTo(1178, 289);
pathB.cubicTo(1175.23853f, 289, 1173, 286.761414f, 1173, 284);
pathB.lineTo(1173, 127);
pathB.cubicTo(1173, 124.238579f, 1175.23853f, 122, 1178, 122);
pathB.lineTo(1340, 122);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@10299 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-07-24 07:01:12 +00:00
}
add florin tests hopefully fix build BUG= Review URL: https://codereview.chromium.org/19183005 git-svn-id: http://skia.googlecode.com/svn/trunk@10079 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-07-15 15:00:58 +00:00
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void loopEdge1(skiatest::Reporter* reporter, const char* filename) {
path ops work in progress path ops work in progress BUG= Review URL: https://codereview.chromium.org/21359002 git-svn-id: http://skia.googlecode.com/svn/trunk@11291 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-09-16 15:55:01 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(0,0);
path.lineTo(3,0);
path.lineTo(3,2);
path.lineTo(1,2);
path.lineTo(1,1);
path.lineTo(2,1);
path.lineTo(2,3);
path.lineTo(0,3);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kEvenOdd_FillType);
pathB.moveTo(1,2);
pathB.lineTo(2,2);
pathB.lineTo(2,4);
pathB.lineTo(1,4);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
path ops work in progress path ops work in progress BUG= Review URL: https://codereview.chromium.org/21359002 git-svn-id: http://skia.googlecode.com/svn/trunk@11291 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-09-16 15:55:01 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void loopEdge2(skiatest::Reporter* reporter, const char* filename) {
path ops work in progress path ops work in progress BUG= Review URL: https://codereview.chromium.org/21359002 git-svn-id: http://skia.googlecode.com/svn/trunk@11291 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-09-16 15:55:01 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(0,0);
path.lineTo(3,0);
path.lineTo(3,2);
path.lineTo(1,2);
path.lineTo(1,1);
path.lineTo(2,1);
path.lineTo(2,3);
path.lineTo(0,3);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kEvenOdd_FillType);
pathB.moveTo(1 - 1e-6f,2);
pathB.lineTo(2 - 1e-6f,2);
pathB.lineTo(2 - 1e-6f,4);
pathB.lineTo(1 - 1e-6f,4);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
path ops work in progress path ops work in progress BUG= Review URL: https://codereview.chromium.org/21359002 git-svn-id: http://skia.googlecode.com/svn/trunk@11291 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-09-16 15:55:01 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp86i(skiatest::Reporter* reporter, const char* filename) {
path ops work in progress path ops work in progress BUG= Review URL: https://codereview.chromium.org/21359002 git-svn-id: http://skia.googlecode.com/svn/trunk@11291 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-09-16 15:55:01 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0, 4);
path.cubicTo(3, 4, 6, 2, 5, 2);
path.close();
pathB.setFillType(SkPath::kEvenOdd_FillType);
pathB.moveTo(2, 6);
pathB.cubicTo(2, 5, 4, 0, 4, 3);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
path ops work in progress path ops work in progress BUG= Review URL: https://codereview.chromium.org/21359002 git-svn-id: http://skia.googlecode.com/svn/trunk@11291 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-09-16 15:55:01 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp87u(skiatest::Reporter* reporter, const char* filename) {
path ops work in progress path ops work in progress BUG= Review URL: https://codereview.chromium.org/21359002 git-svn-id: http://skia.googlecode.com/svn/trunk@11291 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-09-16 15:55:01 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(0,2, 2,0, 6,4);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,2);
pathB.cubicTo(4,6, 1,0, 2,0);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kUnion_SkPathOp, filename);
path ops work in progress path ops work in progress BUG= Review URL: https://codereview.chromium.org/21359002 git-svn-id: http://skia.googlecode.com/svn/trunk@11291 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-09-16 15:55:01 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp88u(skiatest::Reporter* reporter, const char* filename) {
path ops work in progress path ops work in progress BUG= Review URL: https://codereview.chromium.org/21359002 git-svn-id: http://skia.googlecode.com/svn/trunk@11291 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-09-16 15:55:01 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(2,5, 5,0, 6,4);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,5);
pathB.cubicTo(4,6, 1,0, 5,2);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kUnion_SkPathOp, filename);
path ops work in progress path ops work in progress BUG= Review URL: https://codereview.chromium.org/21359002 git-svn-id: http://skia.googlecode.com/svn/trunk@11291 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-09-16 15:55:01 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp89u(skiatest::Reporter* reporter, const char* filename) {
path ops work in progress path ops work in progress BUG= Review URL: https://codereview.chromium.org/21359002 git-svn-id: http://skia.googlecode.com/svn/trunk@11291 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-09-16 15:55:01 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0, 3);
path.cubicTo(1, 6, 5, 0, 6, 3);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 5);
pathB.cubicTo(3, 6, 3, 0, 6, 1);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kUnion_SkPathOp, filename);
path ops work in progress path ops work in progress BUG= Review URL: https://codereview.chromium.org/21359002 git-svn-id: http://skia.googlecode.com/svn/trunk@11291 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-09-16 15:55:01 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp90u(skiatest::Reporter* reporter, const char* filename) {
path ops work in progress path ops work in progress BUG= Review URL: https://codereview.chromium.org/21359002 git-svn-id: http://skia.googlecode.com/svn/trunk@11291 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-09-16 15:55:01 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(0, 5);
path.cubicTo(1, 2, 5, 2, 4, 1);
path.close();
pathB.setFillType(SkPath::kEvenOdd_FillType);
pathB.moveTo(2, 5);
pathB.cubicTo(1, 4, 5, 0, 2, 1);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kUnion_SkPathOp, filename);
path ops work in progress path ops work in progress BUG= Review URL: https://codereview.chromium.org/21359002 git-svn-id: http://skia.googlecode.com/svn/trunk@11291 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-09-16 15:55:01 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp91u(skiatest::Reporter* reporter, const char* filename) {
path ops work in progress path ops work in progress BUG= Review URL: https://codereview.chromium.org/21359002 git-svn-id: http://skia.googlecode.com/svn/trunk@11291 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-09-16 15:55:01 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(1, 6);
path.cubicTo(0, 3, 6, 3, 5, 0);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(3, 6);
pathB.cubicTo(0, 5, 6, 1, 3, 0);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kUnion_SkPathOp, filename);
path ops work in progress path ops work in progress BUG= Review URL: https://codereview.chromium.org/21359002 git-svn-id: http://skia.googlecode.com/svn/trunk@11291 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-09-16 15:55:01 +00:00
}
path ops work in progress make more skps work remove edit files BUG= Review URL: https://codereview.chromium.org/23542056 git-svn-id: http://skia.googlecode.com/svn/trunk@11570 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-02 14:49:34 +00:00
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
static void skpaaalgarve_org53(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(-1.24344979e-014f, 348);
path.lineTo(258, 348);
path.lineTo(258, 322);
path.quadTo(258, 317.857849f, 255.072006f, 314.928009f);
path.quadTo(252.142136f, 312, 248, 312);
path.lineTo(1.77635684e-015f, 312);
path.lineTo(-1.24344979e-014f, 348);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 312);
pathB.lineTo(258, 312);
pathB.lineTo(258, 348);
pathB.lineTo(0, 348);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
}
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
static void skpabcspark_ca103(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(1.99840144e-015f, 494);
path.lineTo(97, 494);
path.quadTo(100.313705f, 494, 102.6576f, 491.657593f);
path.quadTo(105, 489.313721f, 105, 486);
path.lineTo(105, 425);
path.quadTo(105, 421.686279f, 102.6576f, 419.342407f);
path.quadTo(100.313705f, 417, 97, 417);
path.lineTo(2.22044605e-016f, 417);
path.lineTo(1.99840144e-015f, 494);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 417);
pathB.lineTo(105, 417);
pathB.lineTo(105, 494);
pathB.lineTo(0, 494);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
}
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
static void skpacesoftech_com47(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(670.537415f, 285);
path.lineTo(670.387451f, 285);
path.lineTo(596.315186f, 314.850708f);
path.lineTo(626.19696f, 389);
path.lineTo(626.346863f, 389);
path.lineTo(700.419189f, 359.149261f);
path.lineTo(670.537415f, 285);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(663.318542f, 374.100616f);
pathB.quadTo(647.950989f, 380.293671f, 632.705322f, 373.806305f);
pathB.quadTo(617.459595f, 367.318909f, 611.266541f, 351.951355f);
pathB.quadTo(605.073486f, 336.58374f, 611.560913f, 321.338074f);
pathB.quadTo(618.048279f, 306.092407f, 633.415833f, 299.899353f);
pathB.quadTo(648.783447f, 293.706299f, 664.029114f, 300.193665f);
pathB.quadTo(679.27478f, 306.68103f, 685.467834f, 322.048645f);
pathB.quadTo(691.660889f, 337.416199f, 685.173523f, 352.661896f);
pathB.quadTo(678.686157f, 367.907562f, 663.318542f, 374.100616f);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
}
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
static void skpact_com43(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(1.45716772e-016f, 924.336121f);
path.lineTo(-1.11022302e-016f, 920);
path.lineTo(6, 920);
path.lineTo(6, 926);
path.lineTo(1.66389287f, 926);
path.quadTo(1.18842196f, 925.674561f, 0.756800175f, 925.243225f);
path.quadTo(0.325406998f, 924.811523f, 1.45716772e-016f, 924.336121f);
path.close();
path.moveTo(1, 921);
path.lineTo(5, 921);
path.lineTo(5, 925);
path.cubicTo(2.79086018f, 925, 1, 923.209167f, 1, 921);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(-1, 920);
pathB.lineTo(0, 920);
pathB.lineTo(3, 927);
pathB.lineTo(-1, 927);
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
}
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
static void skpadbox_lt8(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(320.097229f, 628.573669f);
path.lineTo(610.227173f, 85.7786865f);
path.lineTo(946.652588f, 265.601807f);
path.lineTo(656.522644f, 808.39679f);
path.lineTo(320.097229f, 628.573669f);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kInverseWinding_FillType);
pathB.moveTo(333.866608f, 623.496155f);
pathB.lineTo(613.368042f, 100.585754f);
pathB.cubicTo(613.685303f, 99.9921265f, 614.423767f, 99.7681885f, 615.017395f, 100.085449f);
pathB.lineTo(932.633057f, 269.854553f);
pathB.cubicTo(933.226685f, 270.171875f, 933.450623f, 270.910278f, 933.133301f, 271.503906f);
pathB.lineTo(653.631897f, 794.414307f);
pathB.cubicTo(653.314636f, 795.007935f, 652.576172f, 795.231934f, 651.982544f, 794.914612f);
pathB.lineTo(334.366943f, 625.145508f);
pathB.cubicTo(333.773315f, 624.828247f, 333.549286f, 624.089783f, 333.866608f, 623.496155f);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
}
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
static void skpadindex_de4(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(0, 926);
path.lineTo(0, 0);
path.lineTo(1280, 0);
path.lineTo(1280, 926);
path.lineTo(0, 926);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 312);
pathB.lineTo(8.20486257e-015f, 178);
pathB.lineTo(49, 178);
pathB.lineTo(49, 312);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
}
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
static void skpadithya_putr4_blogspot_com551(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(205.605804f, 142.334625f);
path.lineTo(254.665359f, 85.6058044f);
path.lineTo(311.394196f, 134.665359f);
path.lineTo(262.334625f, 191.39418f);
path.lineTo(205.605804f, 142.334625f);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(283.407959f, 110.462646f);
pathB.cubicTo(298.864319f, 123.829437f, 300.558258f, 147.195221f, 287.191467f, 162.651581f);
pathB.lineTo(286.537354f, 163.407959f);
pathB.cubicTo(273.170563f, 178.864334f, 249.804779f, 180.558258f, 234.348419f, 167.191467f);
pathB.lineTo(233.592026f, 166.537338f);
pathB.cubicTo(218.135666f, 153.170547f, 216.441727f, 129.804779f, 229.808517f, 114.348412f);
pathB.lineTo(230.462646f, 113.592026f);
pathB.cubicTo(243.829437f, 98.1356659f, 267.195221f, 96.4417267f, 282.651581f, 109.808517f);
pathB.lineTo(283.407959f, 110.462646f);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
}
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
static void skpadspert_de11(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(-4.4408921e-016f, 682.5f);
path.lineTo(30.5f, 682.5f);
path.cubicTo(32.709137f, 682.5f, 34.5f, 680.709167f, 34.5f, 678.5f);
path.lineTo(34.5f, 486.5f);
path.cubicTo(34.5f, 484.290863f, 32.709137f, 482.5f, 30.5f, 482.5f);
path.lineTo(0, 482.5f);
path.lineTo(-4.4408921e-016f, 682.5f);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 482);
pathB.lineTo(35, 482);
pathB.lineTo(35, 683);
pathB.lineTo(0, 683);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
}
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
static void skpaiaigames_com870(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(324.071075f, 845.071045f);
path.cubicTo(324.405151f, 844.737f, 324.715668f, 844.379395f, 325, 844.000977f);
path.lineTo(325, 842.127197f);
path.cubicTo(324.571411f, 842.956238f, 324.017761f, 843.710144f, 323.363953f, 844.363953f);
path.lineTo(324.071075f, 845.071045f);
path.close();
path.moveTo(323.363953f, 714.636047f);
path.lineTo(324.071075f, 713.928955f);
path.cubicTo(324.405151f, 714.263f, 324.715668f, 714.620605f, 325, 714.999023f);
path.lineTo(325, 716.872803f);
path.cubicTo(324.571411f, 716.043762f, 324.017761f, 715.289856f, 323.363953f, 714.636047f);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(317, 711);
pathB.cubicTo(322.522858f, 711, 327, 715.477173f, 327, 721);
pathB.lineTo(327, 838);
pathB.cubicTo(327, 843.522827f, 322.522858f, 848, 317, 848);
pathB.lineTo(155, 848);
pathB.cubicTo(149.477158f, 848, 145, 843.522827f, 145, 838);
pathB.lineTo(145, 721);
pathB.cubicTo(145, 715.477173f, 149.477158f, 711, 155, 711);
pathB.lineTo(317, 711);
pathB.close();
Path ops formerly found the topmost unprocessed edge and determined its angle sort order to initialize the winding. This never worked correctly with cubics and was flaky with paths consisting mostly of vertical edges. This replacement shoots axis-aligned rays through all intersecting edges to find the outermost one either horizontally or vertically. The resulting code is smaller and twice as fast. To support this, most of the horizontal / vertical intersection code was rewritten and standardized, and old code supporting the top-directed winding was deleted. Contours were pointed to by an SkTDArray. Instead, put them in a linked list, and designate the list head with its own class to ensure that methods that take lists of contours start at the top. This change removed a large percentage of memory allocations used by path ops. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1111333002
2015-05-11 14:21:27 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp92i(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0, 1);
path.cubicTo(2, 6, 4, 1, 5, 4);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1, 4);
pathB.cubicTo(4, 5, 1, 0, 6, 2);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp93d(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0, 1);
path.cubicTo(1, 6, 4, 1, 4, 3);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1, 4);
pathB.cubicTo(3, 4, 1, 0, 6, 1);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp94u(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(0, 3);
path.cubicTo(2, 3, 5, 0, 5, 3);
path.close();
pathB.setFillType(SkPath::kEvenOdd_FillType);
pathB.moveTo(0, 5);
pathB.cubicTo(3, 5, 3, 0, 3, 2);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kUnion_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void skpadbox_lt15(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(333.292084f, 624.570984f);
path.lineTo(614.229797f, 98.9735107f);
path.lineTo(933.457764f, 269.604431f);
path.lineTo(652.52002f, 795.201904f);
path.lineTo(333.292084f, 624.570984f);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(613.368042f, 100.585754f);
pathB.cubicTo(613.685303f, 99.9921265f, 614.423767f, 99.7681885f, 615.017395f, 100.085449f);
pathB.lineTo(932.633057f, 269.854553f);
pathB.cubicTo(933.226685f, 270.171875f, 933.450623f, 270.910278f, 933.133301f, 271.503906f);
pathB.lineTo(653.631897f, 794.414307f);
pathB.cubicTo(653.314636f, 795.007935f, 652.576172f, 795.231934f, 651.982544f, 794.914612f);
pathB.lineTo(334.366943f, 625.145508f);
pathB.cubicTo(333.773315f, 624.828247f, 333.549286f, 624.089783f, 333.866608f, 623.496155f);
pathB.lineTo(613.368042f, 100.585754f);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void skpadoption_org196(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(802, 367);
path.lineTo(802, 324);
path.lineTo(956, 324);
path.lineTo(956, 371);
path.quadTo(956, 373.071075f, 954.536011f, 374.536011f);
path.quadTo(953.071045f, 376, 951, 376);
path.lineTo(811, 376);
path.cubicTo(806.029419f, 376, 802, 371.970551f, 802, 367);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kInverseWinding_FillType);
pathB.moveTo(803, 326);
pathB.lineTo(955, 326);
pathB.lineTo(955, 370);
pathB.cubicTo(955, 372.761414f, 952.761414f, 375, 950, 375);
pathB.lineTo(808, 375);
pathB.cubicTo(805.238586f, 375, 803, 372.761414f, 803, 370);
pathB.lineTo(803, 326);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void skpadspert_net23(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(-2.220446e-018f, 483.5f);
path.lineTo(0, 482.5f);
path.lineTo(30.5f, 482.5f);
path.cubicTo(32.709137f, 482.5f, 34.5f, 484.290863f, 34.5f, 486.5f);
path.lineTo(34.5f, 678.5f);
path.cubicTo(34.5f, 680.709167f, 32.709137f, 682.5f, 30.5f, 682.5f);
path.lineTo(-4.4408921e-016f, 682.5f);
path.lineTo(-4.41868766e-016f, 681.5f);
path.lineTo(30.5f, 681.5f);
path.cubicTo(32.1568565f, 681.5f, 33.5f, 680.15686f, 33.5f, 678.5f);
path.lineTo(33.5f, 486.5f);
path.cubicTo(33.5f, 484.84314f, 32.1568565f, 483.5f, 30.5f, 483.5f);
path.lineTo(-2.220446e-018f, 483.5f);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 482);
pathB.lineTo(35, 482);
pathB.lineTo(35, 683);
pathB.lineTo(0, 683);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void skpadventistmission_org572(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(1182.00037f, 926);
path.cubicTo(1181.08813f, 924.785583f, 1179.63586f, 924, 1178, 924);
path.lineTo(938, 924);
path.cubicTo(936.364197f, 924, 934.911865f, 924.785583f, 933.999634f, 926);
path.lineTo(1182.00037f, 926);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(934, 924);
pathB.lineTo(1182, 924);
pathB.lineTo(1182, 926);
pathB.lineTo(934, 926);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void skpagentxsites_com55(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(925, 27);
path.cubicTo(924.447693f, 27, 924, 27.4477158f, 924, 28);
path.lineTo(924, 55);
path.cubicTo(924, 55.5522842f, 924.447693f, 56, 925, 56);
path.lineTo(1103, 56);
path.cubicTo(1103.55225f, 56, 1104, 55.5522842f, 1104, 55);
path.lineTo(1104, 28);
path.cubicTo(1104, 27.4477158f, 1103.55225f, 27, 1103, 27);
path.lineTo(925, 27);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1103, 27);
pathB.cubicTo(1104.10461f, 27, 1105, 27.8954315f, 1105, 29);
pathB.lineTo(1105, 54);
pathB.cubicTo(1105, 55.1045685f, 1104.10461f, 56, 1103, 56);
pathB.lineTo(926, 56);
pathB.cubicTo(924.895447f, 56, 924, 55.1045685f, 924, 54);
pathB.lineTo(924, 29);
pathB.cubicTo(924, 27.8954315f, 924.895447f, 27, 926, 27);
pathB.lineTo(1103, 27);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void skpbakosoft_com10(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(190, 170);
path.cubicTo(178.9543f, 170, 170, 178.9543f, 170, 190);
path.cubicTo(170, 201.0457f, 178.9543f, 210, 190, 210);
path.lineTo(370, 210);
path.cubicTo(381.045685f, 210, 390, 201.0457f, 390, 190);
path.cubicTo(390, 178.9543f, 381.045685f, 170, 370, 170);
path.lineTo(190, 170);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(210, 190);
pathB.quadTo(210, 198.284271f, 204.142136f, 204.142136f);
pathB.quadTo(198.284271f, 210, 190, 210);
pathB.quadTo(181.715729f, 210, 175.857864f, 204.142136f);
pathB.quadTo(170, 198.284271f, 170, 190);
pathB.quadTo(170, 181.715729f, 175.857864f, 175.857864f);
pathB.quadTo(181.715729f, 170, 190, 170);
pathB.quadTo(198.284271f, 170, 204.142136f, 175.857864f);
pathB.quadTo(210, 181.715729f, 210, 190);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void skpbambootheme_com12(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(47.8780937f, 58);
path.lineTo(0, 58);
path.lineTo(-8.65973959e-015f, 96.9914017f);
path.quadTo(20.0654926f, 96.6451874f, 34.3553391f, 82.3553391f);
path.quadTo(44.9466133f, 71.764061f, 47.8780937f, 58);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kEvenOdd_FillType);
pathB.moveTo(-1, -3);
pathB.lineTo(-1, -3);
pathB.cubicTo(26.6142502f, -3, 49, 19.3857498f, 49, 47);
pathB.lineTo(49, 47);
pathB.cubicTo(49, 74.6142502f, 26.6142502f, 97, -1, 97);
pathB.lineTo(-1, 97);
pathB.cubicTo(-28.6142502f, 97, -51, 74.6142502f, -51, 47);
pathB.lineTo(-51, 47);
pathB.cubicTo(-51, 19.3857498f, -28.6142502f, -3, -1, -3);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void skpakmmos_ru100(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(693.000488f, 926);
path.cubicTo(692.164734f, 925.37207f, 691.125793f, 925, 690, 925);
path.lineTo(578, 925);
path.cubicTo(576.874207f, 925, 575.835266f, 925.37207f, 574.999512f, 926);
path.lineTo(693.000488f, 926);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(575, 925);
pathB.lineTo(693, 925);
pathB.lineTo(693, 926);
pathB.lineTo(575, 926);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
}
path ops work in progress make more skps work remove edit files BUG= Review URL: https://codereview.chromium.org/23542056 git-svn-id: http://skia.googlecode.com/svn/trunk@11570 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-02 14:49:34 +00:00
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void skpcarpetplanet_ru22(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(195, 785);
path.cubicTo(124.307556f, 785, 67, 841.859863f, 67, 912);
path.lineTo(67, 913);
path.cubicTo(67, 917.388916f, 67.2243805f, 921.725769f, 67.662384f, 926);
path.lineTo(322, 926);
path.lineTo(322, 896.048035f);
path.cubicTo(314.09201f, 833.437622f, 260.247131f, 785, 195, 785);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(195, 785);
pathB.cubicTo(265.140167f, 785, 322, 842.307556f, 322, 913);
pathB.cubicTo(322, 983.692444f, 265.140167f, 1041, 195, 1041);
pathB.lineTo(194, 1041);
pathB.cubicTo(123.85984f, 1041, 67, 983.692444f, 67, 913);
pathB.cubicTo(67, 842.307556f, 123.85984f, 785, 194, 785);
pathB.lineTo(195, 785);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void skpcarrot_is24(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(945, 597);
path.quadTo(913.93396f, 597, 891.96698f, 618.96698f);
path.quadTo(870, 640.93396f, 870, 672);
path.quadTo(870, 703.06604f, 891.96698f, 725.03302f);
path.quadTo(913.93396f, 747, 945, 747);
path.quadTo(976.06604f, 747, 998.03302f, 725.03302f);
path.quadTo(1020, 703.06604f, 1020, 672);
path.quadTo(1020, 640.93396f, 998.03302f, 618.96698f);
path.quadTo(976.06604f, 597, 945, 597);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(945.080994f, 597.161987f);
pathB.cubicTo(903.659973f, 597.161987f, 870.080994f, 630.73999f, 870.080994f, 672.161987f);
pathB.cubicTo(870.080994f, 676.096008f, 870.387024f, 679.957031f, 870.971008f, 683.726013f);
pathB.cubicTo(876.53302f, 719.656006f, 907.593994f, 747.161987f, 945.080994f, 747.161987f);
pathB.cubicTo(982.567993f, 747.161987f, 1013.62903f, 719.656006f, 1019.19104f, 683.726013f);
pathB.cubicTo(1019.77502f, 679.955017f, 1020.08099f, 676.094971f, 1020.08099f, 672.161987f);
pathB.cubicTo(1020.08002f, 630.73999f, 986.502014f, 597.161987f, 945.080994f, 597.161987f);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void skpbangalorenest_com4(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(0, 926);
path.lineTo(0, 0);
path.lineTo(1265, 0);
path.lineTo(1265, 926);
path.lineTo(0, 926);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 290);
pathB.lineTo(-2.64514972e-014f, 146);
pathB.lineTo(30, 146);
pathB.lineTo(30, 290);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void skpbenzoteh_ru152(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(883, 23);
path.lineTo(883, 0);
path.lineTo(1122.5f, 0);
path.lineTo(1122.5f, 25.2136822f);
path.quadTo(1122.14441f, 25.9271851f, 1121.53601f, 26.5359993f);
path.quadTo(1120.07104f, 28, 1118, 28);
path.lineTo(888, 28);
path.quadTo(885.928955f, 28, 884.463989f, 26.5359993f);
path.quadTo(883, 25.0710678f, 883, 23);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(883, 0);
pathB.lineTo(1123, 0);
pathB.lineTo(1123, 23);
pathB.quadTo(1123, 25.0710678f, 1121.53601f, 26.5359993f);
pathB.quadTo(1120.07104f, 28, 1118, 28);
pathB.lineTo(888, 28);
pathB.quadTo(885.928955f, 28, 884.463989f, 26.5359993f);
pathB.quadTo(883, 25.0710678f, 883, 23);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void skpbestred_ru37(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(883, 23);
path.lineTo(883, 0);
path.lineTo(1122.5f, 0);
path.lineTo(1122.5f, 25.2136822f);
path.quadTo(1122.14441f, 25.9271851f, 1121.53601f, 26.5359993f);
path.quadTo(1120.07104f, 28, 1118, 28);
path.lineTo(888, 28);
path.quadTo(885.928955f, 28, 884.463989f, 26.5359993f);
path.quadTo(883, 25.0710678f, 883, 23);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(883, 0);
pathB.lineTo(1123, 0);
pathB.lineTo(1123, 23);
pathB.quadTo(1123, 25.0710678f, 1121.53601f, 26.5359993f);
pathB.quadTo(1120.07104f, 28, 1118, 28);
pathB.lineTo(888, 28);
pathB.quadTo(885.928955f, 28, 884.463989f, 26.5359993f);
pathB.quadTo(883, 25.0710678f, 883, 23);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void skpbingoentertainment_net189(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(896, 745.38678f);
path.lineTo(896, 873.38678f);
path.lineTo(922.567993f, 876.683716f);
path.lineTo(922.567993f, 748.683716f);
path.lineTo(896, 745.38678f);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(899.200928f, 745.783997f);
pathB.cubicTo(897.119385f, 745.525696f, 895.432007f, 752.031982f, 895.432007f, 760.316284f);
pathB.lineTo(895.432007f, 858.316284f);
pathB.cubicTo(895.432007f, 866.600586f, 897.119385f, 873.525696f, 899.200928f, 873.783997f);
pathB.lineTo(918.799133f, 876.216003f);
pathB.cubicTo(920.880615f, 876.474304f, 922.567993f, 869.968018f, 922.567993f, 861.683716f);
pathB.lineTo(922.567993f, 763.683716f);
pathB.cubicTo(922.567993f, 755.399414f, 920.880615f, 748.474304f, 918.799133f, 748.216003f);
pathB.lineTo(899.200928f, 745.783997f);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void skpcarrefour_ro62(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(1104, 453);
path.lineTo(399, 453);
path.lineTo(399, 657);
path.cubicTo(399, 661.970581f, 403.029449f, 666, 408, 666);
path.lineTo(1095, 666);
path.cubicTo(1099.97058f, 666, 1104, 661.970581f, 1104, 657);
path.lineTo(1104, 453);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kInverseWinding_FillType);
pathB.moveTo(400, 453);
pathB.lineTo(1103, 453);
pathB.lineTo(1103, 666);
pathB.lineTo(406, 666);
pathB.cubicTo(402.686279f, 666, 400, 663.313721f, 400, 660);
pathB.lineTo(400, 453);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void skpcaffelavazzait_com_ua21(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(883, 23);
path.lineTo(883, 0);
path.lineTo(1122.5f, 0);
path.lineTo(1122.5f, 25.2136822f);
path.quadTo(1122.14441f, 25.9271851f, 1121.53601f, 26.5359993f);
path.quadTo(1120.07104f, 28, 1118, 28);
path.lineTo(888, 28);
path.quadTo(885.928955f, 28, 884.463989f, 26.5359993f);
path.quadTo(883, 25.0710678f, 883, 23);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(883, 0);
pathB.lineTo(1123, 0);
pathB.lineTo(1123, 23);
pathB.quadTo(1123, 25.0710678f, 1121.53601f, 26.5359993f);
pathB.quadTo(1120.07104f, 28, 1118, 28);
pathB.lineTo(888, 28);
pathB.quadTo(885.928955f, 28, 884.463989f, 26.5359993f);
pathB.quadTo(883, 25.0710678f, 883, 23);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void skpcamcorder_kz21(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(883, 23);
path.lineTo(883, 0);
path.lineTo(1122.5f, 0);
path.lineTo(1122.5f, 25.2136822f);
path.quadTo(1122.14441f, 25.9271851f, 1121.53601f, 26.5359993f);
path.quadTo(1120.07104f, 28, 1118, 28);
path.lineTo(888, 28);
path.quadTo(885.928955f, 28, 884.463989f, 26.5359993f);
path.quadTo(883, 25.0710678f, 883, 23);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(883, 0);
pathB.lineTo(1123, 0);
pathB.lineTo(1123, 23);
pathB.quadTo(1123, 25.0710678f, 1121.53601f, 26.5359993f);
pathB.quadTo(1120.07104f, 28, 1118, 28);
pathB.lineTo(888, 28);
pathB.quadTo(885.928955f, 28, 884.463989f, 26.5359993f);
pathB.quadTo(883, 25.0710678f, 883, 23);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void skpcavablar_net563(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(160.000488f, 918);
path.cubicTo(159.164749f, 917.37207f, 158.125824f, 917, 157, 917);
path.lineTo(94, 917);
path.cubicTo(92.874176f, 917, 91.8352661f, 917.37207f, 90.9995193f, 918);
path.lineTo(160.000488f, 918);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(91, 917);
pathB.lineTo(160, 917);
pathB.lineTo(160, 918);
pathB.lineTo(91, 918);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void skpinsomnia_gr72(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(1138, 231);
path.lineTo(1137, 243.625748f);
path.lineTo(1137, 926);
path.lineTo(1139, 926);
path.lineTo(1139, 231);
path.lineTo(1138, 231);
path.close();
SkPath pathB;
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1139, 231);
pathB.lineTo(1138, 231);
pathB.lineTo(633, 6101);
pathB.lineTo(1139, 6607);
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp95u(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(0, 2);
path.cubicTo(2, 3, 5, 1, 3, 2);
path.close();
pathB.setFillType(SkPath::kEvenOdd_FillType);
pathB.moveTo(1, 5);
pathB.cubicTo(2, 3, 2, 0, 3, 2);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kUnion_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp96d(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@11585 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-03 07:01:37 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(1, 6);
path.cubicTo(0, 3, 6, 3, 5, 0);
path.close();
pathB.setFillType(SkPath::kEvenOdd_FillType);
pathB.moveTo(3, 6);
pathB.cubicTo(0, 5, 6, 1, 3, 0);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
path ops work in progress make more skps work remove edit files BUG= Review URL: https://codereview.chromium.org/23542056 git-svn-id: http://skia.googlecode.com/svn/trunk@11570 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-02 14:49:34 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp97x(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@12101 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-11-02 07:02:02 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(0, 2);
path.cubicTo(0, 6, 2, 1, 2, 1);
path.close();
pathB.setFillType(SkPath::kEvenOdd_FillType);
pathB.moveTo(1, 2);
pathB.cubicTo(1, 2, 2, 0, 6, 0);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kXOR_SkPathOp, filename);
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@12101 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-11-02 07:02:02 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp98x(skiatest::Reporter* reporter, const char* filename) {
Sanitizing source files in Housekeeper-Nightly git-svn-id: http://skia.googlecode.com/svn/trunk@12101 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-11-02 07:02:02 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(0, 3);
path.cubicTo(3, 6, 4, 1, 6, 3);
path.close();
pathB.setFillType(SkPath::kEvenOdd_FillType);
pathB.moveTo(1, 4);
pathB.cubicTo(3, 6, 3, 0, 6, 3);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kXOR_SkPathOp, filename);
pathops work in progress BUG= Review URL: https://codereview.chromium.org/52653002 git-svn-id: http://skia.googlecode.com/svn/trunk@12089 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-11-01 17:36:03 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp99(skiatest::Reporter* reporter, const char* filename) {
optimize pathops coverage Remove unused code from SkOpSegment.cpp and friends. Add new tests exposed by coverage. Fix a bug exposed by coverage -- removing the need to detect points that are nearby when intersecting. Add gyp rule for building coverage flavor on Mac. R=mtklein@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/75453003 git-svn-id: http://skia.googlecode.com/svn/trunk@12344 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-11-21 17:04:29 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(3,6);
path.cubicTo(0,3, 6,5, 5,4);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(5,6);
pathB.cubicTo(4,5, 6,3, 3,0);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
optimize pathops coverage Remove unused code from SkOpSegment.cpp and friends. Add new tests exposed by coverage. Fix a bug exposed by coverage -- removing the need to detect points that are nearby when intersecting. Add gyp rule for building coverage flavor on Mac. R=mtklein@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/75453003 git-svn-id: http://skia.googlecode.com/svn/trunk@12344 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-11-21 17:04:29 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp100(skiatest::Reporter* reporter, const char* filename) {
optimize pathops coverage Remove unused code from SkOpSegment.cpp and friends. Add new tests exposed by coverage. Fix a bug exposed by coverage -- removing the need to detect points that are nearby when intersecting. Add gyp rule for building coverage flavor on Mac. R=mtklein@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/75453003 git-svn-id: http://skia.googlecode.com/svn/trunk@12344 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-11-21 17:04:29 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(0,2, 2,1, 4,2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1,2);
pathB.cubicTo(2,4, 1,0, 2,0);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
optimize pathops coverage Remove unused code from SkOpSegment.cpp and friends. Add new tests exposed by coverage. Fix a bug exposed by coverage -- removing the need to detect points that are nearby when intersecting. Add gyp rule for building coverage flavor on Mac. R=mtklein@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/75453003 git-svn-id: http://skia.googlecode.com/svn/trunk@12344 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-11-21 17:04:29 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void cubicOp101(skiatest::Reporter* reporter, const char* filename) {
optimize pathops coverage Remove unused code from SkOpSegment.cpp and friends. Add new tests exposed by coverage. Fix a bug exposed by coverage -- removing the need to detect points that are nearby when intersecting. Add gyp rule for building coverage flavor on Mac. R=mtklein@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/75453003 git-svn-id: http://skia.googlecode.com/svn/trunk@12344 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-11-21 17:04:29 +00:00
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0, 1);
path.cubicTo(2, 3, 2, 1, 5, 3);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1, 2);
pathB.cubicTo(3, 5, 1, 0, 3, 2);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
}
static void cubicOp102(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(1,2, 1,0, 3,0);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,1);
pathB.cubicTo(0,3, 1,0, 2,1);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
}
static void cubicOp103(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(1,5, 2,0, 2,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,2);
pathB.cubicTo(1,2, 1,0, 5,1);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
}
static void cubicOp104(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(0,6, 4,0, 6,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,4);
pathB.cubicTo(1,6, 1,0, 6,0);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
}
static void cubicOp105(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(0,4, 6,5, 2,0);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(5,6);
pathB.cubicTo(0,2, 1,0, 4,0);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
}
static void cubicOp106(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0, 1);
path.cubicTo(4, 6, 2, 1, 2, 0);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1, 2);
pathB.cubicTo(0, 2, 1, 0, 6, 4);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
}
static void cubicOp107(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0, 1);
path.cubicTo(4, 6, 2, 1, 2, 0);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1, 2);
pathB.cubicTo(0, 2, 1, 0, 6, 4);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
}
static void cubicOp108(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0, 1);
path.cubicTo(4, 6, 2, 1, 2, 0);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1, 2);
pathB.cubicTo(0, 2, 1, 0, 6, 4);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kUnion_SkPathOp, filename);
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
}
static void cubicOp109(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(4,5, 6,3, 5,4);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(3,6);
pathB.cubicTo(4,5, 1,0, 5,4);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
}
static void cubicOp110(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kEvenOdd_FillType);
path.addRect(0, 0, 1, 1, SkPath::kCW_Direction);
path.addRect(0, 0, 4, 4, SkPath::kCW_Direction);
pathB.setFillType(SkPath::kEvenOdd_FillType);
pathB.addRect(0, 0, 2, 2, SkPath::kCW_Direction);
pathB.addRect(0, 0, 2, 2, SkPath::kCW_Direction);
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
}
static void cubicOp111(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(1,4);
path.cubicTo(0,5, 4,1, 3,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1,4);
pathB.cubicTo(1,3, 4,1, 5,0);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
}
static void xOp1u(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(1, 4);
path.cubicTo(4, 5, 3, 2, 6, 3);
path.close();
pathB.setFillType(SkPath::kEvenOdd_FillType);
pathB.moveTo(2, 3);
pathB.cubicTo(3, 6, 4, 1, 5, 4);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kUnion_SkPathOp, filename);
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
}
static void xOp1i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(1, 4);
path.cubicTo(1, 5, 6, 0, 5, 1);
path.close();
pathB.setFillType(SkPath::kEvenOdd_FillType);
pathB.moveTo(0, 6);
pathB.cubicTo(1, 5, 4, 1, 5, 1);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
}
static void xOp2i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(1, 5);
path.cubicTo(0, 4, 3, 2, 6, 1);
path.close();
pathB.setFillType(SkPath::kEvenOdd_FillType);
pathB.moveTo(2, 3);
pathB.cubicTo(1, 6, 5, 1, 4, 0);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
}
static void xOp3i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(1,4);
path.cubicTo(0,5, 4,1, 3,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1,4);
pathB.cubicTo(1,3, 4,1, 5,0);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
}
static void findFirst1(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(1,6, 5,0, 2,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,5);
pathB.cubicTo(1,2, 1,0, 6,1);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
optimize pathops coverage Remove unused code from SkOpSegment.cpp and friends. Add new tests exposed by coverage. Fix a bug exposed by coverage -- removing the need to detect points that are nearby when intersecting. Add gyp rule for building coverage flavor on Mac. R=mtklein@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/75453003 git-svn-id: http://skia.googlecode.com/svn/trunk@12344 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-11-21 17:04:29 +00:00
}
fix minor skp-found bugs remove globals from pathops_unittest BUG=skia:2460 TBR=mtklein Author: caryclark@google.com Review URL: https://codereview.chromium.org/239563004 git-svn-id: http://skia.googlecode.com/svn/trunk@14378 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-25 12:59:11 +00:00
static void cubicOp112(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(2,4);
path.cubicTo(2,3, 6,4, 1,0);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(4,6);
pathB.cubicTo(0,1, 4,2, 3,2);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
fix minor skp-found bugs remove globals from pathops_unittest BUG=skia:2460 TBR=mtklein Author: caryclark@google.com Review URL: https://codereview.chromium.org/239563004 git-svn-id: http://skia.googlecode.com/svn/trunk@14378 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-25 12:59:11 +00:00
}
static void cubicOp113(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.moveTo(2,4);
path.cubicTo(3,5, 2.33333325f,4.33333349f, 3.83333325f,3.83333349f);
path.close();
pathB.moveTo(3,5);
pathB.cubicTo(2.33333325f,4.33333349f, 3.83333325f,3.83333349f, 2,4);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
fix minor skp-found bugs remove globals from pathops_unittest BUG=skia:2460 TBR=mtklein Author: caryclark@google.com Review URL: https://codereview.chromium.org/239563004 git-svn-id: http://skia.googlecode.com/svn/trunk@14378 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-25 12:59:11 +00:00
}
static void cubicOp114(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0, 1);
path.cubicTo(1, 3, -1, 2, 3.5f, 1.33333337f);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1, 3);
pathB.cubicTo(-1, 2, 3.5f, 1.33333337f, 0, 1);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
fix minor skp-found bugs remove globals from pathops_unittest BUG=skia:2460 TBR=mtklein Author: caryclark@google.com Review URL: https://codereview.chromium.org/239563004 git-svn-id: http://skia.googlecode.com/svn/trunk@14378 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-25 12:59:11 +00:00
}
Fix last pathops skp bug This fixes the last bug discovered by iterating through the 800K skp corpus representing the top 1M websites. For every clip on the stack, the paths are replaced with the pathop intersection. The resulting draw is compared with the original draw for pixel errors. At least two prominent bugs remain. In one, the winding value is confused by a cubic with an inflection. In the other, a quad/cubic pair, nearly coincident, fails to find an intersection. These minor changes include ignoring very tiny self-intersections of cubics, and processing degenerate edges that don't connect to anything else. R=reed@android.com TBR=reed Author: caryclark@google.com Review URL: https://codereview.chromium.org/340103002
2014-06-18 14:24:19 +00:00
static void cubicOp114asQuad(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0, 1);
path.cubicTo(1, 3, -1, 2, 3.5f, 1.33333337f);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1, 3);
pathB.cubicTo(-1, 2, 3.5f, 1.33333337f, 0, 1);
pathB.close();
SkPath qPath, qPathB;
CubicPathToQuads(path, &qPath);
CubicPathToQuads(pathB, &qPathB);
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, qPath, qPathB, kIntersect_SkPathOp, filename);
Fix last pathops skp bug This fixes the last bug discovered by iterating through the 800K skp corpus representing the top 1M websites. For every clip on the stack, the paths are replaced with the pathop intersection. The resulting draw is compared with the original draw for pixel errors. At least two prominent bugs remain. In one, the winding value is confused by a cubic with an inflection. In the other, a quad/cubic pair, nearly coincident, fails to find an intersection. These minor changes include ignoring very tiny self-intersections of cubics, and processing degenerate edges that don't connect to anything else. R=reed@android.com TBR=reed Author: caryclark@google.com Review URL: https://codereview.chromium.org/340103002
2014-06-18 14:24:19 +00:00
}
fix minor skp-found bugs remove globals from pathops_unittest BUG=skia:2460 TBR=mtklein Author: caryclark@google.com Review URL: https://codereview.chromium.org/239563004 git-svn-id: http://skia.googlecode.com/svn/trunk@14378 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-25 12:59:11 +00:00
static void quadOp10i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.moveTo(0, 0);
path.quadTo(1, 8, 3, 5);
path.lineTo(8, 1);
path.close();
pathB.moveTo(0, 0);
pathB.quadTo(8, 1, 4, 8);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
fix minor skp-found bugs remove globals from pathops_unittest BUG=skia:2460 TBR=mtklein Author: caryclark@google.com Review URL: https://codereview.chromium.org/239563004 git-svn-id: http://skia.googlecode.com/svn/trunk@14378 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-25 12:59:11 +00:00
}
fix cubic/line intersection; add skp tests BUG=skia:2488 TBR=reed Author: caryclark@google.com Review URL: https://codereview.chromium.org/252243003 git-svn-id: http://skia.googlecode.com/svn/trunk@14458 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-30 13:37:48 +00:00
static void kari1(skiatest::Reporter* reporter, const char* filename) {
SkPath path1;
path1.moveTo(39.9375, -5.8359375);
path1.lineTo(40.625, -5.7890625);
path1.lineTo(37.7109375, 1.3515625);
path1.lineTo(37.203125, 0.9609375);
path1.close();
SkPath path2;
path2.moveTo(37.52734375f, -1.44140625f);
path2.cubicTo(37.8736991882324f, -1.69921875f, 38.1640625f, -2.140625f, 38.3984375f, -2.765625f);
path2.lineTo(38.640625f, -2.609375f);
path2.cubicTo(38.53125f, -1.89583337306976f, 38.0664443969727f, -0.154893040657043f, 38.0664443969727f, -0.154893040657043f);
path2.cubicTo(38.0664443969727f, -0.154893040657043f, 37.1809883117676f, -1.18359375f, 37.52734375, -1.44140625f);
path2.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path1, path2, kDifference_SkPathOp, filename);
fix cubic/line intersection; add skp tests BUG=skia:2488 TBR=reed Author: caryclark@google.com Review URL: https://codereview.chromium.org/252243003 git-svn-id: http://skia.googlecode.com/svn/trunk@14458 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-30 13:37:48 +00:00
}
When solving the cubic line intersection directly fails, use binary search as a fallback. The cubic line intersection math empirically works 99.99% of the time (fails 3100 out of 1B random tests) but when it fails, an intersection may be missed altogether. The binary search is may not find a solution if the cubic line failed to find any solutions at all, but so far that case hasn't arisen. BUG=skia:2504 TBR=reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/266063003 git-svn-id: http://skia.googlecode.com/svn/trunk@14614 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-05-07 15:31:40 +00:00
static void issue2504(skiatest::Reporter* reporter, const char* filename) {
SkPath path1;
path1.moveTo(34.2421875, -5.976562976837158203125);
path1.lineTo(35.453121185302734375, 0);
path1.lineTo(31.9375, 0);
path1.close();
SkPath path2;
path2.moveTo(36.71843719482421875, 0.8886508941650390625);
path2.cubicTo(36.71843719482421875, 0.8886508941650390625,
35.123386383056640625, 0.554015457630157470703125,
34.511409759521484375, -0.1152553558349609375);
path2.cubicTo(33.899425506591796875, -0.7845261096954345703125,
34.53484344482421875, -5.6777553558349609375,
34.53484344482421875, -5.6777553558349609375);
path2.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path1, path2, kUnion_SkPathOp, filename);
When solving the cubic line intersection directly fails, use binary search as a fallback. The cubic line intersection math empirically works 99.99% of the time (fails 3100 out of 1B random tests) but when it fails, an intersection may be missed altogether. The binary search is may not find a solution if the cubic line failed to find any solutions at all, but so far that case hasn't arisen. BUG=skia:2504 TBR=reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/266063003 git-svn-id: http://skia.googlecode.com/svn/trunk@14614 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-05-07 15:31:40 +00:00
}
Enabling the canvas bit to turn the clip stack into a flat replace exposed around 100 failures when testing the 800K skp set generated from the top 1M web sites. This fixes all but one of those failures. Major changes include: - Replace angle indices with angle pointers. This was motivated by the need to add angles later but not renumber existing angles. - Aggressive segment chase. When the winding is known on a segment, more aggressively passing that winding to adjacent segments allows fragmented data sets to succeed. - Line segments with ends nearly the same are treated as coincident first. - Transfer partial coincidence by observing that if segment A is partially coincident to B and C then B and C may be partially coincident. TBR=reed Author: caryclark@google.com Review URL: https://codereview.chromium.org/272153002
2014-06-17 12:15:38 +00:00
static void issue2540(skiatest::Reporter* reporter, const char* filename) {
SkPath path1;
path1.moveTo(26.5054988861083984375, 85.73960113525390625);
path1.cubicTo(84.19739532470703125, 17.77140045166015625, 16.93920135498046875, 101.86199951171875, 12.631000518798828125, 105.24700164794921875);
path1.cubicTo(11.0819997787475585937500000, 106.46399688720703125, 11.5260000228881835937500000, 104.464996337890625, 11.5260000228881835937500000, 104.464996337890625);
path1.lineTo(23.1654987335205078125, 89.72879791259765625);
path1.cubicTo(23.1654987335205078125, 89.72879791259765625, -10.1713008880615234375, 119.9160003662109375, -17.1620006561279296875, 120.8249969482421875);
path1.cubicTo(-19.1149997711181640625, 121.07900238037109375, -18.0380001068115234375, 119.79299163818359375, -18.0380001068115234375, 119.79299163818359375);
path1.cubicTo(-18.0380001068115234375, 119.79299163818359375, 14.22100067138671875, 90.60700225830078125, 26.5054988861083984375, 85.73960113525390625);
path1.close();
SkPath path2;
path2.moveTo(-25.077999114990234375, 124.9120025634765625);
path2.cubicTo(-25.077999114990234375, 124.9120025634765625, -25.9509983062744140625, 125.95400238037109375, -24.368999481201171875, 125.7480010986328125);
path2.cubicTo(-16.06999969482421875, 124.66899871826171875, 1.2680000066757202148437500, 91.23999786376953125, 37.264003753662109375, 95.35400390625);
path2.cubicTo(37.264003753662109375, 95.35400390625, 11.3710002899169921875, 83.7339935302734375, -25.077999114990234375, 124.9120025634765625);
path2.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path1, path2, kUnion_SkPathOp, filename);
Enabling the canvas bit to turn the clip stack into a flat replace exposed around 100 failures when testing the 800K skp set generated from the top 1M web sites. This fixes all but one of those failures. Major changes include: - Replace angle indices with angle pointers. This was motivated by the need to add angles later but not renumber existing angles. - Aggressive segment chase. When the winding is known on a segment, more aggressively passing that winding to adjacent segments allows fragmented data sets to succeed. - Line segments with ends nearly the same are treated as coincident first. - Transfer partial coincidence by observing that if segment A is partially coincident to B and C then B and C may be partially coincident. TBR=reed Author: caryclark@google.com Review URL: https://codereview.chromium.org/272153002
2014-06-17 12:15:38 +00:00
}
static void rects1(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(0, 0);
path.lineTo(1, 0);
path.lineTo(1, 1);
path.lineTo(0, 1);
path.close();
path.moveTo(0, 0);
path.lineTo(6, 0);
path.lineTo(6, 6);
path.lineTo(0, 6);
path.close();
pathB.setFillType(SkPath::kEvenOdd_FillType);
pathB.moveTo(0, 0);
pathB.lineTo(1, 0);
pathB.lineTo(1, 1);
pathB.lineTo(0, 1);
pathB.close();
pathB.moveTo(0, 0);
pathB.lineTo(2, 0);
pathB.lineTo(2, 2);
pathB.lineTo(0, 2);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kUnion_SkPathOp, filename);
Enabling the canvas bit to turn the clip stack into a flat replace exposed around 100 failures when testing the 800K skp set generated from the top 1M web sites. This fixes all but one of those failures. Major changes include: - Replace angle indices with angle pointers. This was motivated by the need to add angles later but not renumber existing angles. - Aggressive segment chase. When the winding is known on a segment, more aggressively passing that winding to adjacent segments allows fragmented data sets to succeed. - Line segments with ends nearly the same are treated as coincident first. - Transfer partial coincidence by observing that if segment A is partially coincident to B and C then B and C may be partially coincident. TBR=reed Author: caryclark@google.com Review URL: https://codereview.chromium.org/272153002
2014-06-17 12:15:38 +00:00
}
static void rects2(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(0, 0);
path.lineTo(4, 0);
path.lineTo(4, 4);
path.lineTo(0, 4);
path.close();
path.moveTo(3, 3);
path.lineTo(4, 3);
path.lineTo(4, 4);
path.lineTo(3, 4);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(3, 3);
pathB.lineTo(6, 3);
pathB.lineTo(6, 6);
pathB.lineTo(3, 6);
pathB.close();
pathB.moveTo(3, 3);
pathB.lineTo(4, 3);
pathB.lineTo(4, 4);
pathB.lineTo(3, 4);
pathB.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Enabling the canvas bit to turn the clip stack into a flat replace exposed around 100 failures when testing the 800K skp set generated from the top 1M web sites. This fixes all but one of those failures. Major changes include: - Replace angle indices with angle pointers. This was motivated by the need to add angles later but not renumber existing angles. - Aggressive segment chase. When the winding is known on a segment, more aggressively passing that winding to adjacent segments allows fragmented data sets to succeed. - Line segments with ends nearly the same are treated as coincident first. - Transfer partial coincidence by observing that if segment A is partially coincident to B and C then B and C may be partially coincident. TBR=reed Author: caryclark@google.com Review URL: https://codereview.chromium.org/272153002
2014-06-17 12:15:38 +00:00
}
static void rects3(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kEvenOdd_FillType);
path.addRect(0, 0, 1, 1, SkPath::kCW_Direction);
path.addRect(0, 0, 4, 4, SkPath::kCW_Direction);
pathB.setFillType(SkPath::kWinding_FillType);
pathB.addRect(0, 0, 2, 2, SkPath::kCW_Direction);
pathB.addRect(0, 0, 2, 2, SkPath::kCW_Direction);
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Enabling the canvas bit to turn the clip stack into a flat replace exposed around 100 failures when testing the 800K skp set generated from the top 1M web sites. This fixes all but one of those failures. Major changes include: - Replace angle indices with angle pointers. This was motivated by the need to add angles later but not renumber existing angles. - Aggressive segment chase. When the winding is known on a segment, more aggressively passing that winding to adjacent segments allows fragmented data sets to succeed. - Line segments with ends nearly the same are treated as coincident first. - Transfer partial coincidence by observing that if segment A is partially coincident to B and C then B and C may be partially coincident. TBR=reed Author: caryclark@google.com Review URL: https://codereview.chromium.org/272153002
2014-06-17 12:15:38 +00:00
}
static void rects4(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kEvenOdd_FillType);
path.addRect(0, 0, 1, 1, SkPath::kCW_Direction);
path.addRect(0, 0, 2, 2, SkPath::kCW_Direction);
pathB.setFillType(SkPath::kWinding_FillType);
pathB.addRect(0, 0, 2, 2, SkPath::kCW_Direction);
pathB.addRect(0, 0, 3, 3, SkPath::kCW_Direction);
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Enabling the canvas bit to turn the clip stack into a flat replace exposed around 100 failures when testing the 800K skp set generated from the top 1M web sites. This fixes all but one of those failures. Major changes include: - Replace angle indices with angle pointers. This was motivated by the need to add angles later but not renumber existing angles. - Aggressive segment chase. When the winding is known on a segment, more aggressively passing that winding to adjacent segments allows fragmented data sets to succeed. - Line segments with ends nearly the same are treated as coincident first. - Transfer partial coincidence by observing that if segment A is partially coincident to B and C then B and C may be partially coincident. TBR=reed Author: caryclark@google.com Review URL: https://codereview.chromium.org/272153002
2014-06-17 12:15:38 +00:00
}
update pathops core and tests split out skpclip (the test of 1M pictures) into its own project TBR=reed Author: caryclark@google.com Review URL: https://codereview.chromium.org/400033002
2014-07-18 12:08:14 +00:00
static void issue2753(skiatest::Reporter* reporter, const char* filename) {
SkPath path1;
path1.moveTo(142.701f, 110.568f);
path1.lineTo(142.957f, 100);
path1.lineTo(153.835f, 100);
path1.lineTo(154.592f, 108.188f);
path1.cubicTo(154.592f, 108.188f, 153.173f, 108.483f, 152.83f, 109.412f);
path1.cubicTo(152.83f, 109.412f, 142.701f, 110.568f, 142.701f, 110.568f);
path1.close();
SkPath path2;
path2.moveTo(39, 124.001f);
path2.cubicTo(39, 124.001f, 50.6f, 117.001f, 50.6f, 117.001f);
path2.cubicTo(50.6f, 117.001f, 164.601f, 85.2f, 188.201f, 117.601f);
path2.cubicTo(188.201f, 117.601f, 174.801f, 93, 39, 124.001f);
path2.close();
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path1, path2, kUnion_SkPathOp, filename);
update pathops core and tests split out skpclip (the test of 1M pictures) into its own project TBR=reed Author: caryclark@google.com Review URL: https://codereview.chromium.org/400033002
2014-07-18 12:08:14 +00:00
}
relax quadratic binary search test Extreme implicit quartic equations solve to roots that are different enough that they appear to have failed. In this case, fall back on binary searching to find an intersection. Relax the condition when this happens; don't give up just because the computed implicit root points aren't remotely the same. TBR=reed BUG=skia:2808 Author: caryclark@google.com Review URL: https://codereview.chromium.org/456383003
2014-08-12 12:49:37 +00:00
static void issue2808(skiatest::Reporter* reporter, const char* filename) {
SkPath path1, path2;
pathops coincidence and security rewrite Most changes stem from working on an examples bracketed by #if DEBUG_UNDER_DEVELOPMENT // tiger These exposed many problems with coincident curves, as well as errors throughout the code. Fixing these errors also fixed a number of fuzzer-inspired bug reports. * Line/Curve Intersections Check to see if the end of the line nearly intersects the curve. This was a FIXME in the old code. * Performance Use a central chunk allocator. Plumb the allocator into the global variable state so that it can be shared. (Note that 'SkGlobalState' is allocated on the stack and is visible to children functions but not other threads.) * Refactor Let SkOpAngle grow up from a structure to a class. Let SkCoincidentSpans grow up from a structure to a class. Rename enum Alias to AliasMatch. * Coincidence Rewrite Add more debugging to coincidence detection. Parallel debugging routines have read-only logic to report the current coincidence state so that steps through the logic can expose whether things got better or worse. More functions can error-out and cause the pathops engine to non-destructively exit. * Accuracy Remove code that adjusted point locations. Instead, offset the curve part so that sorted curves all use the same origin. Reduce the size (and influence) of magic numbers. * Testing The debug suite with verify and the full release suite ./out/Debug/pathops_unittest -v -V ./out/Release/pathops_unittest -v -V -x expose one error. That error is captured as cubics_d3. This error exists in the checked in code as well. BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 Review-Url: https://codereview.chromium.org/2128633003
2016-07-18 17:01:36 +00:00
path1.moveTo(509.20300293f, 385.601989746f);
path1.quadTo(509.20300293f, 415.68838501f, 487.928710938f, 436.96270752f);
path1.quadTo(466.654388428f, 458.236999512f, 436.567993164f, 458.236999512f);
path1.quadTo(406.4815979f, 458.236999512f, 385.207275391f, 436.96270752f);
path1.quadTo(363.932983398f, 415.68838501f, 363.932983398f, 385.601989746f);
path1.quadTo(363.932983398f, 355.515594482f, 385.207275391f, 334.241271973f);
path1.quadTo(406.4815979f, 312.96697998f, 436.567993164f, 312.96697998f);
path1.quadTo(466.654388428f, 312.96697998f, 487.928710938f, 334.241271973f);
path1.quadTo(509.20300293f, 355.515594482f, 509.20300293f, 385.601989746f);
path1.close();
path2.moveTo(449.033996582f, 290.87298584f);
path2.quadTo(449.033996582f, 301.028259277f, 441.853149414f, 308.209106445f);
path2.quadTo(434.672271729f, 315.389984131f, 424.516998291f, 315.389984131f);
path2.quadTo(414.361724854f, 315.389984131f, 407.180847168f, 308.209106445f);
path2.quadTo(400, 301.028259277f, 400, 290.87298584f);
path2.quadTo(400, 280.717712402f, 407.180847168f, 273.536865234f);
path2.quadTo(414.361724854f, 266.355987549f, 424.516998291f, 266.355987549f);
path2.quadTo(434.672271729f, 266.355987549f, 441.853149414f, 273.536865234f);
path2.quadTo(449.033996582f, 280.717712402f, 449.033996582f, 290.87298584f);
path2.close();
relax quadratic binary search test Extreme implicit quartic equations solve to roots that are different enough that they appear to have failed. In this case, fall back on binary searching to find an intersection. Relax the condition when this happens; don't give up just because the computed implicit root points aren't remotely the same. TBR=reed BUG=skia:2808 Author: caryclark@google.com Review URL: https://codereview.chromium.org/456383003
2014-08-12 12:49:37 +00:00
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path1, path2, kUnion_SkPathOp, filename);
relax quadratic binary search test Extreme implicit quartic equations solve to roots that are different enough that they appear to have failed. In this case, fall back on binary searching to find an intersection. Relax the condition when this happens; don't give up just because the computed implicit root points aren't remotely the same. TBR=reed BUG=skia:2808 Author: caryclark@google.com Review URL: https://codereview.chromium.org/456383003
2014-08-12 12:49:37 +00:00
}
These tests stress pathops by describing the union of circle-like paths that have tiny line segments embedded and double back to create near-coincident conditions. The fixes include - detect when finding the active top loops between two possible answers - preflight chasing winding to ensure answer is consistent - binary search more often when quadratic intersection fails - add more failure paths when an intersect is missed While this fixes the chrome bug, reenabling path ops in svg should be deferred until additional fixes are landed. TBR= BUG=421132 Committed: https://skia.googlesource.com/skia/+/6f726addf3178b01949bb389ef83cf14a1d7b6b2 Review URL: https://codereview.chromium.org/633393002
2014-11-13 14:58:52 +00:00
static void cubicOp115(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(3,4, 2,1, 5,3);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1,2);
pathB.cubicTo(3,5, 1,0, 4,3);
pathB.close();
SkPath path2(path);
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void testRect1(skiatest::Reporter* reporter, const char* filename) {
SkPath path, path2;
path.addRect(0, 0, 60, 60, SkPath::kCCW_Direction);
path.addRect(30, 20, 50, 50, SkPath::kCCW_Direction);
path.addRect(24, 20, 36, 30, SkPath::kCCW_Direction);
// path.addRect(32, 24, 36, 41, SkPath::kCCW_Direction);
testPathOp(reporter, path, path2, kUnion_SkPathOp, filename);
}
static void testRect2(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.addRect(0, 0, 1, 1, SkPath::kCW_Direction);
path.addRect(4, 4, 5, 5, SkPath::kCW_Direction);
pathB.setFillType(SkPath::kEvenOdd_FillType);
pathB.addRect(0, 0, 2, 2, SkPath::kCW_Direction);
pathB.addRect(0, 0, 6, 6, SkPath::kCW_Direction);
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void cubicOp116(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(4,6, 2,0, 2,0);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,2);
pathB.cubicTo(0,2, 1,0, 6,4);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
These tests stress pathops by describing the union of circle-like paths that have tiny line segments embedded and double back to create near-coincident conditions. The fixes include - detect when finding the active top loops between two possible answers - preflight chasing winding to ensure answer is consistent - binary search more often when quadratic intersection fails - add more failure paths when an intersect is missed While this fixes the chrome bug, reenabling path ops in svg should be deferred until additional fixes are landed. TBR= BUG=421132 Committed: https://skia.googlesource.com/skia/+/6f726addf3178b01949bb389ef83cf14a1d7b6b2 Review URL: https://codereview.chromium.org/633393002
2014-11-13 14:58:52 +00:00
}
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
static void cubicOp117(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(4,5, 6,0, 1,0);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,6);
pathB.cubicTo(0,1, 1,0, 5,4);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void cubicOp118(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(4,6, 5,1, 6,2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1,5);
pathB.cubicTo(2,6, 1,0, 6,4);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void loop1(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.moveTo(0,1);
path.cubicTo(1,5, -5.66666651f,3.33333349f, 8.83333302f,2.33333349f);
path.close();
pathB.moveTo(1,5);
pathB.cubicTo(-5.66666651f,3.33333349f, 8.83333302f,2.33333349f, 0,1);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
#include "SkPathOpsCubic.h"
static void loop1asQuad(skiatest::Reporter* reporter, const char* filename) {
SkDCubic c1 = {{{0,1}, {1,5}, {-5.66666651f,3.33333349f}, {8.83333302f,2.33333349f}}};
SkDCubic c2 = {{{1,5}, {-5.66666651f,3.33333349f}, {8.83333302f,2.33333349f}, {0,1}}};
double c1InflectionTs[2], c2InflectionTs[2];
SkDEBUGCODE(int c1InfTCount =) c1.findInflections(c1InflectionTs);
SkASSERT(c1InfTCount == 2);
SkDEBUGCODE(int c2InfTCount =) c2.findInflections(c2InflectionTs);
SkASSERT(c2InfTCount == 1);
SkASSERT(c1InflectionTs[0] > c1InflectionTs[1]);
SkDCubicPair c1pair = c1.chopAt(c1InflectionTs[0]);
SkDCubicPair c1apair = c1pair.first().chopAt(c1InflectionTs[1]);
SkDCubicPair c2pair = c2.chopAt(c2InflectionTs[0]);
SkDQuad q1[2] = { c1pair.first().toQuad(), c1pair.second().toQuad() };
SkDQuad q1a[2] = { c1apair.first().toQuad(), c1apair.second().toQuad() };
SkDQuad q2[2] = { c2pair.first().toQuad(), c2pair.second().toQuad() };
SkPath path, pathB;
path.moveTo(q1a[0].fPts[0].asSkPoint());
path.quadTo(q1a[0].fPts[1].asSkPoint(), q1a[0].fPts[2].asSkPoint());
path.quadTo(q1a[1].fPts[1].asSkPoint(), q1a[1].fPts[2].asSkPoint());
path.quadTo(q1[1].fPts[1].asSkPoint(), q1[1].fPts[2].asSkPoint());
path.close();
pathB.moveTo(q2[0].fPts[0].asSkPoint());
pathB.quadTo(q2[0].fPts[1].asSkPoint(), q2[0].fPts[2].asSkPoint());
pathB.quadTo(q2[1].fPts[1].asSkPoint(), q2[1].fPts[2].asSkPoint());
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void loop2(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.moveTo(0,1);
path.cubicTo(3,4, 3.f,4.f, 4.5f,1.5f);
path.close();
pathB.moveTo(3,4);
pathB.cubicTo(3.f,4.f, 4.5f,1.5f, 0,1);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void loop3(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.moveTo(0,1);
path.cubicTo(3,5, -3.66666651f,0, 10.5f,-1.66666651f);
path.close();
pathB.moveTo(3,5);
pathB.cubicTo(-3.66666651f,0, 10.5f,-1.66666651f, 0,1);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void loop4(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.moveTo(0,5);
path.cubicTo(1,5, 1,4, 0.833333313f,3);
path.close();
pathB.moveTo(1,5);
pathB.cubicTo(1,4, 0.833333313f,3, 0,5);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
#include "SkParsePath.h"
static void issue3517(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
const char str[] = "M31.35 57.75L31.35 57.75C31.9 57.7486 32.45 57.7948 33 57.7413C33.55 57.6878 34.1 57.5014 34.65 57.4291C35.2 57.3569 35.75 57.3223 36.3 57.3079C36.85 57.2935 37.4 57.3143 37.95 57.3428C38.5 57.3712 39.05 57.4112 39.6 57.4786C40.15 57.546 40.7 57.7029 41.25 57.7472C41.8 57.7916 42.35 57.7962 42.9 57.7445C43.45 57.6928 44 57.5345 44.55 57.4373C45.1 57.34 45.65 57.2115 46.2 57.1611C46.75 57.1107 47.3 57.1371 47.85 57.1349C48.4 57.1327 48.95 57.144 49.5 57.1478C50.05 57.1516 50.6 57.1553 51.15 57.1579C51.7 57.1605 52.25 57.1601 52.8 57.1634C53.35 57.1667 53.9 57.1731 54.45 57.1776C55 57.182 55.55 57.1916 56.1 57.19C56.65 57.1884 57.2 57.178 57.75 57.168C58.3 57.158 58.85 57.1355 59.4 57.1299C59.95 57.1243 60.5 57.1338 61.05 57.1345C61.6 57.1352 62.15 57.124 62.7 57.134C63.25 57.1441 63.8 57.1731 64.35 57.195C64.9 57.2169 65.45 57.2532 66 57.2655C66.55 57.2778 67.1 57.2647 67.65 57.2687C68.2 57.2728 68.75 57.267 69.3 57.2896C69.85 57.3122 70.4 57.371 70.95 57.4044C71.5 57.4377 72.05 57.4668 72.6 57.4896C73.15 57.5123 73.7 57.545 74.25 57.5408C74.8 57.5365 75.35 57.5068 75.9 57.4641C76.45 57.4213 77 57.3244 77.55 57.2842C78.1 57.244 78.65 57.2163 79.2 57.2228C79.75 57.2293 80.3 57.29 80.85 57.3232C81.4 57.3563 81.95 57.396 82.5 57.4219C83.05 57.4478 83.6 57.4637 84.15 57.4787C84.7 57.4937 85.25 57.5011 85.8 57.5121C86.35 57.523 86.9 57.5411 87.45 57.5444C88 57.5477 88.55 57.5663 89.1 57.5318C89.65 57.4972 90.2 57.3126 90.75 57.337C91.3 57.3613 91.85 57.6088 92.4 57.6776C92.95 57.7465 93.5 57.7379 94.05 57.75C94.6 57.7621 95.15 57.75 95.7 57.75L95.7 57.75L31.35 57.75Z";
SkParsePath::FromSVGString(str, &path);
const char strB[] = "M31.35 57.75L31.35 57.75C31.9 57.7514 32.45 57.7052 33 57.7587C33.55 57.8122 34.1 57.9986 34.65 58.0709C35.2 58.1431 35.75 58.1777 36.3 58.1921C36.85 58.2065 37.4 58.1857 37.95 58.1572C38.5 58.1288 39.05 58.0888 39.6 58.0214C40.15 57.954 40.7 57.7971 41.25 57.7528C41.8 57.7084 42.35 57.7038 42.9 57.7555C43.45 57.8072 44 57.9655 44.55 58.0627C45.1 58.16 45.65 58.2885 46.2 58.3389C46.75 58.3893 47.3 58.3629 47.85 58.3651C48.4 58.3673 48.95 58.356 49.5 58.3522C50.05 58.3484 50.6 58.3447 51.15 58.3421C51.7 58.3395 52.25 58.3399 52.8 58.3366C53.35 58.3333 53.9 58.3269 54.45 58.3224C55 58.318 55.55 58.3084 56.1 58.31C56.65 58.3116 57.2 58.322 57.75 58.332C58.3 58.342 58.85 58.3645 59.4 58.3701C59.95 58.3757 60.5 58.3662 61.05 58.3655C61.6 58.3648 62.15 58.376 62.7 58.366C63.25 58.3559 63.8 58.3269 64.35 58.305C64.9 58.2831 65.45 58.2468 66 58.2345C66.55 58.2222 67.1 58.2353 67.65 58.2313C68.2 58.2272 68.75 58.233 69.3 58.2104C69.85 58.1878 70.4 58.129 70.95 58.0956C71.5 58.0623 72.05 58.0332 72.6 58.0104C73.15 57.9877 73.7 57.955 74.25 57.9592C74.8 57.9635 75.35 57.9932 75.9 58.0359C76.45 58.0787 77 58.1756 77.55 58.2158C78.1 58.256 78.65 58.2837 79.2 58.2772C79.75 58.2707 80.3 58.21 80.85 58.1768C81.4 58.1437 81.95 58.104 82.5 58.0781C83.05 58.0522 83.6 58.0363 84.15 58.0213C84.7 58.0063 85.25 57.9989 85.8 57.9879C86.35 57.977 86.9 57.9589 87.45 57.9556C88 57.9523 88.55 57.9337 89.1 57.9682C89.65 58.0028 90.2 58.1874 90.75 58.163C91.3 58.1387 91.85 57.8912 92.4 57.8224C92.95 57.7535 93.5 57.7621 94.05 57.75C94.6 57.7379 95.15 57.75 95.7 57.75L95.7 57.75L31.35 57.75Z";
SkParsePath::FromSVGString(strB, &pathB);
deal more consistently with unsortable edges Improve line/curve coincident detection and resolution. This fixed the remaining simple failures. When an edge is unsortable, use the ray intersection to determine the angles' winding. Deal with degenerate segments. TBR=reed@google.com BUG=skia:3588,skia:3762 Review URL: https://codereview.chromium.org/1140813002
2015-05-13 15:23:48 +00:00
testPathOp(reporter, path, pathB, kUnion_SkPathOp, filename);
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
}
static void cubicOp119(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(3,5, 2,1, 3,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1,2);
pathB.cubicTo(1,3, 1,0, 5,3);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void cubicOp120(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(2,4, 2,1, 4,0);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1,2);
pathB.cubicTo(0,4, 1,0, 4,2);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void cubicOp121(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(3,4, 3,2, 4,3);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(2,3);
pathB.cubicTo(3,4, 1,0, 4,3);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
// FIXME : haven't debugged this failure yet
static void cubicOp122(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(3,5, 4,1, 4,0);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1,4);
pathB.cubicTo(0,4, 1,0, 5,3);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void cubicOp123(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(1,5, 2,0, 6,0);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,2);
pathB.cubicTo(0,6, 1,0, 5,1);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void loop5(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.moveTo(0,2);
path.cubicTo(1,2, 1,1.66666663f, 0.833333313f,1.33333325f);
path.close();
pathB.moveTo(1,2);
pathB.cubicTo(1,1.66666663f, 0.833333313f,1.33333325f, 0,2);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void loop6(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.moveTo(0,1);
path.cubicTo(1,3, -1.66666675f,1.66666663f, 4.16666651f,1.00000012f);
path.close();
pathB.moveTo(1,3);
pathB.cubicTo(-1.66666675f,1.66666663f, 4.16666651f,1.00000012f, 0,1);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void cubicOp124(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(1,5, 6,0, 3,0);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,6);
pathB.cubicTo(0,3, 1,0, 5,1);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void cubicOp125(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(3,6, 3,1, 6,2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1,3);
pathB.cubicTo(2,6, 1,0, 6,3);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void cubicOp126(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(0,3, 6,0, 2,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,6);
pathB.cubicTo(1,2, 1,0, 3,0);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void cubicOp127(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(1,5, 6,0, 3,0);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,6);
pathB.cubicTo(0,3, 1,0, 5,1);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void cubicOp128(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(0,3, 3,2, 5,2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(2,3);
pathB.cubicTo(2,5, 1,0, 3,0);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
Now, path ops natively intersect conics, quads, and cubics in any combination. There are still a class of cubic tests that fail and a handful of undiagnosed failures from skps and fuzz tests, but things are much better overall. Extended tests (150M+) run to completion in release in about 6 minutes; the standard test suite exceeds 100K and finishes in a few seconds on desktops. TBR=reed BUG=skia:3588 Review URL: https://codereview.chromium.org/1037953004
2015-04-20 15:31:59 +00:00
static void cubicOp129(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(5,6);
path.cubicTo(3,4, 2,0, 2,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,2);
pathB.cubicTo(1,2, 6,5, 4,3);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void cubicOp130(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(5,6);
path.cubicTo(4,6, 3,0, 2,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,3);
pathB.cubicTo(1,2, 6,5, 6,4);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
#include "SkGeometry.h"
static void complex_to_quads(const SkPoint pts[], SkPath* path) {
SkScalar loopT;
remove unused fields from SkOpSegment TBR=reed@google.com Review URL: https://codereview.chromium.org/1322413002
2015-09-04 12:00:15 +00:00
if (SkDCubic::ComplexBreak(pts, &loopT)) {
Style bikeshed - remove extraneous whitespace GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1842753002 Review URL: https://codereview.chromium.org/1842753002
2016-03-29 16:03:52 +00:00
SkPoint cubicPair[7];
Now, path ops natively intersect conics, quads, and cubics in any combination. There are still a class of cubic tests that fail and a handful of undiagnosed failures from skps and fuzz tests, but things are much better overall. Extended tests (150M+) run to completion in release in about 6 minutes; the standard test suite exceeds 100K and finishes in a few seconds on desktops. TBR=reed BUG=skia:3588 Review URL: https://codereview.chromium.org/1037953004
2015-04-20 15:31:59 +00:00
SkChopCubicAt(pts, cubicPair, loopT);
SkDCubic c1, c2;
c1.set(cubicPair);
c2.set(&cubicPair[3]);
SkDQuad q1 = c1.toQuad();
SkDQuad q2 = c2.toQuad();
path->quadTo(q1[1].asSkPoint(), q1[2].asSkPoint());
path->quadTo(q2[1].asSkPoint(), q2[2].asSkPoint());
} else {
path->cubicTo(pts[1], pts[2], pts[3]);
}
}
static void cubicOp130a(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(5,6);
SkPoint pts[] = { {5,6}, {4,6}, {3,0}, {2,1} };
complex_to_quads(pts, &path);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,3);
SkPoint pts2[] = { {0,3}, {1,2}, {6,5}, {6,4} };
complex_to_quads(pts2, &path);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void cubicOp131(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,1);
path.cubicTo(3,4, 3,0, 6,2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,3);
pathB.cubicTo(2,6, 1,0, 4,3);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void circlesOp1(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.addCircle(0, 1, 2, SkPath::kCCW_Direction);
pathB.setFillType(SkPath::kWinding_FillType);
pathB.addCircle(0, 1, 1, SkPath::kCW_Direction);
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void circlesOp2(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.addCircle(0, 1, 4, SkPath::kCCW_Direction);
pathB.setFillType(SkPath::kWinding_FillType);
pathB.addCircle(0, 4, 3, SkPath::kCW_Direction);
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void rRect1x(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(20.65f, 5.65f);
path.conicTo(20.65f, 1.13612f, 25.1404f, 0.65f, 0.888488f);
path.lineTo(25.65f, 0.65f);
path.lineTo(26.1596f, 0.67604f);
path.conicTo(30.65f, 1.13612f, 30.65f, 5.65f, 0.888488f);
path.lineTo(30.65f, 25.65f);
path.conicTo(30.65f, 20.65f, 25.65f, 20.65f, 0.707107f);
path.lineTo(20.65f, 20.65f);
path.lineTo(20.65f, 5.65f);
path.close();
path.moveTo(20.65f, 20.65f);
path.lineTo(5.65f, 20.65f);
path.conicTo(0.65f, 20.65f, 0.65f, 25.65f, 0.707107f);
path.lineTo(0.65f, 45.65f);
path.conicTo(0.65f, 50.65f, 5.65f, 50.65f, 0.707107f);
path.lineTo(25.65f, 50.65f);
path.conicTo(30.65f, 50.65f, 30.65f, 45.65f, 0.707107f);
path.lineTo(30.65f, 25.65f);
path.conicTo(30.65f, 30.65f, 25.65f, 30.65f, 0.707107f);
path.conicTo(20.65f, 30.65f, 20.65f, 25.65f, 0.707107f);
path.lineTo(20.65f, 20.65f);
path.close();
SkPath path1(path);
path.reset();
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(20.65f, 45.65f);
path.lineTo(20.65f, 25.65f);
path.conicTo(20.65f, 20.65f, 25.65f, 20.65f, 0.707107f);
path.lineTo(45.65f, 20.65f);
path.conicTo(50.65f, 20.65f, 50.65f, 25.65f, 0.707107f);
path.lineTo(50.65f, 45.65f);
path.conicTo(50.65f, 50.65f, 45.65f, 50.65f, 0.707107f);
path.lineTo(25.65f, 50.65f);
path.conicTo(20.65f, 50.65f, 20.65f, 45.65f, 0.707107f);
path.close();
SkPath path2(path);
testPathOp(reporter, path1, path2, kDifference_SkPathOp, filename);
}
static void loop7(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.moveTo(0,1);
path.cubicTo(3,4, -1,0, 8.5f,-2.5f);
path.close();
pathB.moveTo(3,4);
pathB.cubicTo(-1,0, 8.5f,-2.5f, 0,1);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void rects5(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.addRect(5, 5, 6, 6, SkPath::kCW_Direction);
path.addRect(5, 5, 6, 6, SkPath::kCW_Direction);
pathB.setFillType(SkPath::kEvenOdd_FillType);
pathB.addRect(0, 0, 6, 6, SkPath::kCW_Direction);
pathB.addRect(5, 5, 6, 6, SkPath::kCW_Direction);
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void loop8(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.moveTo(0,1);
path.cubicTo(1,4, -3.83333325f,0.166666627f, 6,-1);
path.close();
pathB.moveTo(1,4);
pathB.cubicTo(-3.83333325f,0.166666627f, 6,-1, 0,1);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void loop9(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.moveTo(0,1);
path.cubicTo(1,3, -2.5f,0, 3.33333325f,-0.666666627f);
path.close();
pathB.moveTo(1,3);
pathB.cubicTo(-2.5f,0, 3.33333325f,-0.666666627f, 0,1);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void circlesOp3(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.addCircle(0, 1, 2, SkPath::kCCW_Direction);
pathB.setFillType(SkPath::kWinding_FillType);
pathB.addCircle(3, 5, 3, SkPath::kCW_Direction);
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void loop10(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.moveTo(5,6);
path.cubicTo(1,2, 1,2, -3.66666651f,13.333334f);
path.close();
pathB.moveTo(1,2);
pathB.cubicTo(1,2, -3.66666651f,13.333334f, 5,6);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void loop11(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.moveTo(0,1);
path.cubicTo(1,3, -1.83333349f,1.33333337f, 4,-1);
path.close();
pathB.moveTo(1,3);
pathB.cubicTo(-1.83333349f,1.33333337f, 4,-1, 0,1);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void cubicOp132(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(5,6);
path.cubicTo(3,4, 3,0, 3,2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,3);
pathB.cubicTo(2,3, 6,5, 4,3);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
working on initial winding for cubics Path ops works well for all tests except for cubics. Isolate failures caused by cubics, and do a better job of computing the initial winding for cubics. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1096923003
2015-04-23 16:13:37 +00:00
static void loop12(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.moveTo(1,2);
path.cubicTo(0,6, -3.16666675f,3.66666675f, 6.33333349f,3.33333349f);
path.close();
pathB.moveTo(0,6);
pathB.cubicTo(-3.16666675f,3.66666675f, 6.33333349f,3.33333349f, 1,2);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void cubicOp133(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(5,6);
path.cubicTo(5,6, 5,0, 4,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,5);
pathB.cubicTo(1,4, 6,5, 6,5);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void cubicOp134(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(5,6);
path.cubicTo(5,6, 6,0, 3,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,6);
pathB.cubicTo(1,3, 6,5, 6,5);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void cubicOp135(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(5,6);
path.cubicTo(5,6, 6,0, 4,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,6);
pathB.cubicTo(1,4, 6,5, 6,5);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void cubicOp136(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(5,6);
path.cubicTo(5,6, 5,0, 3,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,5);
pathB.cubicTo(1,3, 6,5, 6,5);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void cubicOp136a(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(5,6);
path.quadTo(5,0, 3,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0,5);
pathB.cubicTo(1,3, 6,5, 6,5);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void cubics137(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0, 5);
path.cubicTo(3, 6, 1, 0, 3, 2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 1);
pathB.cubicTo(2, 3, 5, 0, 6, 3);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void cubics138(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0, 5);
path.cubicTo(3, 6, 1, 0, 4, 2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 1);
pathB.cubicTo(2, 4, 5, 0, 6, 3);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
// three curves intersect successfully nearby -- the angle only gets 2 of the 3 pts
static void cubicOp139(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,2);
path.cubicTo(0,4, 3,1, 5,1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1,3);
pathB.cubicTo(1,5, 2,0, 4,0);
pathB.close();
fix multiple intersection logic When three or more curves intersect at the same point, ensure that each curve records the intersections of the others. This fixes a number of cubic tests. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1105943002
2015-04-24 16:08:57 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
working on initial winding for cubics Path ops works well for all tests except for cubics. Isolate failures caused by cubics, and do a better job of computing the initial winding for cubics. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1096923003
2015-04-23 16:13:37 +00:00
}
static void cubicOp140(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,2);
path.cubicTo(1,2, 5,4, 3,2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(4,5);
pathB.cubicTo(2,3, 2,0, 2,1);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void cubicOp141(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0,2);
path.cubicTo(1,2, 6,4, 3,2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(4,6);
pathB.cubicTo(2,3, 2,0, 2,1);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void quadRect1(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.moveTo(6,15);
path.quadTo(16,0, 8,4);
path.quadTo(2,7, 12,12);
path.close();
pathB.addRect(4,11, 13,16);
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void quadRect2(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.moveTo(5,12);
path.quadTo(15,7, 9,4);
path.quadTo(1,0, 11,15);
path.close();
pathB.addRect(4,11, 13,16);
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void quadRect3(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.moveTo(12,12);
path.quadTo(2,7, 8,4);
path.quadTo(16,0, 6,15);
path.close();
pathB.addRect(4,11, 13,16);
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void quadRect4(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.moveTo(11,15);
path.quadTo(1,0, 9,4);
path.quadTo(15,7, 5,12);
path.close();
pathB.addRect(4,11, 13,16);
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void quadRect5(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.moveTo(11,13);
path.quadTo(4,4, 8,4);
path.quadTo(12,4, 5,13);
path.close();
pathB.addRect(4,11, 13,16);
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void quadRect6(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.moveTo(5,13);
path.quadTo(12,4, 8,4);
path.quadTo(4,4, 11,13);
path.close();
pathB.addRect(4,11, 13,16);
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void loops4i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0, 3);
path.cubicTo(0, 2, 0, 2, -1.66666663f, 2.16666675f);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 2);
pathB.cubicTo(0, 2, -1.66666663f, 2.16666675f, 0, 3);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void loops5i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(1, 2);
path.cubicTo(0, 2, 0, 2, 0.166666672f, 2.66666675f);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 2);
pathB.cubicTo(0, 2, 0.166666672f, 2.66666675f, 1, 2);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
fix multiple intersection logic When three or more curves intersect at the same point, ensure that each curve records the intersections of the others. This fixes a number of cubic tests. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1105943002
2015-04-24 16:08:57 +00:00
static void cubicOp142(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(5,6);
path.cubicTo(2,5, 2,1, 1,0);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1,2);
pathB.cubicTo(0,1, 6,5, 5,2);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void cubics6d(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(3, 5);
path.cubicTo(1, 5, 4, 2, 4, 0);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(2, 4);
pathB.cubicTo(0, 4, 5, 3, 5, 1);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void cubics7d(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(2, 6);
path.cubicTo(2, 4, 5, 1, 3, 1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1, 5);
pathB.cubicTo(1, 3, 6, 2, 4, 2);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void cubics8d(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(2, 5);
path.cubicTo(2, 4, 5, 1, 3, 2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1, 5);
pathB.cubicTo(2, 3, 5, 2, 4, 2);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void cubics9d(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(2, 4);
path.cubicTo(2, 6, 3, 1, 5, 1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1, 3);
pathB.cubicTo(1, 5, 4, 2, 6, 2);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void cubics10u(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(2, 4);
path.cubicTo(1, 6, 4, 1, 5, 1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1, 4);
pathB.cubicTo(1, 5, 4, 2, 6, 1);
pathB.close();
testPathOp(reporter, path, pathB, kUnion_SkPathOp, filename);
}
static void cubics11i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(2, 4);
path.cubicTo(2, 5, 3, 2, 5, 1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(2, 3);
pathB.cubicTo(1, 5, 4, 2, 5, 2);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void cubics12d(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(2, 4);
path.cubicTo(0, 4, 5, 3, 5, 1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(3, 5);
pathB.cubicTo(1, 5, 4, 2, 4, 0);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void cubics13d(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(2, 3);
path.cubicTo(1, 5, 4, 2, 5, 2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(2, 4);
pathB.cubicTo(2, 5, 3, 2, 5, 1);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void cubics14d(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(2, 3);
path.cubicTo(0, 4, 3, 1, 3, 0);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1, 3);
pathB.cubicTo(0, 3, 3, 2, 4, 0);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void cubics15d(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(1, 5);
path.cubicTo(3, 5, 4, 0, 4, 2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 4);
pathB.cubicTo(2, 4, 5, 1, 5, 3);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void cubics16i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(1, 5);
path.cubicTo(2, 5, 5, 0, 4, 2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 5);
pathB.cubicTo(2, 4, 5, 1, 5, 2);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void cubics17d(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(1, 5);
path.cubicTo(3, 4, 4, 1, 4, 2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1, 4);
pathB.cubicTo(2, 4, 5, 1, 4, 3);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void cubics18d(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(1, 5);
path.cubicTo(1, 3, 4, 0, 2, 0);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 4);
pathB.cubicTo(0, 2, 5, 1, 3, 1);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void cubics19d(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(1, 5);
path.cubicTo(2, 3, 5, 2, 4, 2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(2, 5);
pathB.cubicTo(2, 4, 5, 1, 3, 2);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void cubicOp157(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(1,5);
path.cubicTo(1,3, 6,2, 4,2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(2,6);
pathB.cubicTo(2,4, 5,1, 3,1);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
minor fixes to cubics code and overall alignment of how bounds and tops are computed for all curve types All but 17 extended tests work. A helper function is privately added to SkPath.h to permit a test to modify a given point in a path. BUG=skia:3588 Review URL: https://codereview.chromium.org/1107353004
2015-04-29 15:28:30 +00:00
static void cubics20d(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(1, 2);
path.cubicTo(0, 3, 6, 0, 3, 2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 6);
pathB.cubicTo(2, 3, 2, 1, 3, 0);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void loops20i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(1, 2);
path.cubicTo(0, 2, 0.833333313f, 2, 1, 3.66666651f);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 2);
pathB.cubicTo(0.833333313f, 2, 1, 3.66666651f, 1, 2);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void loops21i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(1, 2);
path.cubicTo(0, 2, 0.833333313f, 2, 1, 4);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 2);
pathB.cubicTo(0.833333313f, 2, 1, 4, 1, 2);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void loops22i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(1, 3);
path.cubicTo(0, 3, 0.833333313f, 3, 1, 4.66666651f);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 3);
pathB.cubicTo(0.833333313f, 3, 1, 4.66666651f, 1, 3);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void loops23i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(1, 5);
path.cubicTo(0, 1, 6.16666698f, 5.66666698f, -5.66666651f, 6.66666651f);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 1);
pathB.cubicTo(6.16666698f, 5.66666698f, -5.66666651f, 6.66666651f, 1, 5);
pathB.close();
Path ops formerly found the topmost unprocessed edge and determined its angle sort order to initialize the winding. This never worked correctly with cubics and was flaky with paths consisting mostly of vertical edges. This replacement shoots axis-aligned rays through all intersecting edges to find the outermost one either horizontally or vertically. The resulting code is smaller and twice as fast. To support this, most of the horizontal / vertical intersection code was rewritten and standardized, and old code supporting the top-directed winding was deleted. Contours were pointed to by an SkTDArray. Instead, put them in a linked list, and designate the list head with its own class to ensure that methods that take lists of contours start at the top. This change removed a large percentage of memory allocations used by path ops. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1111333002
2015-05-11 14:21:27 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
minor fixes to cubics code and overall alignment of how bounds and tops are computed for all curve types All but 17 extended tests work. A helper function is privately added to SkPath.h to permit a test to modify a given point in a path. BUG=skia:3588 Review URL: https://codereview.chromium.org/1107353004
2015-04-29 15:28:30 +00:00
}
static void loops24i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(1, 2);
path.cubicTo(0, 2, 0.833333313f, 2, 1, 3);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 2);
pathB.cubicTo(0.833333313f, 2, 1, 3, 1, 2);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void loops25i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(1, 5);
path.cubicTo(0, 5, 0.833333313f, 5, 1, 7);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 5);
pathB.cubicTo(0.833333313f, 5, 1, 7, 1, 5);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void loops26i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(1, 6);
path.cubicTo(0, 2, 6.16666698f, 6.66666698f, -5.66666651f, 7.66666651f);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 2);
pathB.cubicTo(6.16666698f, 6.66666698f, -5.66666651f, 7.66666651f, 1, 6);
pathB.close();
Path ops formerly found the topmost unprocessed edge and determined its angle sort order to initialize the winding. This never worked correctly with cubics and was flaky with paths consisting mostly of vertical edges. This replacement shoots axis-aligned rays through all intersecting edges to find the outermost one either horizontally or vertically. The resulting code is smaller and twice as fast. To support this, most of the horizontal / vertical intersection code was rewritten and standardized, and old code supporting the top-directed winding was deleted. Contours were pointed to by an SkTDArray. Instead, put them in a linked list, and designate the list head with its own class to ensure that methods that take lists of contours start at the top. This change removed a large percentage of memory allocations used by path ops. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1111333002
2015-05-11 14:21:27 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
minor fixes to cubics code and overall alignment of how bounds and tops are computed for all curve types All but 17 extended tests work. A helper function is privately added to SkPath.h to permit a test to modify a given point in a path. BUG=skia:3588 Review URL: https://codereview.chromium.org/1107353004
2015-04-29 15:28:30 +00:00
}
static void loops27i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(1, 3);
path.cubicTo(0, 3, 0.833333313f, 3, 1, 4.33333349f);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 3);
pathB.cubicTo(0.833333313f, 3, 1, 4.33333349f, 1, 3);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void loops28i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(2, 3);
path.cubicTo(1, 3, 1.83333337f, 3, 2, 4.66666651f);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1, 3);
pathB.cubicTo(1.83333337f, 3, 2, 4.66666651f, 2, 3);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void loops29i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(2, 4);
path.cubicTo(0, 4, 1.66666663f, 4, 2, 7.33333302f);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 4);
pathB.cubicTo(1.66666663f, 4, 2, 7.33333302f, 2, 4);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void loops30i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(2, 4);
path.cubicTo(0, 4, 1.66666663f, 4, 2, 8);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 4);
pathB.cubicTo(1.66666663f, 4, 2, 8, 2, 4);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void loops31i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(2, 5);
path.cubicTo(1, 5, 1.83333337f, 5, 2, 6.66666651f);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1, 5);
pathB.cubicTo(1.83333337f, 5, 2, 6.66666651f, 2, 5);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void loops32i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(2, 6);
path.cubicTo(1, 6, 1.83333337f, 6, 2, 8);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1, 6);
pathB.cubicTo(1.83333337f, 6, 2, 8, 2, 6);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void loops33i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(2, 6);
path.cubicTo(1, 2, 7.16666698f, 6.66666698f, -4.66666651f, 7.66666651f);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1, 2);
pathB.cubicTo(7.16666698f, 6.66666698f, -4.66666651f, 7.66666651f, 2, 6);
pathB.close();
Path ops formerly found the topmost unprocessed edge and determined its angle sort order to initialize the winding. This never worked correctly with cubics and was flaky with paths consisting mostly of vertical edges. This replacement shoots axis-aligned rays through all intersecting edges to find the outermost one either horizontally or vertically. The resulting code is smaller and twice as fast. To support this, most of the horizontal / vertical intersection code was rewritten and standardized, and old code supporting the top-directed winding was deleted. Contours were pointed to by an SkTDArray. Instead, put them in a linked list, and designate the list head with its own class to ensure that methods that take lists of contours start at the top. This change removed a large percentage of memory allocations used by path ops. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1111333002
2015-05-11 14:21:27 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
minor fixes to cubics code and overall alignment of how bounds and tops are computed for all curve types All but 17 extended tests work. A helper function is privately added to SkPath.h to permit a test to modify a given point in a path. BUG=skia:3588 Review URL: https://codereview.chromium.org/1107353004
2015-04-29 15:28:30 +00:00
}
static void loops33iMod(skiatest::Reporter* reporter, const char* filename) {
SkPoint pts[] = {{2, 6}, {1, 2}, {7.16666698f, 6.66666698f}, {-4.66666651f, 7.66666651f},
{1, 2}, {7.16666698f, 6.66666698f}, {-4.66666651f, 7.66666651f}, {2, 6}};
bool up = false;
float offset = 0.0380172729f;
float step = 7.62939453e-006f;
bool lastResult = true;
// for (int i = 0; i < 30; ++i) {
SkString name(filename);
// name.appendS32(i);
// if (i > 0) {
// SkDebugf("\n\n<div id=\"%s\">\n", name.c_str());
// }
pts[5].fY = 6.66666698f + offset;
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(pts[0]);
path.cubicTo(pts[1], pts[2], pts[3]);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(pts[4]);
pathB.cubicTo(pts[5], pts[6], pts[7]);
pathB.close();
Path ops formerly found the topmost unprocessed edge and determined its angle sort order to initialize the winding. This never worked correctly with cubics and was flaky with paths consisting mostly of vertical edges. This replacement shoots axis-aligned rays through all intersecting edges to find the outermost one either horizontally or vertically. The resulting code is smaller and twice as fast. To support this, most of the horizontal / vertical intersection code was rewritten and standardized, and old code supporting the top-directed winding was deleted. Contours were pointed to by an SkTDArray. Instead, put them in a linked list, and designate the list head with its own class to ensure that methods that take lists of contours start at the top. This change removed a large percentage of memory allocations used by path ops. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1111333002
2015-05-11 14:21:27 +00:00
bool result = testPathOp(reporter, path, pathB, kIntersect_SkPathOp, name.c_str());
minor fixes to cubics code and overall alignment of how bounds and tops are computed for all curve types All but 17 extended tests work. A helper function is privately added to SkPath.h to permit a test to modify a given point in a path. BUG=skia:3588 Review URL: https://codereview.chromium.org/1107353004
2015-04-29 15:28:30 +00:00
if (lastResult != result) {
up = !up;
}
step /= 2;
offset += up ? step : -step;
lastResult = result;
// }
}
static void loops33iAsQuads(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(2, 6);
path.cubicTo(1, 2, 7.16666698f, 6.66666698f, -4.66666651f, 7.66666651f);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1, 2);
pathB.cubicTo(7.16666698f, 6.66666698f, -4.66666651f, 7.66666651f, 2, 6);
pathB.close();
SkPath qPath, qPathB;
CubicPathToQuads(path, &qPath);
CubicPathToQuads(pathB, &qPathB);
testPathOp(reporter, qPath, qPathB, kIntersect_SkPathOp, filename);
}
static void loops34i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(3, 4);
path.cubicTo(0, 4, 2.5f, 4, 3, 9);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 4);
pathB.cubicTo(2.5f, 4, 3, 9, 3, 4);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void loops35i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(3, 4);
path.cubicTo(0, 4, 2.5f, 4, 3, 10);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 4);
pathB.cubicTo(2.5f, 4, 3, 10, 3, 4);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void loops36i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(3, 4);
path.cubicTo(1, 4, 2.66666675f, 4, 3, 8);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1, 4);
pathB.cubicTo(2.66666675f, 4, 3, 8, 3, 4);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void loops37i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(2, 4);
path.cubicTo(1, 4, 1.83333337f, 4, 2, 5.33333349f);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1, 4);
pathB.cubicTo(1.83333337f, 4, 2, 5.33333349f, 2, 4);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void loops38i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(3, 4);
path.cubicTo(2, 4, 2.83333325f, 4, 3, 6);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(2, 4);
pathB.cubicTo(2.83333325f, 4, 3, 6, 3, 4);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void loops39i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(3, 5);
path.cubicTo(0, 5, 2.5f, 5, 3, 10);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 5);
pathB.cubicTo(2.5f, 5, 3, 10, 3, 5);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void loops40i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(3, 5);
path.cubicTo(0, 5, 2.5f, 5, 3, 11);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 5);
pathB.cubicTo(2.5f, 5, 3, 11, 3, 5);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void loops40iAsQuads(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(3, 5);
path.cubicTo(0, 5, 2.5f, 5, 3, 11);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 5);
pathB.cubicTo(2.5f, 5, 3, 11, 3, 5);
pathB.close();
SkPath qPath, qPathB;
CubicPathToQuads(path, &qPath);
CubicPathToQuads(pathB, &qPathB);
testPathOp(reporter, qPath, qPathB, kIntersect_SkPathOp, filename);
}
static void loops44i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(1, 5);
path.cubicTo(0, 1, 7.33333302f, 5.33333349f, -7, 7);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 1);
pathB.cubicTo(7.33333302f, 5.33333349f, -7, 7, 1, 5);
pathB.close();
Path ops formerly found the topmost unprocessed edge and determined its angle sort order to initialize the winding. This never worked correctly with cubics and was flaky with paths consisting mostly of vertical edges. This replacement shoots axis-aligned rays through all intersecting edges to find the outermost one either horizontally or vertically. The resulting code is smaller and twice as fast. To support this, most of the horizontal / vertical intersection code was rewritten and standardized, and old code supporting the top-directed winding was deleted. Contours were pointed to by an SkTDArray. Instead, put them in a linked list, and designate the list head with its own class to ensure that methods that take lists of contours start at the top. This change removed a large percentage of memory allocations used by path ops. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1111333002
2015-05-11 14:21:27 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
minor fixes to cubics code and overall alignment of how bounds and tops are computed for all curve types All but 17 extended tests work. A helper function is privately added to SkPath.h to permit a test to modify a given point in a path. BUG=skia:3588 Review URL: https://codereview.chromium.org/1107353004
2015-04-29 15:28:30 +00:00
}
static void loops45i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(1, 6);
path.cubicTo(0, 2, 7.33333302f, 6.33333302f, -7, 8);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 2);
pathB.cubicTo(7.33333302f, 6.33333302f, -7, 8, 1, 6);
pathB.close();
Path ops formerly found the topmost unprocessed edge and determined its angle sort order to initialize the winding. This never worked correctly with cubics and was flaky with paths consisting mostly of vertical edges. This replacement shoots axis-aligned rays through all intersecting edges to find the outermost one either horizontally or vertically. The resulting code is smaller and twice as fast. To support this, most of the horizontal / vertical intersection code was rewritten and standardized, and old code supporting the top-directed winding was deleted. Contours were pointed to by an SkTDArray. Instead, put them in a linked list, and designate the list head with its own class to ensure that methods that take lists of contours start at the top. This change removed a large percentage of memory allocations used by path ops. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1111333002
2015-05-11 14:21:27 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
minor fixes to cubics code and overall alignment of how bounds and tops are computed for all curve types All but 17 extended tests work. A helper function is privately added to SkPath.h to permit a test to modify a given point in a path. BUG=skia:3588 Review URL: https://codereview.chromium.org/1107353004
2015-04-29 15:28:30 +00:00
}
static void loops46i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(2, 6);
path.cubicTo(1, 2, 8.33333302f, 6.33333302f, -6, 8);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1, 2);
pathB.cubicTo(8.33333302f, 6.33333302f, -6, 8, 2, 6);
pathB.close();
Path ops formerly found the topmost unprocessed edge and determined its angle sort order to initialize the winding. This never worked correctly with cubics and was flaky with paths consisting mostly of vertical edges. This replacement shoots axis-aligned rays through all intersecting edges to find the outermost one either horizontally or vertically. The resulting code is smaller and twice as fast. To support this, most of the horizontal / vertical intersection code was rewritten and standardized, and old code supporting the top-directed winding was deleted. Contours were pointed to by an SkTDArray. Instead, put them in a linked list, and designate the list head with its own class to ensure that methods that take lists of contours start at the top. This change removed a large percentage of memory allocations used by path ops. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1111333002
2015-05-11 14:21:27 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
minor fixes to cubics code and overall alignment of how bounds and tops are computed for all curve types All but 17 extended tests work. A helper function is privately added to SkPath.h to permit a test to modify a given point in a path. BUG=skia:3588 Review URL: https://codereview.chromium.org/1107353004
2015-04-29 15:28:30 +00:00
}
/*
pathops coincidence and security rewrite Most changes stem from working on an examples bracketed by #if DEBUG_UNDER_DEVELOPMENT // tiger These exposed many problems with coincident curves, as well as errors throughout the code. Fixing these errors also fixed a number of fuzzer-inspired bug reports. * Line/Curve Intersections Check to see if the end of the line nearly intersects the curve. This was a FIXME in the old code. * Performance Use a central chunk allocator. Plumb the allocator into the global variable state so that it can be shared. (Note that 'SkGlobalState' is allocated on the stack and is visible to children functions but not other threads.) * Refactor Let SkOpAngle grow up from a structure to a class. Let SkCoincidentSpans grow up from a structure to a class. Rename enum Alias to AliasMatch. * Coincidence Rewrite Add more debugging to coincidence detection. Parallel debugging routines have read-only logic to report the current coincidence state so that steps through the logic can expose whether things got better or worse. More functions can error-out and cause the pathops engine to non-destructively exit. * Accuracy Remove code that adjusted point locations. Instead, offset the curve part so that sorted curves all use the same origin. Reduce the size (and influence) of magic numbers. * Testing The debug suite with verify and the full release suite ./out/Debug/pathops_unittest -v -V ./out/Release/pathops_unittest -v -V -x expose one error. That error is captured as cubics_d3. This error exists in the checked in code as well. BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 Review-Url: https://codereview.chromium.org/2128633003
2016-07-18 17:01:36 +00:00
FAILED: d:\cygwin\puregit\tests\pathopsextendedtest.cpp:346 0 */
minor fixes to cubics code and overall alignment of how bounds and tops are computed for all curve types All but 17 extended tests work. A helper function is privately added to SkPath.h to permit a test to modify a given point in a path. BUG=skia:3588 Review URL: https://codereview.chromium.org/1107353004
2015-04-29 15:28:30 +00:00
static void loops47i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(2, 4);
path.cubicTo(0, 1, 6, 5.83333302f, -4, 8);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 1);
pathB.cubicTo(6, 5.83333302f, -4, 8, 2, 4);
pathB.close();
Path ops formerly found the topmost unprocessed edge and determined its angle sort order to initialize the winding. This never worked correctly with cubics and was flaky with paths consisting mostly of vertical edges. This replacement shoots axis-aligned rays through all intersecting edges to find the outermost one either horizontally or vertically. The resulting code is smaller and twice as fast. To support this, most of the horizontal / vertical intersection code was rewritten and standardized, and old code supporting the top-directed winding was deleted. Contours were pointed to by an SkTDArray. Instead, put them in a linked list, and designate the list head with its own class to ensure that methods that take lists of contours start at the top. This change removed a large percentage of memory allocations used by path ops. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1111333002
2015-05-11 14:21:27 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
minor fixes to cubics code and overall alignment of how bounds and tops are computed for all curve types All but 17 extended tests work. A helper function is privately added to SkPath.h to permit a test to modify a given point in a path. BUG=skia:3588 Review URL: https://codereview.chromium.org/1107353004
2015-04-29 15:28:30 +00:00
}
static void loops48i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(2, 6);
path.cubicTo(0, 1, 9.33333302f, 6.83333302f, -8.33333302f, 9.16666603f);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 1);
pathB.cubicTo(9.33333302f, 6.83333302f, -8.33333302f, 9.16666603f, 2, 6);
pathB.close();
Path ops formerly found the topmost unprocessed edge and determined its angle sort order to initialize the winding. This never worked correctly with cubics and was flaky with paths consisting mostly of vertical edges. This replacement shoots axis-aligned rays through all intersecting edges to find the outermost one either horizontally or vertically. The resulting code is smaller and twice as fast. To support this, most of the horizontal / vertical intersection code was rewritten and standardized, and old code supporting the top-directed winding was deleted. Contours were pointed to by an SkTDArray. Instead, put them in a linked list, and designate the list head with its own class to ensure that methods that take lists of contours start at the top. This change removed a large percentage of memory allocations used by path ops. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1111333002
2015-05-11 14:21:27 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
minor fixes to cubics code and overall alignment of how bounds and tops are computed for all curve types All but 17 extended tests work. A helper function is privately added to SkPath.h to permit a test to modify a given point in a path. BUG=skia:3588 Review URL: https://codereview.chromium.org/1107353004
2015-04-29 15:28:30 +00:00
}
static void loops49i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0, 2);
path.cubicTo(1, 4, -0.166666687f, 2.66666675f, 1.66666675f, 2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1, 4);
pathB.cubicTo(-0.166666687f, 2.66666675f, 1.66666675f, 2, 0, 2);
pathB.close();
Path ops formerly found the topmost unprocessed edge and determined its angle sort order to initialize the winding. This never worked correctly with cubics and was flaky with paths consisting mostly of vertical edges. This replacement shoots axis-aligned rays through all intersecting edges to find the outermost one either horizontally or vertically. The resulting code is smaller and twice as fast. To support this, most of the horizontal / vertical intersection code was rewritten and standardized, and old code supporting the top-directed winding was deleted. Contours were pointed to by an SkTDArray. Instead, put them in a linked list, and designate the list head with its own class to ensure that methods that take lists of contours start at the top. This change removed a large percentage of memory allocations used by path ops. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1111333002
2015-05-11 14:21:27 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
minor fixes to cubics code and overall alignment of how bounds and tops are computed for all curve types All but 17 extended tests work. A helper function is privately added to SkPath.h to permit a test to modify a given point in a path. BUG=skia:3588 Review URL: https://codereview.chromium.org/1107353004
2015-04-29 15:28:30 +00:00
}
static void loops50i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0, 3);
path.cubicTo(1, 5, -0.166666687f, 3.66666675f, 1.66666675f, 3);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1, 5);
pathB.cubicTo(-0.166666687f, 3.66666675f, 1.66666675f, 3, 0, 3);
pathB.close();
Path ops formerly found the topmost unprocessed edge and determined its angle sort order to initialize the winding. This never worked correctly with cubics and was flaky with paths consisting mostly of vertical edges. This replacement shoots axis-aligned rays through all intersecting edges to find the outermost one either horizontally or vertically. The resulting code is smaller and twice as fast. To support this, most of the horizontal / vertical intersection code was rewritten and standardized, and old code supporting the top-directed winding was deleted. Contours were pointed to by an SkTDArray. Instead, put them in a linked list, and designate the list head with its own class to ensure that methods that take lists of contours start at the top. This change removed a large percentage of memory allocations used by path ops. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1111333002
2015-05-11 14:21:27 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
minor fixes to cubics code and overall alignment of how bounds and tops are computed for all curve types All but 17 extended tests work. A helper function is privately added to SkPath.h to permit a test to modify a given point in a path. BUG=skia:3588 Review URL: https://codereview.chromium.org/1107353004
2015-04-29 15:28:30 +00:00
}
static void loops51i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(1, 2);
path.cubicTo(2, 4, 0.833333313f, 2.66666675f, 2.66666675f, 2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(2, 4);
pathB.cubicTo(0.833333313f, 2.66666675f, 2.66666675f, 2, 1, 2);
pathB.close();
Path ops formerly found the topmost unprocessed edge and determined its angle sort order to initialize the winding. This never worked correctly with cubics and was flaky with paths consisting mostly of vertical edges. This replacement shoots axis-aligned rays through all intersecting edges to find the outermost one either horizontally or vertically. The resulting code is smaller and twice as fast. To support this, most of the horizontal / vertical intersection code was rewritten and standardized, and old code supporting the top-directed winding was deleted. Contours were pointed to by an SkTDArray. Instead, put them in a linked list, and designate the list head with its own class to ensure that methods that take lists of contours start at the top. This change removed a large percentage of memory allocations used by path ops. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1111333002
2015-05-11 14:21:27 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
minor fixes to cubics code and overall alignment of how bounds and tops are computed for all curve types All but 17 extended tests work. A helper function is privately added to SkPath.h to permit a test to modify a given point in a path. BUG=skia:3588 Review URL: https://codereview.chromium.org/1107353004
2015-04-29 15:28:30 +00:00
}
static void loops52i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(1, 3);
path.cubicTo(2, 5, 0.833333313f, 3.66666675f, 2.66666675f, 3);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(2, 5);
pathB.cubicTo(0.833333313f, 3.66666675f, 2.66666675f, 3, 1, 3);
pathB.close();
Path ops formerly found the topmost unprocessed edge and determined its angle sort order to initialize the winding. This never worked correctly with cubics and was flaky with paths consisting mostly of vertical edges. This replacement shoots axis-aligned rays through all intersecting edges to find the outermost one either horizontally or vertically. The resulting code is smaller and twice as fast. To support this, most of the horizontal / vertical intersection code was rewritten and standardized, and old code supporting the top-directed winding was deleted. Contours were pointed to by an SkTDArray. Instead, put them in a linked list, and designate the list head with its own class to ensure that methods that take lists of contours start at the top. This change removed a large percentage of memory allocations used by path ops. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1111333002
2015-05-11 14:21:27 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
minor fixes to cubics code and overall alignment of how bounds and tops are computed for all curve types All but 17 extended tests work. A helper function is privately added to SkPath.h to permit a test to modify a given point in a path. BUG=skia:3588 Review URL: https://codereview.chromium.org/1107353004
2015-04-29 15:28:30 +00:00
}
static void loops53i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(2, 3);
path.cubicTo(3, 5, 1.83333325f, 3.66666675f, 3.66666651f, 3);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(3, 5);
pathB.cubicTo(1.83333325f, 3.66666675f, 3.66666651f, 3, 2, 3);
pathB.close();
Path ops formerly found the topmost unprocessed edge and determined its angle sort order to initialize the winding. This never worked correctly with cubics and was flaky with paths consisting mostly of vertical edges. This replacement shoots axis-aligned rays through all intersecting edges to find the outermost one either horizontally or vertically. The resulting code is smaller and twice as fast. To support this, most of the horizontal / vertical intersection code was rewritten and standardized, and old code supporting the top-directed winding was deleted. Contours were pointed to by an SkTDArray. Instead, put them in a linked list, and designate the list head with its own class to ensure that methods that take lists of contours start at the top. This change removed a large percentage of memory allocations used by path ops. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1111333002
2015-05-11 14:21:27 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
minor fixes to cubics code and overall alignment of how bounds and tops are computed for all curve types All but 17 extended tests work. A helper function is privately added to SkPath.h to permit a test to modify a given point in a path. BUG=skia:3588 Review URL: https://codereview.chromium.org/1107353004
2015-04-29 15:28:30 +00:00
}
static void loops54i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0, 2);
path.cubicTo(1, 4, 0, 3, 1.66666675f, 2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1, 4);
pathB.cubicTo(0, 3, 1.66666675f, 2, 0, 2);
pathB.close();
Path ops formerly found the topmost unprocessed edge and determined its angle sort order to initialize the winding. This never worked correctly with cubics and was flaky with paths consisting mostly of vertical edges. This replacement shoots axis-aligned rays through all intersecting edges to find the outermost one either horizontally or vertically. The resulting code is smaller and twice as fast. To support this, most of the horizontal / vertical intersection code was rewritten and standardized, and old code supporting the top-directed winding was deleted. Contours were pointed to by an SkTDArray. Instead, put them in a linked list, and designate the list head with its own class to ensure that methods that take lists of contours start at the top. This change removed a large percentage of memory allocations used by path ops. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1111333002
2015-05-11 14:21:27 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
minor fixes to cubics code and overall alignment of how bounds and tops are computed for all curve types All but 17 extended tests work. A helper function is privately added to SkPath.h to permit a test to modify a given point in a path. BUG=skia:3588 Review URL: https://codereview.chromium.org/1107353004
2015-04-29 15:28:30 +00:00
}
static void loops55i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0, 3);
path.cubicTo(1, 5, 0, 4, 1.66666675f, 3);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1, 5);
pathB.cubicTo(0, 4, 1.66666675f, 3, 0, 3);
pathB.close();
Path ops formerly found the topmost unprocessed edge and determined its angle sort order to initialize the winding. This never worked correctly with cubics and was flaky with paths consisting mostly of vertical edges. This replacement shoots axis-aligned rays through all intersecting edges to find the outermost one either horizontally or vertically. The resulting code is smaller and twice as fast. To support this, most of the horizontal / vertical intersection code was rewritten and standardized, and old code supporting the top-directed winding was deleted. Contours were pointed to by an SkTDArray. Instead, put them in a linked list, and designate the list head with its own class to ensure that methods that take lists of contours start at the top. This change removed a large percentage of memory allocations used by path ops. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1111333002
2015-05-11 14:21:27 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
minor fixes to cubics code and overall alignment of how bounds and tops are computed for all curve types All but 17 extended tests work. A helper function is privately added to SkPath.h to permit a test to modify a given point in a path. BUG=skia:3588 Review URL: https://codereview.chromium.org/1107353004
2015-04-29 15:28:30 +00:00
}
static void loops56i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(1, 2);
path.cubicTo(2, 4, 0.99999994f, 3, 2.66666675f, 2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(2, 4);
pathB.cubicTo(0.99999994f, 3, 2.66666675f, 2, 1, 2);
pathB.close();
Path ops formerly found the topmost unprocessed edge and determined its angle sort order to initialize the winding. This never worked correctly with cubics and was flaky with paths consisting mostly of vertical edges. This replacement shoots axis-aligned rays through all intersecting edges to find the outermost one either horizontally or vertically. The resulting code is smaller and twice as fast. To support this, most of the horizontal / vertical intersection code was rewritten and standardized, and old code supporting the top-directed winding was deleted. Contours were pointed to by an SkTDArray. Instead, put them in a linked list, and designate the list head with its own class to ensure that methods that take lists of contours start at the top. This change removed a large percentage of memory allocations used by path ops. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1111333002
2015-05-11 14:21:27 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
minor fixes to cubics code and overall alignment of how bounds and tops are computed for all curve types All but 17 extended tests work. A helper function is privately added to SkPath.h to permit a test to modify a given point in a path. BUG=skia:3588 Review URL: https://codereview.chromium.org/1107353004
2015-04-29 15:28:30 +00:00
}
static void loops57i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(1, 3);
path.cubicTo(2, 5, 0.99999994f, 4, 2.66666675f, 3);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(2, 5);
pathB.cubicTo(0.99999994f, 4, 2.66666675f, 3, 1, 3);
pathB.close();
Path ops formerly found the topmost unprocessed edge and determined its angle sort order to initialize the winding. This never worked correctly with cubics and was flaky with paths consisting mostly of vertical edges. This replacement shoots axis-aligned rays through all intersecting edges to find the outermost one either horizontally or vertically. The resulting code is smaller and twice as fast. To support this, most of the horizontal / vertical intersection code was rewritten and standardized, and old code supporting the top-directed winding was deleted. Contours were pointed to by an SkTDArray. Instead, put them in a linked list, and designate the list head with its own class to ensure that methods that take lists of contours start at the top. This change removed a large percentage of memory allocations used by path ops. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1111333002
2015-05-11 14:21:27 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
minor fixes to cubics code and overall alignment of how bounds and tops are computed for all curve types All but 17 extended tests work. A helper function is privately added to SkPath.h to permit a test to modify a given point in a path. BUG=skia:3588 Review URL: https://codereview.chromium.org/1107353004
2015-04-29 15:28:30 +00:00
}
static void loops58i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(2, 3);
path.cubicTo(3, 5, 2, 4, 3.66666651f, 3);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(3, 5);
pathB.cubicTo(2, 4, 3.66666651f, 3, 2, 3);
pathB.close();
Path ops formerly found the topmost unprocessed edge and determined its angle sort order to initialize the winding. This never worked correctly with cubics and was flaky with paths consisting mostly of vertical edges. This replacement shoots axis-aligned rays through all intersecting edges to find the outermost one either horizontally or vertically. The resulting code is smaller and twice as fast. To support this, most of the horizontal / vertical intersection code was rewritten and standardized, and old code supporting the top-directed winding was deleted. Contours were pointed to by an SkTDArray. Instead, put them in a linked list, and designate the list head with its own class to ensure that methods that take lists of contours start at the top. This change removed a large percentage of memory allocations used by path ops. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1111333002
2015-05-11 14:21:27 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void loops58iAsQuads(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(2, 3);
path.cubicTo(3, 5, 2, 4, 3.66666651f, 3);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(3, 5);
pathB.cubicTo(2, 4, 3.66666651f, 3, 2, 3);
pathB.close();
SkPath qPath, qPathB;
CubicPathToQuads(path, &qPath);
CubicPathToQuads(pathB, &qPathB);
// SkPoint from = {2.61714339f,1.90228665f};
// SkPoint to = {2.617045833359139f,1.9013528935803314f};
// path_edit(from, to, &qPathB);
testPathOp(reporter, qPath, qPathB, kIntersect_SkPathOp, filename);
minor fixes to cubics code and overall alignment of how bounds and tops are computed for all curve types All but 17 extended tests work. A helper function is privately added to SkPath.h to permit a test to modify a given point in a path. BUG=skia:3588 Review URL: https://codereview.chromium.org/1107353004
2015-04-29 15:28:30 +00:00
}
static void loops59i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0, 6);
path.cubicTo(1, 2, 7.33333302f, 1.66666663f, -7.5f, 2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1, 2);
pathB.cubicTo(7.33333302f, 1.66666663f, -7.5f, 2, 0, 6);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void loops59iasQuads(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0, 6);
path.cubicTo(1, 2, 7.33333302f, 1.66666663f, -7.5f, 2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1, 2);
pathB.cubicTo(7.33333302f, 1.66666663f, -7.5f, 2, 0, 6);
pathB.close();
SkPath qPath, qPathB;
CubicPathToQuads(path, &qPath);
CubicPathToQuads(pathB, &qPathB);
SkPoint from = {2.61714339f,1.90228665f};
SkPoint to = {2.617045833359139f,1.9013528935803314f};
path_edit(from, to, &qPathB);
testPathOp(reporter, qPath, qPathB, kIntersect_SkPathOp, filename);
}
static void cubics41d(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0, 1);
path.cubicTo(1, 4, 3, 0, 3, 1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 3);
pathB.cubicTo(1, 3, 1, 0, 4, 1);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
Path ops formerly found the topmost unprocessed edge and determined its angle sort order to initialize the winding. This never worked correctly with cubics and was flaky with paths consisting mostly of vertical edges. This replacement shoots axis-aligned rays through all intersecting edges to find the outermost one either horizontally or vertically. The resulting code is smaller and twice as fast. To support this, most of the horizontal / vertical intersection code was rewritten and standardized, and old code supporting the top-directed winding was deleted. Contours were pointed to by an SkTDArray. Instead, put them in a linked list, and designate the list head with its own class to ensure that methods that take lists of contours start at the top. This change removed a large percentage of memory allocations used by path ops. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1111333002
2015-05-11 14:21:27 +00:00
void loops61i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0, 1);
path.cubicTo(1, 5, -6.33333302f, 0.666666627f, 8, -1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1, 5);
pathB.cubicTo(-6.33333302f, 0.666666627f, 8, -1, 0, 1);
pathB.close();
deal more consistently with unsortable edges Improve line/curve coincident detection and resolution. This fixed the remaining simple failures. When an edge is unsortable, use the ray intersection to determine the angles' winding. Deal with degenerate segments. TBR=reed@google.com BUG=skia:3588,skia:3762 Review URL: https://codereview.chromium.org/1140813002
2015-05-13 15:23:48 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Path ops formerly found the topmost unprocessed edge and determined its angle sort order to initialize the winding. This never worked correctly with cubics and was flaky with paths consisting mostly of vertical edges. This replacement shoots axis-aligned rays through all intersecting edges to find the outermost one either horizontally or vertically. The resulting code is smaller and twice as fast. To support this, most of the horizontal / vertical intersection code was rewritten and standardized, and old code supporting the top-directed winding was deleted. Contours were pointed to by an SkTDArray. Instead, put them in a linked list, and designate the list head with its own class to ensure that methods that take lists of contours start at the top. This change removed a large percentage of memory allocations used by path ops. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1111333002
2015-05-11 14:21:27 +00:00
}
static void loops62i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0, 2);
path.cubicTo(1, 6, -6.33333302f, 1.66666663f, 8, 0);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1, 6);
pathB.cubicTo(-6.33333302f, 1.66666663f, 8, 0, 0, 2);
pathB.close();
deal more consistently with unsortable edges Improve line/curve coincident detection and resolution. This fixed the remaining simple failures. When an edge is unsortable, use the ray intersection to determine the angles' winding. Deal with degenerate segments. TBR=reed@google.com BUG=skia:3588,skia:3762 Review URL: https://codereview.chromium.org/1140813002
2015-05-13 15:23:48 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Path ops formerly found the topmost unprocessed edge and determined its angle sort order to initialize the winding. This never worked correctly with cubics and was flaky with paths consisting mostly of vertical edges. This replacement shoots axis-aligned rays through all intersecting edges to find the outermost one either horizontally or vertically. The resulting code is smaller and twice as fast. To support this, most of the horizontal / vertical intersection code was rewritten and standardized, and old code supporting the top-directed winding was deleted. Contours were pointed to by an SkTDArray. Instead, put them in a linked list, and designate the list head with its own class to ensure that methods that take lists of contours start at the top. This change removed a large percentage of memory allocations used by path ops. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1111333002
2015-05-11 14:21:27 +00:00
}
static void loops63i(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0, 1);
path.cubicTo(2, 4, -4, -0.833333254f, 6, -3);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(2, 4);
pathB.cubicTo(-4, -0.833333254f, 6, -3, 0, 1);
pathB.close();
deal more consistently with unsortable edges Improve line/curve coincident detection and resolution. This fixed the remaining simple failures. When an edge is unsortable, use the ray intersection to determine the angles' winding. Deal with degenerate segments. TBR=reed@google.com BUG=skia:3588,skia:3762 Review URL: https://codereview.chromium.org/1140813002
2015-05-13 15:23:48 +00:00
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
Path ops formerly found the topmost unprocessed edge and determined its angle sort order to initialize the winding. This never worked correctly with cubics and was flaky with paths consisting mostly of vertical edges. This replacement shoots axis-aligned rays through all intersecting edges to find the outermost one either horizontally or vertically. The resulting code is smaller and twice as fast. To support this, most of the horizontal / vertical intersection code was rewritten and standardized, and old code supporting the top-directed winding was deleted. Contours were pointed to by an SkTDArray. Instead, put them in a linked list, and designate the list head with its own class to ensure that methods that take lists of contours start at the top. This change removed a large percentage of memory allocations used by path ops. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1111333002
2015-05-11 14:21:27 +00:00
}
static void cubics44d(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(3, 4);
path.cubicTo(2, 5, 3, 1, 6, 2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1, 3);
pathB.cubicTo(2, 6, 4, 3, 5, 2);
pathB.close();
deal more consistently with unsortable edges Improve line/curve coincident detection and resolution. This fixed the remaining simple failures. When an edge is unsortable, use the ray intersection to determine the angles' winding. Deal with degenerate segments. TBR=reed@google.com BUG=skia:3588,skia:3762 Review URL: https://codereview.chromium.org/1140813002
2015-05-13 15:23:48 +00:00
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
Path ops formerly found the topmost unprocessed edge and determined its angle sort order to initialize the winding. This never worked correctly with cubics and was flaky with paths consisting mostly of vertical edges. This replacement shoots axis-aligned rays through all intersecting edges to find the outermost one either horizontally or vertically. The resulting code is smaller and twice as fast. To support this, most of the horizontal / vertical intersection code was rewritten and standardized, and old code supporting the top-directed winding was deleted. Contours were pointed to by an SkTDArray. Instead, put them in a linked list, and designate the list head with its own class to ensure that methods that take lists of contours start at the top. This change removed a large percentage of memory allocations used by path ops. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1111333002
2015-05-11 14:21:27 +00:00
}
static void cubics45u(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(1, 3);
path.cubicTo(2, 6, 4, 3, 5, 2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(3, 4);
pathB.cubicTo(2, 5, 3, 1, 6, 2);
pathB.close();
deal more consistently with unsortable edges Improve line/curve coincident detection and resolution. This fixed the remaining simple failures. When an edge is unsortable, use the ray intersection to determine the angles' winding. Deal with degenerate segments. TBR=reed@google.com BUG=skia:3588,skia:3762 Review URL: https://codereview.chromium.org/1140813002
2015-05-13 15:23:48 +00:00
testPathOp(reporter, path, pathB, kUnion_SkPathOp, filename);
Path ops formerly found the topmost unprocessed edge and determined its angle sort order to initialize the winding. This never worked correctly with cubics and was flaky with paths consisting mostly of vertical edges. This replacement shoots axis-aligned rays through all intersecting edges to find the outermost one either horizontally or vertically. The resulting code is smaller and twice as fast. To support this, most of the horizontal / vertical intersection code was rewritten and standardized, and old code supporting the top-directed winding was deleted. Contours were pointed to by an SkTDArray. Instead, put them in a linked list, and designate the list head with its own class to ensure that methods that take lists of contours start at the top. This change removed a large percentage of memory allocations used by path ops. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1111333002
2015-05-11 14:21:27 +00:00
}
fix pathops fuzz failures If a curve has the identical start and control points, the initial or final tangent can't be trivally determined. The perpendicular to the tangent is used to measure coincidence. Add logic for cubics, quadratics, and conics, to use the secondary control points or the end points if the initial control point alone can't determine the tangent. Add debugging (currently untriggered by exhaustive testing) to detect zero-length tangents which are not at the curve endpoints. Increase the number of temporary intersecions gathered from 10 to 12 but reduce the max passed in by cubic intersection from 27 to 12. Also, add checks if the max passed exceeds the storage allocated. When cleaning up parallel lines, choose the intersection which is on the end of both segments over the intersection which is on the end of a single segment. TBR=reed@google.com BUG=425140,516266 Review URL: https://codereview.chromium.org/1288863004
2015-08-18 14:12:43 +00:00
static void fuzz38(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.moveTo(100.34f, 303.312f);
path.lineTo(-1e+08, 303.312f);
path.lineTo(102, 310.156f);
path.lineTo(100.34f, 310.156f);
path.lineTo(100.34f, 303.312f);
path.close();
pathops coincidence and security rewrite Most changes stem from working on an examples bracketed by #if DEBUG_UNDER_DEVELOPMENT // tiger These exposed many problems with coincident curves, as well as errors throughout the code. Fixing these errors also fixed a number of fuzzer-inspired bug reports. * Line/Curve Intersections Check to see if the end of the line nearly intersects the curve. This was a FIXME in the old code. * Performance Use a central chunk allocator. Plumb the allocator into the global variable state so that it can be shared. (Note that 'SkGlobalState' is allocated on the stack and is visible to children functions but not other threads.) * Refactor Let SkOpAngle grow up from a structure to a class. Let SkCoincidentSpans grow up from a structure to a class. Rename enum Alias to AliasMatch. * Coincidence Rewrite Add more debugging to coincidence detection. Parallel debugging routines have read-only logic to report the current coincidence state so that steps through the logic can expose whether things got better or worse. More functions can error-out and cause the pathops engine to non-destructively exit. * Accuracy Remove code that adjusted point locations. Instead, offset the curve part so that sorted curves all use the same origin. Reduce the size (and influence) of magic numbers. * Testing The debug suite with verify and the full release suite ./out/Debug/pathops_unittest -v -V ./out/Release/pathops_unittest -v -V -x expose one error. That error is captured as cubics_d3. This error exists in the checked in code as well. BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 Review-Url: https://codereview.chromium.org/2128633003
2016-07-18 17:01:36 +00:00
testPathOpCheck(reporter, path, pathB, kUnion_SkPathOp, filename, true);
fix pathops fuzz failures If a curve has the identical start and control points, the initial or final tangent can't be trivally determined. The perpendicular to the tangent is used to measure coincidence. Add logic for cubics, quadratics, and conics, to use the secondary control points or the end points if the initial control point alone can't determine the tangent. Add debugging (currently untriggered by exhaustive testing) to detect zero-length tangents which are not at the curve endpoints. Increase the number of temporary intersecions gathered from 10 to 12 but reduce the max passed in by cubic intersection from 27 to 12. Also, add checks if the max passed exceeds the storage allocated. When cleaning up parallel lines, choose the intersection which is on the end of both segments over the intersection which is on the end of a single segment. TBR=reed@google.com BUG=425140,516266 Review URL: https://codereview.chromium.org/1288863004
2015-08-18 14:12:43 +00:00
}
link test to bug crbug.com/526025 includes a minimized SVG test case. Translating that test case into native code (fuzzTNG) did not reproduce the bug. That test case should have not been included with skia issue 1323813003, and is deleted here. Running the minimal test case in a modified version of chrome isolated the bug. The modified version generated the test fuzz763_3 with the edit #define DEBUGGING_PATHOPS_FROM_HOST 1 in src/pathops/SkPathopsOp.cpp line 188. Rename fuzz763_3 to issue_526025 to associate the test with the bug. Note that the bug contains the body of the CL in comment $5. R=reed@google.com Review URL: https://codereview.chromium.org/1315503005
2015-09-01 13:22:36 +00:00
static void crbug_526025(skiatest::Reporter* reporter, const char* filename) {
remove duplicate linked list adjustment The list of intersection points on a curve segment may have entries that can be safely removed when nearby points have nearly the same t value and point value. When a path includes very large curves as well as small ones, as is the case with this fuzzer, additional points may lie between the similar points that do not meet the nearby criteria. After merging the nearby point with its doppelganger, SkOpSegment::moveNearby() unnecessarily set the doppelganger's next pointer to the one following the nearby point. While this usually has no effect, since the merge already updated the linked list, the explicit call removes the additional outlier points from the segment. TBR=reed@google.com BUG=526025 Review URL: https://codereview.chromium.org/1323813003
2015-08-31 16:46:20 +00:00
SkPath path;
path.setFillType((SkPath::FillType) 1);
path.moveTo(SkBits2Float(0x43b40000), SkBits2Float(0xcf000000)); // 360, -2.14748e+09f
path.cubicTo(SkBits2Float(0x4e0d628f), SkBits2Float(0xceffffff), SkBits2Float(0x4e800003), SkBits2Float(0xcec6b143), SkBits2Float(0x4e800002), SkBits2Float(0xce7ffffc)); // 5.93012e+08f, -2.14748e+09f, 1.07374e+09f, -1.66675e+09f, 1.07374e+09f, -1.07374e+09f
path.cubicTo(SkBits2Float(0x4e800002), SkBits2Float(0xcde53aee), SkBits2Float(0x4e0d6292), SkBits2Float(0xc307820e), SkBits2Float(0x44627d00), SkBits2Float(0x437ffff2)); // 1.07374e+09f, -4.80731e+08f, 5.93012e+08f, -135.508f, 905.953f, 256
path.lineTo(SkBits2Float(0x444bf3bc), SkBits2Float(0x4460537e)); // 815.808f, 897.305f
path.lineTo(SkBits2Float(0x43553abd), SkBits2Float(0x440f3cbd)); // 213.229f, 572.949f
path.lineTo(SkBits2Float(0x42000000), SkBits2Float(0x41800000)); // 32, 16
path.lineTo(SkBits2Float(0x42c80000), SkBits2Float(0x44000000)); // 100, 512
path.lineTo(SkBits2Float(0x43553abd), SkBits2Float(0x440f3cbd)); // 213.229f, 572.949f
path.lineTo(SkBits2Float(0x43b40000), SkBits2Float(0x44800000)); // 360, 1024
path.lineTo(SkBits2Float(0x43b40000), SkBits2Float(0x45816000)); // 360, 4140
SkPath path1(path);
path.reset();
path.setFillType((SkPath::FillType) 0);
path.moveTo(SkBits2Float(0x42fe0000), SkBits2Float(0x43a08000)); // 127, 321
path.lineTo(SkBits2Float(0x45d5c000), SkBits2Float(0x43870000)); // 6840, 270
path.lineTo(SkBits2Float(0xd0a00000), SkBits2Float(0x4cbebc20)); // -2.14748e+10f, 1e+08
path.lineTo(SkBits2Float(0x451f7000), SkBits2Float(0x42800000)); // 2551, 64
path.lineTo(SkBits2Float(0x42fe0000), SkBits2Float(0x43a08000)); // 127, 321
path.close();
SkPath path2(path);
testPathOp(reporter, path1, path2, (SkPathOp) 2, filename);
}
fix security bug This fix is a tradeoff. It changes intersection to treat a case where one coincident run is intersected at one point and the other edge is not as continuing to be a span. The old code tried to treat this as a single point. The old code is probably right, but this change alone made the data structures inconsistent. Later, extending the coincident runs would fail by incorrectly discarding the single point intersection. As a result, this fixes the security test and one other, but makes a different test fail. Isolating the failure uncovered a reduced case that fails with and without the change, so there are more serious problems here. Those problems are addressed in a separate CL. Many of the test edits below remove ill-thought out debugging messaging that fire off global state, which isn't usable in a multi-threaded test environment. In the end, with this fix, all existing tests (modulo one new failure and one new non-failure) pass in debug and in the extended release test suites. TBR=reed@google.com BUG=614248 GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2018513003 Review-Url: https://codereview.chromium.org/2018513003
2016-05-26 16:01:47 +00:00
static void fuzzX_392(skiatest::Reporter* reporter, const char* filename) {
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(SkBits2Float(0x41e80000), SkBits2Float(0x43bde212)); // 29, 379.766f
path.lineTo(SkBits2Float(0x41e80000), SkBits2Float(0x43bdc7ef)); // 29, 379.562f
path.conicTo(SkBits2Float(0x42a5861e), SkBits2Float(0x43c61f86), SkBits2Float(0x430b0610), SkBits2Float(0x43c61f86), SkBits2Float(0x3f7d23f3)); // 82.7619f, 396.246f, 139.024f, 396.246f, 0.98883f
path.conicTo(SkBits2Float(0x42a58e20), SkBits2Float(0x43c61f86), SkBits2Float(0x41e80000), SkBits2Float(0x43bde212), SkBits2Float(0x3f7d2cf5)); // 82.7776f, 396.246f, 29, 379.766f, 0.988967f
path.close();
SkPath path1(path);
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(SkBits2Float(0xc36c7bd8), SkBits2Float(0xc3a31d72)); // -236.484f, -326.23f
path.lineTo(SkBits2Float(0xc367a4ae), SkBits2Float(0xc3a31d72)); // -231.643f, -326.23f
path.lineTo(SkBits2Float(0x430b0610), SkBits2Float(0x43c61f86)); // 139.024f, 396.246f
path.lineTo(SkBits2Float(0xc36c7bd8), SkBits2Float(0x43c61f86)); // -236.484f, 396.246f
SkPath path2(path);
testPathOp(reporter, path1, path2, kIntersect_SkPathOp, filename);
}
pathops coincidence and security rewrite Most changes stem from working on an examples bracketed by #if DEBUG_UNDER_DEVELOPMENT // tiger These exposed many problems with coincident curves, as well as errors throughout the code. Fixing these errors also fixed a number of fuzzer-inspired bug reports. * Line/Curve Intersections Check to see if the end of the line nearly intersects the curve. This was a FIXME in the old code. * Performance Use a central chunk allocator. Plumb the allocator into the global variable state so that it can be shared. (Note that 'SkGlobalState' is allocated on the stack and is visible to children functions but not other threads.) * Refactor Let SkOpAngle grow up from a structure to a class. Let SkCoincidentSpans grow up from a structure to a class. Rename enum Alias to AliasMatch. * Coincidence Rewrite Add more debugging to coincidence detection. Parallel debugging routines have read-only logic to report the current coincidence state so that steps through the logic can expose whether things got better or worse. More functions can error-out and cause the pathops engine to non-destructively exit. * Accuracy Remove code that adjusted point locations. Instead, offset the curve part so that sorted curves all use the same origin. Reduce the size (and influence) of magic numbers. * Testing The debug suite with verify and the full release suite ./out/Debug/pathops_unittest -v -V ./out/Release/pathops_unittest -v -V -x expose one error. That error is captured as cubics_d3. This error exists in the checked in code as well. BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 Review-Url: https://codereview.chromium.org/2128633003
2016-07-18 17:01:36 +00:00
static void dean2(skiatest::Reporter* reporter, const char* filename) {
SkPath path;
path.setFillType((SkPath::FillType) 0);
path.moveTo(SkBits2Float(0x3f2b74b3), SkBits2Float(0x4154a02b)); // 0.669749f, 13.2891f
path.cubicTo(SkBits2Float(0x3f2b74b3), SkBits2Float(0x4154a02b), SkBits2Float(0x41531912), SkBits2Float(0x3f130322), SkBits2Float(0x4154a02b), SkBits2Float(0x3f2b74b3)); // 0.669749f, 13.2891f, 13.1936f, 0.574267f, 13.2891f, 0.669749f
path.cubicTo(SkBits2Float(0x414a835a), SkBits2Float(0x3ec07ba6), SkBits2Float(0x413fcc0d), SkBits2Float(0x3e193319), SkBits2Float(0x4134a02b), SkBits2Float(0x00000000)); // 12.6571f, 0.375943f, 11.9873f, 0.149609f, 11.2891f, 0
path.lineTo(SkBits2Float(0x3f2b74b3), SkBits2Float(0x4154a02b)); // 0.669749f, 13.2891f
path.close();
SkPath path1(path);
path.reset();
path.setFillType((SkPath::FillType) 0);
path.moveTo(SkBits2Float(0x3f2b74b3), SkBits2Float(0x4154a02b)); // 0.669749f, 13.2891f
path.cubicTo(SkBits2Float(0x3f2b74b3), SkBits2Float(0x4154a02b), SkBits2Float(0x41531912), SkBits2Float(0x3f130322), SkBits2Float(0x4154a02b), SkBits2Float(0x3f2b74b3)); // 0.669749f, 13.2891f, 13.1936f, 0.574267f, 13.2891f, 0.669749f
path.lineTo(SkBits2Float(0x417ab74b), SkBits2Float(0x4154a02b)); // 15.6697f, 13.2891f
path.lineTo(SkBits2Float(0x3f2b74b3), SkBits2Float(0x4154a02b)); // 0.669749f, 13.2891f
path.close();
SkPath path2(path);
testPathOp(reporter, path1, path2, kIntersect_SkPathOp, filename);
}
static void cubics_d(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0, 1);
path.cubicTo(3, 5, 1, 0, 3, 0);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 1);
pathB.cubicTo(0, 3, 1, 0, 5, 3);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void cubics_d2(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(0, 1);
path.cubicTo(2, 5, 2, 0, 2, 1);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 2);
pathB.cubicTo(1, 2, 1, 0, 5, 2);
pathB.close();
testPathOp(reporter, path, pathB, kDifference_SkPathOp, filename);
}
static void loops_i1(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(2, 3);
path.cubicTo(0, 4, -0.333333343f, 4.66666651f, 3, 5.83333349f);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 4);
pathB.cubicTo(-0.333333343f, 4.66666651f, 3, 5.83333349f, 2, 3);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void loops_i2(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(2, 4);
path.cubicTo(0, 5, -0.333333343f, 5.66666651f, 3, 6.83333302f);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 5);
pathB.cubicTo(-0.333333343f, 5.66666651f, 3, 6.83333302f, 2, 4);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void loops_i3(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(2, 5);
path.cubicTo(0, 6, -0.333333343f, 6.66666651f, 3, 7.83333302f);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 6);
pathB.cubicTo(-0.333333343f, 6.66666651f, 3, 7.83333302f, 2, 5);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void loops_i4(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(3, 4);
path.cubicTo(1, 5, 0.666666627f, 5.66666651f, 4, 6.83333302f);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1, 5);
pathB.cubicTo(0.666666627f, 5.66666651f, 4, 6.83333302f, 3, 4);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void loops_i5(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(3, 5);
path.cubicTo(1, 6, 0.666666627f, 6.66666651f, 4, 7.83333302f);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1, 6);
pathB.cubicTo(0.666666627f, 6.66666651f, 4, 7.83333302f, 3, 5);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void loops_i6(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(4, 5);
path.cubicTo(2, 6, 1.66666663f, 6.66666651f, 5, 7.83333302f);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(2, 6);
pathB.cubicTo(1.66666663f, 6.66666651f, 5, 7.83333302f, 4, 5);
pathB.close();
testPathOp(reporter, path, pathB, kIntersect_SkPathOp, filename);
}
static void cubics_d3(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(3, 4);
path.cubicTo(0, 6, 6, 1, 4, 2);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(1, 6);
pathB.cubicTo(2, 4, 4, 3, 6, 0);
pathB.close();
// DEBUG_UNDER_DEVELOPMENT cubics_d3 disable expectation check for now
fix fuzzer bug Fix another fuzzer bug. Some PathOps asserts only make sense if the incoming data is well-behaved. Well-behaved tests set debugging state to trigger these additional asserts. Formalize this by creating macros similar to SkASSERT that check to see if the assert should be skipped. TBR=reed@google.com BUG=629962 GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2169863002 Review-Url: https://codereview.chromium.org/2169863002
2016-07-21 12:48:43 +00:00
// fails because segment 3 is marked as unorderable ?
pathops coincidence and security rewrite Most changes stem from working on an examples bracketed by #if DEBUG_UNDER_DEVELOPMENT // tiger These exposed many problems with coincident curves, as well as errors throughout the code. Fixing these errors also fixed a number of fuzzer-inspired bug reports. * Line/Curve Intersections Check to see if the end of the line nearly intersects the curve. This was a FIXME in the old code. * Performance Use a central chunk allocator. Plumb the allocator into the global variable state so that it can be shared. (Note that 'SkGlobalState' is allocated on the stack and is visible to children functions but not other threads.) * Refactor Let SkOpAngle grow up from a structure to a class. Let SkCoincidentSpans grow up from a structure to a class. Rename enum Alias to AliasMatch. * Coincidence Rewrite Add more debugging to coincidence detection. Parallel debugging routines have read-only logic to report the current coincidence state so that steps through the logic can expose whether things got better or worse. More functions can error-out and cause the pathops engine to non-destructively exit. * Accuracy Remove code that adjusted point locations. Instead, offset the curve part so that sorted curves all use the same origin. Reduce the size (and influence) of magic numbers. * Testing The debug suite with verify and the full release suite ./out/Debug/pathops_unittest -v -V ./out/Release/pathops_unittest -v -V -x expose one error. That error is captured as cubics_d3. This error exists in the checked in code as well. BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 Review-Url: https://codereview.chromium.org/2128633003
2016-07-18 17:01:36 +00:00
testPathOpCheck(reporter, path, pathB, kDifference_SkPathOp, filename, !FLAGS_runFail);
}
static void cubics_o(skiatest::Reporter* reporter, const char* filename) {
SkPath path, pathB;
path.setFillType(SkPath::kWinding_FillType);
path.moveTo(1, 4);
path.cubicTo(2, 6, 5, 0, 5, 3);
path.close();
pathB.setFillType(SkPath::kWinding_FillType);
pathB.moveTo(0, 5);
pathB.cubicTo(3, 5, 4, 1, 6, 2);
pathB.close();
testPathOp(reporter, path, pathB, kXOR_SkPathOp, filename);
}
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
static void (*skipTest)(skiatest::Reporter* , const char* filename) = 0;
pathops coincidence and security rewrite Most changes stem from working on an examples bracketed by #if DEBUG_UNDER_DEVELOPMENT // tiger These exposed many problems with coincident curves, as well as errors throughout the code. Fixing these errors also fixed a number of fuzzer-inspired bug reports. * Line/Curve Intersections Check to see if the end of the line nearly intersects the curve. This was a FIXME in the old code. * Performance Use a central chunk allocator. Plumb the allocator into the global variable state so that it can be shared. (Note that 'SkGlobalState' is allocated on the stack and is visible to children functions but not other threads.) * Refactor Let SkOpAngle grow up from a structure to a class. Let SkCoincidentSpans grow up from a structure to a class. Rename enum Alias to AliasMatch. * Coincidence Rewrite Add more debugging to coincidence detection. Parallel debugging routines have read-only logic to report the current coincidence state so that steps through the logic can expose whether things got better or worse. More functions can error-out and cause the pathops engine to non-destructively exit. * Accuracy Remove code that adjusted point locations. Instead, offset the curve part so that sorted curves all use the same origin. Reduce the size (and influence) of magic numbers. * Testing The debug suite with verify and the full release suite ./out/Debug/pathops_unittest -v -V ./out/Release/pathops_unittest -v -V -x expose one error. That error is captured as cubics_d3. This error exists in the checked in code as well. BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 Review-Url: https://codereview.chromium.org/2128633003
2016-07-18 17:01:36 +00:00
static void (*firstTest)(skiatest::Reporter* , const char* filename) = cubics_d3;
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void (*stopTest)(skiatest::Reporter* , const char* filename) = 0;
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
Path ops formerly found the topmost unprocessed edge and determined its angle sort order to initialize the winding. This never worked correctly with cubics and was flaky with paths consisting mostly of vertical edges. This replacement shoots axis-aligned rays through all intersecting edges to find the outermost one either horizontally or vertically. The resulting code is smaller and twice as fast. To support this, most of the horizontal / vertical intersection code was rewritten and standardized, and old code supporting the top-directed winding was deleted. Contours were pointed to by an SkTDArray. Instead, put them in a linked list, and designate the list head with its own class to ensure that methods that take lists of contours start at the top. This change removed a large percentage of memory allocations used by path ops. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1111333002
2015-05-11 14:21:27 +00:00
#define TEST(name) { name, #name }
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
static struct TestDesc tests[] = {
pathops coincidence and security rewrite Most changes stem from working on an examples bracketed by #if DEBUG_UNDER_DEVELOPMENT // tiger These exposed many problems with coincident curves, as well as errors throughout the code. Fixing these errors also fixed a number of fuzzer-inspired bug reports. * Line/Curve Intersections Check to see if the end of the line nearly intersects the curve. This was a FIXME in the old code. * Performance Use a central chunk allocator. Plumb the allocator into the global variable state so that it can be shared. (Note that 'SkGlobalState' is allocated on the stack and is visible to children functions but not other threads.) * Refactor Let SkOpAngle grow up from a structure to a class. Let SkCoincidentSpans grow up from a structure to a class. Rename enum Alias to AliasMatch. * Coincidence Rewrite Add more debugging to coincidence detection. Parallel debugging routines have read-only logic to report the current coincidence state so that steps through the logic can expose whether things got better or worse. More functions can error-out and cause the pathops engine to non-destructively exit. * Accuracy Remove code that adjusted point locations. Instead, offset the curve part so that sorted curves all use the same origin. Reduce the size (and influence) of magic numbers. * Testing The debug suite with verify and the full release suite ./out/Debug/pathops_unittest -v -V ./out/Release/pathops_unittest -v -V -x expose one error. That error is captured as cubics_d3. This error exists in the checked in code as well. BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 Review-Url: https://codereview.chromium.org/2128633003
2016-07-18 17:01:36 +00:00
TEST(loops_i1),
TEST(loops_i2),
TEST(loops_i3),
TEST(loops_i4),
TEST(loops_i5),
TEST(loops_i6),
TEST(cubics_d3),
TEST(cubics_o),
TEST(cubics_d2),
TEST(cubics_d),
TEST(dean2),
fix security bug This fix is a tradeoff. It changes intersection to treat a case where one coincident run is intersected at one point and the other edge is not as continuing to be a span. The old code tried to treat this as a single point. The old code is probably right, but this change alone made the data structures inconsistent. Later, extending the coincident runs would fail by incorrectly discarding the single point intersection. As a result, this fixes the security test and one other, but makes a different test fail. Isolating the failure uncovered a reduced case that fails with and without the change, so there are more serious problems here. Those problems are addressed in a separate CL. Many of the test edits below remove ill-thought out debugging messaging that fire off global state, which isn't usable in a multi-threaded test environment. In the end, with this fix, all existing tests (modulo one new failure and one new non-failure) pass in debug and in the extended release test suites. TBR=reed@google.com BUG=614248 GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2018513003 Review-Url: https://codereview.chromium.org/2018513003
2016-05-26 16:01:47 +00:00
TEST(fuzzX_392),
link test to bug crbug.com/526025 includes a minimized SVG test case. Translating that test case into native code (fuzzTNG) did not reproduce the bug. That test case should have not been included with skia issue 1323813003, and is deleted here. Running the minimal test case in a modified version of chrome isolated the bug. The modified version generated the test fuzz763_3 with the edit #define DEBUGGING_PATHOPS_FROM_HOST 1 in src/pathops/SkPathopsOp.cpp line 188. Rename fuzz763_3 to issue_526025 to associate the test with the bug. Note that the bug contains the body of the CL in comment $5. R=reed@google.com Review URL: https://codereview.chromium.org/1315503005
2015-09-01 13:22:36 +00:00
TEST(crbug_526025),
fix pathops fuzz failures If a curve has the identical start and control points, the initial or final tangent can't be trivally determined. The perpendicular to the tangent is used to measure coincidence. Add logic for cubics, quadratics, and conics, to use the secondary control points or the end points if the initial control point alone can't determine the tangent. Add debugging (currently untriggered by exhaustive testing) to detect zero-length tangents which are not at the curve endpoints. Increase the number of temporary intersecions gathered from 10 to 12 but reduce the max passed in by cubic intersection from 27 to 12. Also, add checks if the max passed exceeds the storage allocated. When cleaning up parallel lines, choose the intersection which is on the end of both segments over the intersection which is on the end of a single segment. TBR=reed@google.com BUG=425140,516266 Review URL: https://codereview.chromium.org/1288863004
2015-08-18 14:12:43 +00:00
TEST(fuzz38),
Path ops formerly found the topmost unprocessed edge and determined its angle sort order to initialize the winding. This never worked correctly with cubics and was flaky with paths consisting mostly of vertical edges. This replacement shoots axis-aligned rays through all intersecting edges to find the outermost one either horizontally or vertically. The resulting code is smaller and twice as fast. To support this, most of the horizontal / vertical intersection code was rewritten and standardized, and old code supporting the top-directed winding was deleted. Contours were pointed to by an SkTDArray. Instead, put them in a linked list, and designate the list head with its own class to ensure that methods that take lists of contours start at the top. This change removed a large percentage of memory allocations used by path ops. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1111333002
2015-05-11 14:21:27 +00:00
TEST(cubics44d),
TEST(cubics45u),
TEST(loops61i),
TEST(loops62i),
TEST(loops63i),
TEST(loops58iAsQuads),
minor fixes to cubics code and overall alignment of how bounds and tops are computed for all curve types All but 17 extended tests work. A helper function is privately added to SkPath.h to permit a test to modify a given point in a path. BUG=skia:3588 Review URL: https://codereview.chromium.org/1107353004
2015-04-29 15:28:30 +00:00
TEST(cubics41d),
TEST(loops59iasQuads),
TEST(loops59i),
TEST(loops44i),
TEST(loops45i),
TEST(loops46i),
TEST(loops47i),
TEST(loops48i),
TEST(loops49i),
TEST(loops50i),
TEST(loops51i),
TEST(loops52i),
TEST(loops53i),
TEST(loops54i),
TEST(loops55i),
TEST(loops56i),
TEST(loops57i),
TEST(loops58i),
TEST(loops33iMod),
TEST(loops33iAsQuads),
TEST(loops33i),
TEST(loops40i),
TEST(loops40iAsQuads),
TEST(loops39i),
TEST(loops38i),
TEST(loops37i),
TEST(loops36i),
TEST(loops35i),
TEST(loops34i),
TEST(loops32i),
TEST(loops31i),
TEST(loops30i),
TEST(loops29i),
TEST(loops28i),
TEST(loops27i),
TEST(loops26i),
TEST(loops25i),
TEST(loops24i),
TEST(loops23i),
TEST(loops22i),
TEST(loops21i),
TEST(loops20i),
TEST(cubics20d),
fix multiple intersection logic When three or more curves intersect at the same point, ensure that each curve records the intersections of the others. This fixes a number of cubic tests. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1105943002
2015-04-24 16:08:57 +00:00
TEST(cubics6d),
TEST(cubics7d),
TEST(cubics8d),
TEST(cubics9d),
TEST(cubics10u),
TEST(cubics11i),
TEST(cubics12d),
TEST(cubics13d),
TEST(cubics14d),
TEST(cubics15d),
TEST(cubics16i),
TEST(cubics17d),
TEST(cubics18d),
TEST(cubics19d),
TEST(cubicOp157),
TEST(cubicOp142),
working on initial winding for cubics Path ops works well for all tests except for cubics. Isolate failures caused by cubics, and do a better job of computing the initial winding for cubics. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1096923003
2015-04-23 16:13:37 +00:00
TEST(loops4i),
TEST(quadRect1),
TEST(quadRect2),
TEST(quadRect3),
TEST(quadRect4),
TEST(quadRect5),
TEST(quadRect6),
TEST(cubicOp141),
TEST(cubicOp58d),
TEST(loops5i),
TEST(cubicOp140),
TEST(cubicOp139),
TEST(cubics138),
TEST(cubics137),
TEST(cubicOp136a),
TEST(cubicOp136),
TEST(cubicOp135),
TEST(cubicOp134),
TEST(cubicOp133),
TEST(loop12),
Now, path ops natively intersect conics, quads, and cubics in any combination. There are still a class of cubic tests that fail and a handful of undiagnosed failures from skps and fuzz tests, but things are much better overall. Extended tests (150M+) run to completion in release in about 6 minutes; the standard test suite exceeds 100K and finishes in a few seconds on desktops. TBR=reed BUG=skia:3588 Review URL: https://codereview.chromium.org/1037953004
2015-04-20 15:31:59 +00:00
TEST(cubicOp132),
TEST(loop11),
TEST(loop10),
TEST(circlesOp3),
TEST(loop9),
TEST(loop8),
TEST(rects5),
TEST(loop7),
TEST(cubicOp130a),
TEST(rRect1x),
TEST(circlesOp2),
TEST(circlesOp1),
TEST(cubicOp131),
TEST(cubicOp130),
TEST(cubicOp129),
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
TEST(cubicOp128),
TEST(cubicOp127),
TEST(cubicOp126),
TEST(cubicOp125),
TEST(cubicOp124),
TEST(loop6),
TEST(loop5),
TEST(cubicOp123),
TEST(cubicOp122),
TEST(cubicOp121),
TEST(cubicOp120),
TEST(cubicOp119),
TEST(loop4),
TEST(loop3),
TEST(loop2),
TEST(loop1asQuad),
TEST(loop1),
TEST(issue3517),
TEST(cubicOp118),
TEST(cubicOp117),
TEST(cubicOp116),
TEST(testRect2),
TEST(testRect1),
These tests stress pathops by describing the union of circle-like paths that have tiny line segments embedded and double back to create near-coincident conditions. The fixes include - detect when finding the active top loops between two possible answers - preflight chasing winding to ensure answer is consistent - binary search more often when quadratic intersection fails - add more failure paths when an intersect is missed While this fixes the chrome bug, reenabling path ops in svg should be deferred until additional fixes are landed. TBR= BUG=421132 Committed: https://skia.googlesource.com/skia/+/6f726addf3178b01949bb389ef83cf14a1d7b6b2 Review URL: https://codereview.chromium.org/633393002
2014-11-13 14:58:52 +00:00
TEST(cubicOp115),
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
TEST(issue2753),
working on initial winding for cubics Path ops works well for all tests except for cubics. Isolate failures caused by cubics, and do a better job of computing the initial winding for cubics. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1096923003
2015-04-23 16:13:37 +00:00
TEST(cubicOp114),
relax quadratic binary search test Extreme implicit quartic equations solve to roots that are different enough that they appear to have failed. In this case, fall back on binary searching to find an intersection. Relax the condition when this happens; don't give up just because the computed implicit root points aren't remotely the same. TBR=reed BUG=skia:2808 Author: caryclark@google.com Review URL: https://codereview.chromium.org/456383003
2014-08-12 12:49:37 +00:00
TEST(issue2808),
Fix last pathops skp bug This fixes the last bug discovered by iterating through the 800K skp corpus representing the top 1M websites. For every clip on the stack, the paths are replaced with the pathop intersection. The resulting draw is compared with the original draw for pixel errors. At least two prominent bugs remain. In one, the winding value is confused by a cubic with an inflection. In the other, a quad/cubic pair, nearly coincident, fails to find an intersection. These minor changes include ignoring very tiny self-intersections of cubics, and processing degenerate edges that don't connect to anything else. R=reed@android.com TBR=reed Author: caryclark@google.com Review URL: https://codereview.chromium.org/340103002
2014-06-18 14:24:19 +00:00
TEST(cubicOp114asQuad),
Enabling the canvas bit to turn the clip stack into a flat replace exposed around 100 failures when testing the 800K skp set generated from the top 1M web sites. This fixes all but one of those failures. Major changes include: - Replace angle indices with angle pointers. This was motivated by the need to add angles later but not renumber existing angles. - Aggressive segment chase. When the winding is known on a segment, more aggressively passing that winding to adjacent segments allows fragmented data sets to succeed. - Line segments with ends nearly the same are treated as coincident first. - Transfer partial coincidence by observing that if segment A is partially coincident to B and C then B and C may be partially coincident. TBR=reed Author: caryclark@google.com Review URL: https://codereview.chromium.org/272153002
2014-06-17 12:15:38 +00:00
TEST(rects4),
TEST(rects3),
TEST(rects2),
TEST(rects1),
TEST(issue2540),
When solving the cubic line intersection directly fails, use binary search as a fallback. The cubic line intersection math empirically works 99.99% of the time (fails 3100 out of 1B random tests) but when it fails, an intersection may be missed altogether. The binary search is may not find a solution if the cubic line failed to find any solutions at all, but so far that case hasn't arisen. BUG=skia:2504 TBR=reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/266063003 git-svn-id: http://skia.googlecode.com/svn/trunk@14614 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-05-07 15:31:40 +00:00
TEST(issue2504),
fix cubic/line intersection; add skp tests BUG=skia:2488 TBR=reed Author: caryclark@google.com Review URL: https://codereview.chromium.org/252243003 git-svn-id: http://skia.googlecode.com/svn/trunk@14458 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-30 13:37:48 +00:00
TEST(kari1),
fix minor skp-found bugs remove globals from pathops_unittest BUG=skia:2460 TBR=mtklein Author: caryclark@google.com Review URL: https://codereview.chromium.org/239563004 git-svn-id: http://skia.googlecode.com/svn/trunk@14378 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-25 12:59:11 +00:00
TEST(quadOp10i),
TEST(cubicOp113),
path ops work in progress make more skps work remove edit files BUG= Review URL: https://codereview.chromium.org/23542056 git-svn-id: http://skia.googlecode.com/svn/trunk@11570 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-02 14:49:34 +00:00
TEST(skpcarrot_is24),
TEST(issue1417),
fix minor skp-found bugs remove globals from pathops_unittest BUG=skia:2460 TBR=mtklein Author: caryclark@google.com Review URL: https://codereview.chromium.org/239563004 git-svn-id: http://skia.googlecode.com/svn/trunk@14378 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-25 12:59:11 +00:00
TEST(cubicOp112),
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
TEST(skpadspert_net23),
TEST(skpadspert_de11),
TEST(findFirst1),
TEST(xOp2i),
TEST(xOp3i),
TEST(xOp1u),
TEST(xOp1i),
TEST(cubicOp111),
TEST(cubicOp110),
TEST(cubicOp109),
TEST(cubicOp108),
TEST(cubicOp107),
TEST(cubicOp106),
TEST(cubicOp105),
TEST(cubicOp104),
TEST(cubicOp103),
TEST(cubicOp102),
TEST(cubicOp101),
TEST(cubicOp100),
TEST(cubicOp99),
TEST(issue1435),
pathops work in progress BUG= Review URL: https://codereview.chromium.org/52653002 git-svn-id: http://skia.googlecode.com/svn/trunk@12089 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-11-01 17:36:03 +00:00
TEST(cubicOp98x),
TEST(cubicOp97x),
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
TEST(skpcarpetplanet_ru22),
path ops work in progress make more skps work remove edit files BUG= Review URL: https://codereview.chromium.org/23542056 git-svn-id: http://skia.googlecode.com/svn/trunk@11570 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-10-02 14:49:34 +00:00
TEST(cubicOp96d),
TEST(cubicOp95u),
TEST(skpadbox_lt15),
TEST(skpagentxsites_com55),
TEST(skpadventistmission_org572),
TEST(skpadoption_org196),
TEST(skpbambootheme_com12),
TEST(skpbakosoft_com10),
TEST(skpakmmos_ru100),
TEST(skpbangalorenest_com4),
TEST(skpbingoentertainment_net189),
TEST(skpbestred_ru37),
TEST(skpbenzoteh_ru152),
TEST(skpcamcorder_kz21),
TEST(skpcaffelavazzait_com_ua21),
TEST(skpcarrefour_ro62),
TEST(skpcavablar_net563),
TEST(skpinsomnia_gr72),
TEST(skpadbox_lt8),
TEST(skpact_com43),
TEST(skpacesoftech_com47),
TEST(skpabcspark_ca103),
TEST(cubicOp94u),
TEST(cubicOp93d),
TEST(cubicOp92i),
TEST(skpadithya_putr4_blogspot_com551),
TEST(skpadindex_de4),
TEST(skpaiaigames_com870),
TEST(skpaaalgarve_org53),
TEST(skpkkiste_to716),
path ops work in progress path ops work in progress BUG= Review URL: https://codereview.chromium.org/21359002 git-svn-id: http://skia.googlecode.com/svn/trunk@11291 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-09-16 15:55:01 +00:00
TEST(cubicOp91u),
TEST(cubicOp90u),
TEST(cubicOp89u),
TEST(cubicOp88u),
TEST(cubicOp87u),
TEST(cubicOp86i),
TEST(loopEdge2),
TEST(loopEdge1),
TEST(rectOp3x),
TEST(rectOp2i),
TEST(rectOp1i),
TEST(issue1418b),
turn off debugging printfs fix pathops issues 1417, 1418 be more rigorous about pulling intersections of lines to end points rewrite cubic/line and quad/line intersections to share style BUG= Review URL: https://codereview.chromium.org/19543005 git-svn-id: http://skia.googlecode.com/svn/trunk@10270 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-07-23 15:27:41 +00:00
TEST(cubicOp85i),
TEST(issue1418),
TEST(skpkkiste_to98),
path ops near exact Modify line intersections to first - match exact ends - compute intersections - match near ends where the exact ends are preferred, then near matches, then computed matches. This pulls matches towards existing end points when possible, and keeps intersection distances consistent with different line/line line/quad and line/cubic computations. BUG= Review URL: https://codereview.chromium.org/19183003 git-svn-id: http://skia.googlecode.com/svn/trunk@10073 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-07-15 13:29:13 +00:00
TEST(skpahrefs_com29),
TEST(cubicOp85d),
TEST(skpahrefs_com88),
path ops work in progress BUG= Review URL: https://codereview.chromium.org/18058007 git-svn-id: http://skia.googlecode.com/svn/trunk@9908 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-07-08 17:17:02 +00:00
TEST(skphealth_com76),
TEST(skpancestry_com1),
TEST(skpbyte_com1),
TEST(skpeldorado_com_ua1),
TEST(skp96prezzi1),
path ops -- rewrite angle sort This is a major change resulting from a minor tweak. In the old code, the intersection point of two curves was shared between them, but the intersection points and end points of sorted edges was computed directly from the intersection T value. In this CL, both intersection points and sorted points are the same, and intermediate control points are computed to preserve their slope. The sort itself has been completely rewritten to be more robust and remove 'magic' checks, conditions that empirically worked but couldn't be rationalized. This CL was triggered by errors generated computing the clips of SKP files. At this point, all 73M standard tests work and at least the first troublesome SKPs work. Review URL: https://codereview.chromium.org/15338003 git-svn-id: http://skia.googlecode.com/svn/trunk@9432 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-06-04 17:59:42 +00:00
TEST(skpClip2),
TEST(skpClip1),
TEST(cubicOp84d),
TEST(cubicOp83i),
bump picture version since SkPath has changed (conics) enable conics in SkPath git-svn-id: http://skia.googlecode.com/svn/trunk@9370 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-31 15:17:50 +00:00
TEST(cubicOp82i),
TEST(cubicOp81d),
TEST(cubicOp80i),
TEST(cubicOp79u),
TEST(cubicOp78u),
TEST(cubicOp77i),
path ops -- rewrite angle sort This is a major change resulting from a minor tweak. In the old code, the intersection point of two curves was shared between them, but the intersection points and end points of sorted edges was computed directly from the intersection T value. In this CL, both intersection points and sorted points are the same, and intermediate control points are computed to preserve their slope. The sort itself has been completely rewritten to be more robust and remove 'magic' checks, conditions that empirically worked but couldn't be rationalized. This CL was triggered by errors generated computing the clips of SKP files. At this point, all 73M standard tests work and at least the first troublesome SKPs work. Review URL: https://codereview.chromium.org/15338003 git-svn-id: http://skia.googlecode.com/svn/trunk@9432 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-06-04 17:59:42 +00:00
TEST(cubicOp76u),
bump picture version since SkPath has changed (conics) enable conics in SkPath git-svn-id: http://skia.googlecode.com/svn/trunk@9370 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-31 15:17:50 +00:00
TEST(cubicOp75d),
path ops -- rewrite angle sort This is a major change resulting from a minor tweak. In the old code, the intersection point of two curves was shared between them, but the intersection points and end points of sorted edges was computed directly from the intersection T value. In this CL, both intersection points and sorted points are the same, and intermediate control points are computed to preserve their slope. The sort itself has been completely rewritten to be more robust and remove 'magic' checks, conditions that empirically worked but couldn't be rationalized. This CL was triggered by errors generated computing the clips of SKP files. At this point, all 73M standard tests work and at least the first troublesome SKPs work. Review URL: https://codereview.chromium.org/15338003 git-svn-id: http://skia.googlecode.com/svn/trunk@9432 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-06-04 17:59:42 +00:00
TEST(cubicOp74d),
bump picture version since SkPath has changed (conics) enable conics in SkPath git-svn-id: http://skia.googlecode.com/svn/trunk@9370 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-31 15:17:50 +00:00
TEST(cubicOp73d),
TEST(cubicOp72i),
TEST(cubicOp71d),
path ops -- fix skp bugs This fixes a series of bugs discovered by running the small set of Skia skp files through pathops to flatten the clips. Review URL: https://codereview.chromium.org/14798004 git-svn-id: http://skia.googlecode.com/svn/trunk@9042 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-07 18:51:31 +00:00
TEST(skp5),
TEST(skp4),
TEST(skp3),
TEST(skp2),
TEST(skp1),
path ops : fix empty-diff bug, op-in-place add some debugging around reverse diff, inverse Review URL: https://codereview.chromium.org/13851015 git-svn-id: http://skia.googlecode.com/svn/trunk@8852 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-25 11:51:54 +00:00
TEST(rRect1),
path ops -- fix skp bugs This fixes a series of bugs discovered by running the small set of Skia skp files through pathops to flatten the clips. Review URL: https://codereview.chromium.org/14798004 git-svn-id: http://skia.googlecode.com/svn/trunk@9042 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-05-07 18:51:31 +00:00
TEST(cubicOp70d),
path ops work in progress fix bugs in tests on 32 bit release Most changes revolve around pinning computed t values very close to zero and one. git-svn-id: http://skia.googlecode.com/svn/trunk@8745 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-18 15:58:21 +00:00
TEST(cubicOp69d),
TEST(cubicOp68u),
shape ops work in progress Try to fix the 32 bit build by making some math decisions more robust. Rewrite the cubic intersection special case that detects if only end points are shared. Rewrite the angle sort setup that computes whether a cubic bends to the left or right. git-svn-id: http://skia.googlecode.com/svn/trunk@8726 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-17 15:49:16 +00:00
TEST(cubicOp67u),
TEST(cubicOp66u),
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
TEST(rectOp1d),
TEST(cubicOp65d),
TEST(cubicOp64d),
TEST(cubicOp63d),
TEST(cubicOp62d),
TEST(cubicOp61d),
TEST(cubicOp60d),
TEST(cubicOp59d),
TEST(cubicOp57d),
TEST(cubicOp56d),
TEST(cubicOp55d),
TEST(cubicOp54d),
TEST(cubicOp53d),
TEST(cubicOp52d),
TEST(cubicOp51d),
TEST(cubicOp50d),
TEST(cubicOp49d),
TEST(cubicOp48d),
TEST(cubicOp47d),
TEST(cubicOp46d),
TEST(cubicOp45d),
TEST(cubicOp44d),
TEST(cubicOp43d),
TEST(cubicOp42d),
TEST(cubicOp41i),
TEST(cubicOp40d),
TEST(cubicOp39d),
TEST(cubicOp38d),
TEST(cubicOp37d),
TEST(cubicOp36u),
TEST(cubicOp35d),
TEST(cubicOp34d),
TEST(cubicOp33i),
TEST(cubicOp32d),
TEST(cubicOp31d),
TEST(cubicOp31x),
TEST(cubicOp31u),
TEST(cubicOp30d),
TEST(cubicOp29d),
TEST(cubicOp28u),
TEST(cubicOp27d),
TEST(cubicOp26d),
TEST(cubicOp25i),
TEST(testOp8d),
TEST(testDiff1),
TEST(testIntersect1),
TEST(testUnion1),
TEST(testXor1),
TEST(testDiff2),
TEST(testIntersect2),
TEST(testUnion2),
TEST(testXor2),
TEST(testOp1d),
TEST(testOp2d),
TEST(testOp3d),
TEST(testOp1u),
TEST(testOp4d),
TEST(testOp5d),
TEST(testOp6d),
TEST(testOp7d),
TEST(testOp2u),
TEST(cubicOp24d),
TEST(cubicOp23d),
TEST(cubicOp22d),
TEST(cubicOp21d),
TEST(cubicOp20d),
TEST(cubicOp19i),
TEST(cubicOp18d),
TEST(cubicOp17d),
TEST(cubicOp16d),
TEST(cubicOp15d),
TEST(cubicOp14d),
TEST(cubicOp13d),
TEST(cubicOp12d),
TEST(cubicOp11d),
TEST(cubicOp10d),
TEST(cubicOp1i),
TEST(cubicOp9d),
TEST(quadOp9d),
TEST(lineOp9d),
TEST(cubicOp8d),
TEST(cubicOp7d),
TEST(cubicOp6d),
TEST(cubicOp5d),
TEST(cubicOp3d),
TEST(cubicOp2d),
TEST(cubicOp1d),
};
path ops work in progress standardize tests use SK_ARRAY_COUNT everywhere debug why x87 differs from SIMD 64 various platform specific fixes git-svn-id: http://skia.googlecode.com/svn/trunk@8689 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-15 19:13:59 +00:00
static const size_t testCount = SK_ARRAY_COUNT(tests);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
static struct TestDesc subTests[] = {
deal more consistently with unsortable edges Improve line/curve coincident detection and resolution. This fixed the remaining simple failures. When an edge is unsortable, use the ray intersection to determine the angles' winding. Deal with degenerate segments. TBR=reed@google.com BUG=skia:3588,skia:3762 Review URL: https://codereview.chromium.org/1140813002
2015-05-13 15:23:48 +00:00
TEST(loops47i),
Path ops formerly found the topmost unprocessed edge and determined its angle sort order to initialize the winding. This never worked correctly with cubics and was flaky with paths consisting mostly of vertical edges. This replacement shoots axis-aligned rays through all intersecting edges to find the outermost one either horizontally or vertically. The resulting code is smaller and twice as fast. To support this, most of the horizontal / vertical intersection code was rewritten and standardized, and old code supporting the top-directed winding was deleted. Contours were pointed to by an SkTDArray. Instead, put them in a linked list, and designate the list head with its own class to ensure that methods that take lists of contours start at the top. This change removed a large percentage of memory allocations used by path ops. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1111333002
2015-05-11 14:21:27 +00:00
TEST(loops61i),
TEST(loops62i),
deal more consistently with unsortable edges Improve line/curve coincident detection and resolution. This fixed the remaining simple failures. When an edge is unsortable, use the ray intersection to determine the angles' winding. Deal with degenerate segments. TBR=reed@google.com BUG=skia:3588,skia:3762 Review URL: https://codereview.chromium.org/1140813002
2015-05-13 15:23:48 +00:00
TEST(issue3517),
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
};
path ops work in progress standardize tests use SK_ARRAY_COUNT everywhere debug why x87 differs from SIMD 64 various platform specific fixes git-svn-id: http://skia.googlecode.com/svn/trunk@8689 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-15 19:13:59 +00:00
static const size_t subTestCount = SK_ARRAY_COUNT(subTests);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void (*firstSubTest)(skiatest::Reporter* , const char* filename) = 0;
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
deal more consistently with unsortable edges Improve line/curve coincident detection and resolution. This fixed the remaining simple failures. When an edge is unsortable, use the ray intersection to determine the angles' winding. Deal with degenerate segments. TBR=reed@google.com BUG=skia:3588,skia:3762 Review URL: https://codereview.chromium.org/1140813002
2015-05-13 15:23:48 +00:00
static bool runSubTests = false;
working on initial winding for cubics Path ops works well for all tests except for cubics. Isolate failures caused by cubics, and do a better job of computing the initial winding for cubics. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1096923003
2015-04-23 16:13:37 +00:00
static bool runSubTestsFirst = true;
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
static bool runReverse = false;
Get rid of DEFINE_TESTCLASS_SHORT() macro. Instead tests should be written using DEF_TEST() macro, which is much nicer and simplifies the process of setting up an unit test. BUG=None TEST=skpskgr_test, pathops_unittest R=mtklein@google.com Review URL: https://codereview.chromium.org/117863005 git-svn-id: http://skia.googlecode.com/svn/trunk@12870 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-01-02 21:45:03 +00:00
DEF_TEST(PathOpsOp, reporter) {
path ops work in progress BUG= Review URL: https://codereview.chromium.org/18058007 git-svn-id: http://skia.googlecode.com/svn/trunk@9908 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-07-08 17:17:02 +00:00
#if DEBUG_SHOW_TEST_NAME
strncpy(DEBUG_FILENAME_STRING, "", DEBUG_FILENAME_STRING_LENGTH);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
#endif
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
if (runSubTests && runSubTestsFirst) {
Style Change: NULL->nullptr DOCS_PREVIEW= https://skia.org/?cl=1316233002 Review URL: https://codereview.chromium.org/1316233002
2015-08-27 14:41:13 +00:00
RunTestSet(reporter, subTests, subTestCount, firstSubTest, nullptr, stopTest, runReverse);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
cumulative pathops patch Replace the implicit curve intersection with a geometric curve intersection. The implicit intersection proved mathematically unstable and took a long time to zero in on an answer. Use pointers instead of indices to refer to parts of curves. Indices required awkward renumbering. Unify t and point values so that small intervals can be eliminated in one pass. Break cubics up front to eliminate loops and cusps. Make the Simplify and Op code more regular and eliminate arbitrary differences. Add a builder that takes an array of paths and operators. Delete unused code. BUG=skia:3588 R=reed@google.com Review URL: https://codereview.chromium.org/1037573004
2015-03-26 14:52:43 +00:00
RunTestSet(reporter, tests, testCount, firstTest, skipTest, stopTest, runReverse);
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
if (runSubTests && !runSubTestsFirst) {
Style Change: NULL->nullptr DOCS_PREVIEW= https://skia.org/?cl=1316233002 Review URL: https://codereview.chromium.org/1316233002
2015-08-27 14:41:13 +00:00
RunTestSet(reporter, subTests, subTestCount, firstSubTest, nullptr, stopTest, runReverse);
Add implementation of path ops This CL depends on https://codereview.chromium.org/12880016/ "Add intersections for path ops" Given a path, iterate through its contour, and construct an array of segments containing its curves. Intersect each curve with every other curve, and for cubics, with itself. Given the set of intersections, find one with the smallest y and sort the curves eminating from the intersection. Assign each curve a winding value. Operate on the curves, keeping and discarding them according to the current operation and the sum of the winding values. Assemble the kept curves into an output path. Review URL: https://codereview.chromium.org/13094010 git-svn-id: http://skia.googlecode.com/svn/trunk@8553 2bbb7eff-a529-9590-31e7-b0007b416f81
2013-04-08 11:50:46 +00:00
}
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
fix path ops fuzz bug If one path is empty and the other has extreme values, the intermediate coincident paths cannot be resolved, but triggers an assert that a data structure unexpectedly has zero-length. Tunnel this failure back up to the top and return that the entire path op fails. A future optimization could detect the empty path and avoid this, allowing the op to succeed -- not sure that it's worth the additional logic though. TBR=reed@google.com BUG=535151 GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1730293002 Review URL: https://codereview.chromium.org/1730293002
2016-02-24 17:03:07 +00:00
static void fuzz535151(skiatest::Reporter* reporter, const char* filename) {
SkPath one;
one.setFillType(SkPath::kWinding_FillType);
SkPath two;
two.setFillType(SkPath::kWinding_FillType);
two.moveTo(0, 0);
two.lineTo(0, 50);
two.lineTo(4.29497e+09f, 50);
SkPath dummy;
pathops coincidence and security rewrite Most changes stem from working on an examples bracketed by #if DEBUG_UNDER_DEVELOPMENT // tiger These exposed many problems with coincident curves, as well as errors throughout the code. Fixing these errors also fixed a number of fuzzer-inspired bug reports. * Line/Curve Intersections Check to see if the end of the line nearly intersects the curve. This was a FIXME in the old code. * Performance Use a central chunk allocator. Plumb the allocator into the global variable state so that it can be shared. (Note that 'SkGlobalState' is allocated on the stack and is visible to children functions but not other threads.) * Refactor Let SkOpAngle grow up from a structure to a class. Let SkCoincidentSpans grow up from a structure to a class. Rename enum Alias to AliasMatch. * Coincidence Rewrite Add more debugging to coincidence detection. Parallel debugging routines have read-only logic to report the current coincidence state so that steps through the logic can expose whether things got better or worse. More functions can error-out and cause the pathops engine to non-destructively exit. * Accuracy Remove code that adjusted point locations. Instead, offset the curve part so that sorted curves all use the same origin. Reduce the size (and influence) of magic numbers. * Testing The debug suite with verify and the full release suite ./out/Debug/pathops_unittest -v -V ./out/Release/pathops_unittest -v -V -x expose one error. That error is captured as cubics_d3. This error exists in the checked in code as well. BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 Review-Url: https://codereview.chromium.org/2128633003
2016-07-18 17:01:36 +00:00
testPathOp(reporter, one, two, kIntersect_SkPathOp, filename);
fix path ops fuzz bug If one path is empty and the other has extreme values, the intermediate coincident paths cannot be resolved, but triggers an assert that a data structure unexpectedly has zero-length. Tunnel this failure back up to the top and return that the entire path op fails. A future optimization could detect the empty path and avoid this, allowing the op to succeed -- not sure that it's worth the additional logic though. TBR=reed@google.com BUG=535151 GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1730293002 Review URL: https://codereview.chromium.org/1730293002
2016-02-24 17:03:07 +00:00
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static void bufferOverflow(skiatest::Reporter* reporter, const char* filename) {
SkPath path;
path.addRect(0,0, 300,170141183460469231731687303715884105728.f);
SkPath pathB;
pathB.addRect(0,0, 300,16);
pathops coincidence and security rewrite Most changes stem from working on an examples bracketed by #if DEBUG_UNDER_DEVELOPMENT // tiger These exposed many problems with coincident curves, as well as errors throughout the code. Fixing these errors also fixed a number of fuzzer-inspired bug reports. * Line/Curve Intersections Check to see if the end of the line nearly intersects the curve. This was a FIXME in the old code. * Performance Use a central chunk allocator. Plumb the allocator into the global variable state so that it can be shared. (Note that 'SkGlobalState' is allocated on the stack and is visible to children functions but not other threads.) * Refactor Let SkOpAngle grow up from a structure to a class. Let SkCoincidentSpans grow up from a structure to a class. Rename enum Alias to AliasMatch. * Coincidence Rewrite Add more debugging to coincidence detection. Parallel debugging routines have read-only logic to report the current coincidence state so that steps through the logic can expose whether things got better or worse. More functions can error-out and cause the pathops engine to non-destructively exit. * Accuracy Remove code that adjusted point locations. Instead, offset the curve part so that sorted curves all use the same origin. Reduce the size (and influence) of magic numbers. * Testing The debug suite with verify and the full release suite ./out/Debug/pathops_unittest -v -V ./out/Release/pathops_unittest -v -V -x expose one error. That error is captured as cubics_d3. This error exists in the checked in code as well. BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 Review-Url: https://codereview.chromium.org/2128633003
2016-07-18 17:01:36 +00:00
testPathOpCheck(reporter, path, pathB, kUnion_SkPathOp, filename, true);
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
}
fail when coincidence is too far apart TBR= BUG=410552 Author: caryclark@google.com Review URL: https://codereview.chromium.org/556433002
2014-09-08 17:25:38 +00:00
// m 100,0 60,170 -160,-110 200,0 -170,11000000000 z
static void fuzz433(skiatest::Reporter* reporter, const char* filename) {
SkPath path1, path2;
path1.moveTo(100,0);
path1.lineTo(60,170);
path1.lineTo(-160,-110);
path1.lineTo(200,0);
path1.lineTo(-170,11000000000.0f);
path1.close();
path2.moveTo(100 + 20,0 + 20);
path2.lineTo(60 + 20,170 + 20);
path2.lineTo(-160 + 20,-110 + 20);
path2.lineTo(200 + 20,0 + 20);
path2.lineTo(-170 + 20,11000000000.0f + 20);
path2.close();
pathops coincidence and security rewrite Most changes stem from working on an examples bracketed by #if DEBUG_UNDER_DEVELOPMENT // tiger These exposed many problems with coincident curves, as well as errors throughout the code. Fixing these errors also fixed a number of fuzzer-inspired bug reports. * Line/Curve Intersections Check to see if the end of the line nearly intersects the curve. This was a FIXME in the old code. * Performance Use a central chunk allocator. Plumb the allocator into the global variable state so that it can be shared. (Note that 'SkGlobalState' is allocated on the stack and is visible to children functions but not other threads.) * Refactor Let SkOpAngle grow up from a structure to a class. Let SkCoincidentSpans grow up from a structure to a class. Rename enum Alias to AliasMatch. * Coincidence Rewrite Add more debugging to coincidence detection. Parallel debugging routines have read-only logic to report the current coincidence state so that steps through the logic can expose whether things got better or worse. More functions can error-out and cause the pathops engine to non-destructively exit. * Accuracy Remove code that adjusted point locations. Instead, offset the curve part so that sorted curves all use the same origin. Reduce the size (and influence) of magic numbers. * Testing The debug suite with verify and the full release suite ./out/Debug/pathops_unittest -v -V ./out/Release/pathops_unittest -v -V -x expose one error. That error is captured as cubics_d3. This error exists in the checked in code as well. BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 Review-Url: https://codereview.chromium.org/2128633003
2016-07-18 17:01:36 +00:00
testPathOpCheck(reporter, path1, path2, kIntersect_SkPathOp, filename, true);
fail when coincidence is too far apart TBR= BUG=410552 Author: caryclark@google.com Review URL: https://codereview.chromium.org/556433002
2014-09-08 17:25:38 +00:00
}
static void fuzz433b(skiatest::Reporter* reporter, const char* filename) {
SkPath path1, path2;
path1.setFillType(SkPath::kEvenOdd_FillType);
path1.moveTo(140, 40);
path1.lineTo(200, 210);
path1.lineTo(40, 100);
path1.lineTo(240, 100);
path1.lineTo(70, 1.1e+10f);
path1.lineTo(140, 40);
path1.close();
path1.setFillType(SkPath::kWinding_FillType);
path2.moveTo(190, 60);
path2.lineTo(250, 230);
path2.lineTo(90, 120);
path2.lineTo(290, 120);
path2.lineTo(120, 1.1e+10f);
path2.lineTo(190, 60);
path2.close();
pathops coincidence and security rewrite Most changes stem from working on an examples bracketed by #if DEBUG_UNDER_DEVELOPMENT // tiger These exposed many problems with coincident curves, as well as errors throughout the code. Fixing these errors also fixed a number of fuzzer-inspired bug reports. * Line/Curve Intersections Check to see if the end of the line nearly intersects the curve. This was a FIXME in the old code. * Performance Use a central chunk allocator. Plumb the allocator into the global variable state so that it can be shared. (Note that 'SkGlobalState' is allocated on the stack and is visible to children functions but not other threads.) * Refactor Let SkOpAngle grow up from a structure to a class. Let SkCoincidentSpans grow up from a structure to a class. Rename enum Alias to AliasMatch. * Coincidence Rewrite Add more debugging to coincidence detection. Parallel debugging routines have read-only logic to report the current coincidence state so that steps through the logic can expose whether things got better or worse. More functions can error-out and cause the pathops engine to non-destructively exit. * Accuracy Remove code that adjusted point locations. Instead, offset the curve part so that sorted curves all use the same origin. Reduce the size (and influence) of magic numbers. * Testing The debug suite with verify and the full release suite ./out/Debug/pathops_unittest -v -V ./out/Release/pathops_unittest -v -V -x expose one error. That error is captured as cubics_d3. This error exists in the checked in code as well. BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 Review-Url: https://codereview.chromium.org/2128633003
2016-07-18 17:01:36 +00:00
testPathOpCheck(reporter, path1, path2, kUnion_SkPathOp, filename, true);
fail when coincidence is too far apart TBR= BUG=410552 Author: caryclark@google.com Review URL: https://codereview.chromium.org/556433002
2014-09-08 17:25:38 +00:00
}
fail early if coincidence can't be resolved Bail out if a very large value causes coincidence resolution to fail. TBR= BUG=415866 Author: caryclark@google.com Review URL: https://codereview.chromium.org/585913002
2014-09-19 13:33:31 +00:00
static void fuzz487a(skiatest::Reporter* reporter, const char* filename) {
SkPath path;
path.setFillType((SkPath::FillType) 0);
path.moveTo(SkBits2Float(0x432c8000), SkBits2Float(0x42c00000));
path.lineTo(SkBits2Float(0x4309999a), SkBits2Float(0x42c00000));
path.cubicTo(SkBits2Float(0x4309999a), SkBits2Float(0x429a6666), SkBits2Float(0x42f9999a), SkBits2Float(0x4275999a), SkBits2Float(0x42d70001), SkBits2Float(0x42633333));
path.lineTo(SkBits2Float(0x42e90001), SkBits2Float(0x41b8cccc));
path.cubicTo(SkBits2Float(0x42dc6667), SkBits2Float(0x41ab3332), SkBits2Float(0x42cf3334), SkBits2Float(0x41a3ffff), SkBits2Float(0x42c20001), SkBits2Float(0x41a3ffff));
path.lineTo(SkBits2Float(0x42c20001), SkBits2Float(0x425d999a));
path.lineTo(SkBits2Float(0x42c20001), SkBits2Float(0x425d999a));
path.cubicTo(SkBits2Float(0x429c6668), SkBits2Float(0x425d999a), SkBits2Float(0x4279999c), SkBits2Float(0x42886667), SkBits2Float(0x42673335), SkBits2Float(0x42ab0000));
path.lineTo(SkBits2Float(0x41c0ccd0), SkBits2Float(0x42990000));
path.cubicTo(SkBits2Float(0x41b33336), SkBits2Float(0x42a5999a), SkBits2Float(0x41ac0003), SkBits2Float(0x42b2cccd), SkBits2Float(0x41ac0003), SkBits2Float(0x42c00000));
path.lineTo(SkBits2Float(0x4261999c), SkBits2Float(0x42c00000));
path.lineTo(SkBits2Float(0x4261999c), SkBits2Float(0x42c00000));
path.cubicTo(SkBits2Float(0x4261999c), SkBits2Float(0x434d3333), SkBits2Float(0x4364e667), SkBits2Float(0x4346b333), SkBits2Float(0x4364e667), SkBits2Float(0x43400000));
path.lineTo(SkBits2Float(0x432c8000), SkBits2Float(0x42c00000));
path.close();
SkPath path1(path);
path.reset();
path.setFillType((SkPath::FillType) 0);
path.moveTo(SkBits2Float(0x432c8000), SkBits2Float(0x42c00000));
path.lineTo(SkBits2Float(0x4309999a), SkBits2Float(0x42c00000));
path.cubicTo(SkBits2Float(0x4309999a), SkBits2Float(0x42a20000), SkBits2Float(0x43016667), SkBits2Float(0x4287cccd), SkBits2Float(0x42ea999a), SkBits2Float(0x4273999a));
path.lineTo(SkBits2Float(0x4306cccd), SkBits2Float(0x41f5999a));
path.cubicTo(SkBits2Float(0x42f76667), SkBits2Float(0x41c26667), SkBits2Float(0x42dd999a), SkBits2Float(0x41a4cccd), SkBits2Float(0x42c23334), SkBits2Float(0x41a4cccd));
path.lineTo(SkBits2Float(0x42c23334), SkBits2Float(0x425e0000));
path.cubicTo(SkBits2Float(0x42a43334), SkBits2Float(0x425e0000), SkBits2Float(0x428a0001), SkBits2Float(0x427ecccd), SkBits2Float(0x42780002), SkBits2Float(0x4297999a));
path.lineTo(SkBits2Float(0x41fccccd), SkBits2Float(0x42693333));
path.cubicTo(SkBits2Float(0x41c9999a), SkBits2Float(0x428acccd), SkBits2Float(0x41ac0000), SkBits2Float(0x42a4999a), SkBits2Float(0x41ac0000), SkBits2Float(0x42c00000));
path.lineTo(SkBits2Float(0x4261999a), SkBits2Float(0x42c00000));
path.cubicTo(SkBits2Float(0x4261999a), SkBits2Float(0x42de0000), SkBits2Float(0x42813333), SkBits2Float(0x42f83333), SkBits2Float(0x42996666), SkBits2Float(0x4303199a));
path.cubicTo(SkBits2Float(0x4272cccc), SkBits2Float(0x4303199a), SkBits2Float(0x423d3332), SkBits2Float(0x430de667), SkBits2Float(0x422d9999), SkBits2Float(0x431cb334));
path.lineTo(SkBits2Float(0x7086a1dc), SkBits2Float(0x42eecccd));
path.lineTo(SkBits2Float(0x41eb3333), SkBits2Float(0xc12ccccd));
path.lineTo(SkBits2Float(0x42053333), SkBits2Float(0xc1cccccd));
path.lineTo(SkBits2Float(0x42780000), SkBits2Float(0xc18f3334));
path.cubicTo(SkBits2Float(0x43206666), SkBits2Float(0x43134ccd), SkBits2Float(0x43213333), SkBits2Float(0x430db333), SkBits2Float(0x43213333), SkBits2Float(0x43080000));
path.lineTo(SkBits2Float(0x432c8000), SkBits2Float(0x42c00000));
path.close();
SkPath path2(path);
pathops coincidence and security rewrite Most changes stem from working on an examples bracketed by #if DEBUG_UNDER_DEVELOPMENT // tiger These exposed many problems with coincident curves, as well as errors throughout the code. Fixing these errors also fixed a number of fuzzer-inspired bug reports. * Line/Curve Intersections Check to see if the end of the line nearly intersects the curve. This was a FIXME in the old code. * Performance Use a central chunk allocator. Plumb the allocator into the global variable state so that it can be shared. (Note that 'SkGlobalState' is allocated on the stack and is visible to children functions but not other threads.) * Refactor Let SkOpAngle grow up from a structure to a class. Let SkCoincidentSpans grow up from a structure to a class. Rename enum Alias to AliasMatch. * Coincidence Rewrite Add more debugging to coincidence detection. Parallel debugging routines have read-only logic to report the current coincidence state so that steps through the logic can expose whether things got better or worse. More functions can error-out and cause the pathops engine to non-destructively exit. * Accuracy Remove code that adjusted point locations. Instead, offset the curve part so that sorted curves all use the same origin. Reduce the size (and influence) of magic numbers. * Testing The debug suite with verify and the full release suite ./out/Debug/pathops_unittest -v -V ./out/Release/pathops_unittest -v -V -x expose one error. That error is captured as cubics_d3. This error exists in the checked in code as well. BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 Review-Url: https://codereview.chromium.org/2128633003
2016-07-18 17:01:36 +00:00
testPathOpCheck(reporter, path1, path2, (SkPathOp) 2, filename, true);
fail early if coincidence can't be resolved Bail out if a very large value causes coincidence resolution to fail. TBR= BUG=415866 Author: caryclark@google.com Review URL: https://codereview.chromium.org/585913002
2014-09-19 13:33:31 +00:00
}
static void fuzz487b(skiatest::Reporter* reporter, const char* filename) {
SkPath path;
path.setFillType((SkPath::FillType) 0);
path.moveTo(SkBits2Float(0x432c8000), SkBits2Float(0x42c00000));
path.lineTo(SkBits2Float(0x4309999a), SkBits2Float(0x42c00000));
path.cubicTo(SkBits2Float(0x4309999a), SkBits2Float(0x429a6666), SkBits2Float(0x42f9999a), SkBits2Float(0x4275999a), SkBits2Float(0x42d70001), SkBits2Float(0x42633333));
path.lineTo(SkBits2Float(0x42e90001), SkBits2Float(0x41b8cccc));
path.cubicTo(SkBits2Float(0x42dc6667), SkBits2Float(0x41ab3332), SkBits2Float(0x42cf3334), SkBits2Float(0x41a3ffff), SkBits2Float(0x42c20001), SkBits2Float(0x41a3ffff));
path.lineTo(SkBits2Float(0x42c20001), SkBits2Float(0x425d999a));
path.lineTo(SkBits2Float(0x42c20001), SkBits2Float(0x425d999a));
path.cubicTo(SkBits2Float(0x429c6668), SkBits2Float(0x425d999a), SkBits2Float(0x4279999c), SkBits2Float(0x42886667), SkBits2Float(0x42673335), SkBits2Float(0x42ab0000));
path.lineTo(SkBits2Float(0x41c0ccd0), SkBits2Float(0x42990000));
path.cubicTo(SkBits2Float(0x41b33336), SkBits2Float(0x42a5999a), SkBits2Float(0x41ac0003), SkBits2Float(0x42b2cccd), SkBits2Float(0x41ac0003), SkBits2Float(0x42c00000));
path.lineTo(SkBits2Float(0x4261999c), SkBits2Float(0x42c00000));
path.lineTo(SkBits2Float(0x4261999c), SkBits2Float(0x42c00000));
path.cubicTo(SkBits2Float(0x4261999c), SkBits2Float(0x434d3333), SkBits2Float(0x4364e667), SkBits2Float(0x4346b333), SkBits2Float(0x4364e667), SkBits2Float(0x43400000));
path.lineTo(SkBits2Float(0x432c8000), SkBits2Float(0x42c00000));
path.close();
SkPath path1(path);
path.reset();
path.setFillType((SkPath::FillType) 0);
path.moveTo(SkBits2Float(0x432c8000), SkBits2Float(0x42c00000));
path.lineTo(SkBits2Float(0x4309999a), SkBits2Float(0x42c00000));
path.cubicTo(SkBits2Float(0x4309999a), SkBits2Float(0x42a20000), SkBits2Float(0x43016667), SkBits2Float(0x4287cccd), SkBits2Float(0x42ea999a), SkBits2Float(0x4273999a));
path.lineTo(SkBits2Float(0x4306cccd), SkBits2Float(0x41f5999a));
path.cubicTo(SkBits2Float(0x42f76667), SkBits2Float(0x41c26667), SkBits2Float(0x42dd999a), SkBits2Float(0x41a4cccd), SkBits2Float(0x42c23334), SkBits2Float(0x41a4cccd));
path.lineTo(SkBits2Float(0x42c23334), SkBits2Float(0x425e0000));
path.cubicTo(SkBits2Float(0x42a43334), SkBits2Float(0x425e0000), SkBits2Float(0x428a0001), SkBits2Float(0x427ecccd), SkBits2Float(0x42780002), SkBits2Float(0x4297999a));
path.lineTo(SkBits2Float(0x41fccccd), SkBits2Float(0x42693333));
path.cubicTo(SkBits2Float(0x41c9999a), SkBits2Float(0x428acccd), SkBits2Float(0x41ac0000), SkBits2Float(0x42a4999a), SkBits2Float(0x41ac0000), SkBits2Float(0x42c00000));
path.lineTo(SkBits2Float(0x4261999a), SkBits2Float(0x42c00000));
path.cubicTo(SkBits2Float(0x4261999a), SkBits2Float(0x42de0000), SkBits2Float(0x42813333), SkBits2Float(0x42f83333), SkBits2Float(0x42996666), SkBits2Float(0x4303199a));
path.cubicTo(SkBits2Float(0x4272cccc), SkBits2Float(0x4303199a), SkBits2Float(0x423d3332), SkBits2Float(0x430de667), SkBits2Float(0x422d9999), SkBits2Float(0x431cb334));
path.lineTo(SkBits2Float(0x7086a1dc), SkBits2Float(0x42eecccd));
path.lineTo(SkBits2Float(0x41eb3333), SkBits2Float(0xc12ccccd));
path.lineTo(SkBits2Float(0x42053333), SkBits2Float(0xc1cccccd));
path.lineTo(SkBits2Float(0x42780000), SkBits2Float(0xc18f3334));
path.cubicTo(SkBits2Float(0x43206666), SkBits2Float(0x43134ccd), SkBits2Float(0x43213333), SkBits2Float(0x430db333), SkBits2Float(0x43213333), SkBits2Float(0x43080000));
path.lineTo(SkBits2Float(0x432c8000), SkBits2Float(0x42c00000));
path.close();
SkPath path2(path);
pathops coincidence and security rewrite Most changes stem from working on an examples bracketed by #if DEBUG_UNDER_DEVELOPMENT // tiger These exposed many problems with coincident curves, as well as errors throughout the code. Fixing these errors also fixed a number of fuzzer-inspired bug reports. * Line/Curve Intersections Check to see if the end of the line nearly intersects the curve. This was a FIXME in the old code. * Performance Use a central chunk allocator. Plumb the allocator into the global variable state so that it can be shared. (Note that 'SkGlobalState' is allocated on the stack and is visible to children functions but not other threads.) * Refactor Let SkOpAngle grow up from a structure to a class. Let SkCoincidentSpans grow up from a structure to a class. Rename enum Alias to AliasMatch. * Coincidence Rewrite Add more debugging to coincidence detection. Parallel debugging routines have read-only logic to report the current coincidence state so that steps through the logic can expose whether things got better or worse. More functions can error-out and cause the pathops engine to non-destructively exit. * Accuracy Remove code that adjusted point locations. Instead, offset the curve part so that sorted curves all use the same origin. Reduce the size (and influence) of magic numbers. * Testing The debug suite with verify and the full release suite ./out/Debug/pathops_unittest -v -V ./out/Release/pathops_unittest -v -V -x expose one error. That error is captured as cubics_d3. This error exists in the checked in code as well. BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 Review-Url: https://codereview.chromium.org/2128633003
2016-07-18 17:01:36 +00:00
testPathOpCheck(reporter, path1, path2, (SkPathOp) 2, filename, true);
fail early if coincidence can't be resolved Bail out if a very large value causes coincidence resolution to fail. TBR= BUG=415866 Author: caryclark@google.com Review URL: https://codereview.chromium.org/585913002
2014-09-19 13:33:31 +00:00
}
fail on extremely large coincident curves TBR= BUG=418381 Author: caryclark@google.com Review URL: https://codereview.chromium.org/607913007
2014-09-29 13:58:41 +00:00
static void fuzz714(skiatest::Reporter* reporter, const char* filename) {
SkPath path;
path.setFillType((SkPath::FillType) 1);
path.moveTo(SkBits2Float(0x430c0000), SkBits2Float(0x42200000));
path.lineTo(SkBits2Float(0x43480000), SkBits2Float(0x43520000));
path.lineTo(SkBits2Float(0x42200000), SkBits2Float(0x42c80000));
fix multiple intersection logic When three or more curves intersect at the same point, ensure that each curve records the intersections of the others. This fixes a number of cubic tests. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1105943002
2015-04-24 16:08:57 +00:00
path.lineTo(SkBits2Float(0x64969569), SkBits2Float(0x42c80000)); // 2.22222e+022f
path.lineTo(SkBits2Float(0x64969569), SkBits2Float(0x43520000)); // 2.22222e+022f
fail on extremely large coincident curves TBR= BUG=418381 Author: caryclark@google.com Review URL: https://codereview.chromium.org/607913007
2014-09-29 13:58:41 +00:00
path.lineTo(SkBits2Float(0x430c0000), SkBits2Float(0x42200000));
path.close();
SkPath path1(path);
path.reset();
path.setFillType((SkPath::FillType) 0);
path.moveTo(SkBits2Float(0x43200000), SkBits2Float(0x42700000));
path.lineTo(SkBits2Float(0x435c0000), SkBits2Float(0x43660000));
path.lineTo(SkBits2Float(0x42700000), SkBits2Float(0x42f00000));
fix multiple intersection logic When three or more curves intersect at the same point, ensure that each curve records the intersections of the others. This fixes a number of cubic tests. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1105943002
2015-04-24 16:08:57 +00:00
path.lineTo(SkBits2Float(0x64969569), SkBits2Float(0x42f00000)); // 2.22222e+022f
path.lineTo(SkBits2Float(0x64969569), SkBits2Float(0x43660000)); // 2.22222e+022f
fail on extremely large coincident curves TBR= BUG=418381 Author: caryclark@google.com Review URL: https://codereview.chromium.org/607913007
2014-09-29 13:58:41 +00:00
path.lineTo(SkBits2Float(0x43200000), SkBits2Float(0x42700000));
path.close();
SkPath path2(path);
pathops coincidence and security rewrite Most changes stem from working on an examples bracketed by #if DEBUG_UNDER_DEVELOPMENT // tiger These exposed many problems with coincident curves, as well as errors throughout the code. Fixing these errors also fixed a number of fuzzer-inspired bug reports. * Line/Curve Intersections Check to see if the end of the line nearly intersects the curve. This was a FIXME in the old code. * Performance Use a central chunk allocator. Plumb the allocator into the global variable state so that it can be shared. (Note that 'SkGlobalState' is allocated on the stack and is visible to children functions but not other threads.) * Refactor Let SkOpAngle grow up from a structure to a class. Let SkCoincidentSpans grow up from a structure to a class. Rename enum Alias to AliasMatch. * Coincidence Rewrite Add more debugging to coincidence detection. Parallel debugging routines have read-only logic to report the current coincidence state so that steps through the logic can expose whether things got better or worse. More functions can error-out and cause the pathops engine to non-destructively exit. * Accuracy Remove code that adjusted point locations. Instead, offset the curve part so that sorted curves all use the same origin. Reduce the size (and influence) of magic numbers. * Testing The debug suite with verify and the full release suite ./out/Debug/pathops_unittest -v -V ./out/Release/pathops_unittest -v -V -x expose one error. That error is captured as cubics_d3. This error exists in the checked in code as well. BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 Review-Url: https://codereview.chromium.org/2128633003
2016-07-18 17:01:36 +00:00
testPathOpCheck(reporter, path1, path2, (SkPathOp) 2, filename, true);
fail on extremely large coincident curves TBR= BUG=418381 Author: caryclark@google.com Review URL: https://codereview.chromium.org/607913007
2014-09-29 13:58:41 +00:00
}
abort op early if path isn't parseable TBR= BUG=419649 Review URL: https://codereview.chromium.org/623943002
2014-10-03 12:36:27 +00:00
static void fuzz1(skiatest::Reporter* reporter, const char* filename) {
SkPath path;
path.setFillType((SkPath::FillType) 0);
path.moveTo(SkBits2Float(0x7f800000), SkBits2Float(0x7f800000));
path.quadTo(SkBits2Float(0x7f800000), SkBits2Float(0x7f800000), SkBits2Float(0x7f800000), SkBits2Float(0x7f800000));
path.quadTo(SkBits2Float(0x7f800000), SkBits2Float(0x7f800000), SkBits2Float(0x7f800000), SkBits2Float(0x7f800000));
path.quadTo(SkBits2Float(0xffc00000), SkBits2Float(0x7f800000), SkBits2Float(0xffc00000), SkBits2Float(0x7f800000));
path.quadTo(SkBits2Float(0xff000001), SkBits2Float(0x7f800000), SkBits2Float(0xff000001), SkBits2Float(0x7f800000));
path.quadTo(SkBits2Float(0xff000001), SkBits2Float(0xffc00000), SkBits2Float(0xffc00000), SkBits2Float(0xffc00000));
path.quadTo(SkBits2Float(0xffc00000), SkBits2Float(0xff000001), SkBits2Float(0x7f800000), SkBits2Float(0xff000001));
path.quadTo(SkBits2Float(0x7f800000), SkBits2Float(0xff000001), SkBits2Float(0x7f800000), SkBits2Float(0xffc00000));
path.quadTo(SkBits2Float(0x7f800000), SkBits2Float(0xffc00000), SkBits2Float(0x7f800000), SkBits2Float(0x7f800000));
path.close();
SkPath path1(path);
path.reset();
path.setFillType((SkPath::FillType) 0);
SkPath path2(path);
pathops coincidence and security rewrite Most changes stem from working on an examples bracketed by #if DEBUG_UNDER_DEVELOPMENT // tiger These exposed many problems with coincident curves, as well as errors throughout the code. Fixing these errors also fixed a number of fuzzer-inspired bug reports. * Line/Curve Intersections Check to see if the end of the line nearly intersects the curve. This was a FIXME in the old code. * Performance Use a central chunk allocator. Plumb the allocator into the global variable state so that it can be shared. (Note that 'SkGlobalState' is allocated on the stack and is visible to children functions but not other threads.) * Refactor Let SkOpAngle grow up from a structure to a class. Let SkCoincidentSpans grow up from a structure to a class. Rename enum Alias to AliasMatch. * Coincidence Rewrite Add more debugging to coincidence detection. Parallel debugging routines have read-only logic to report the current coincidence state so that steps through the logic can expose whether things got better or worse. More functions can error-out and cause the pathops engine to non-destructively exit. * Accuracy Remove code that adjusted point locations. Instead, offset the curve part so that sorted curves all use the same origin. Reduce the size (and influence) of magic numbers. * Testing The debug suite with verify and the full release suite ./out/Debug/pathops_unittest -v -V ./out/Release/pathops_unittest -v -V -x expose one error. That error is captured as cubics_d3. This error exists in the checked in code as well. BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 Review-Url: https://codereview.chromium.org/2128633003
2016-07-18 17:01:36 +00:00
testPathOpFail(reporter, path1, path2, (SkPathOp) 2, filename);
abort op early if path isn't parseable TBR= BUG=419649 Review URL: https://codereview.chromium.org/623943002
2014-10-03 12:36:27 +00:00
}
path ops: fix fuzz-found divide by zero TBR=reed@google.com BUG=542391 Review URL: https://codereview.chromium.org/1427913005
2015-11-02 15:32:58 +00:00
static void fuzz753_91(skiatest::Reporter* reporter, const char* filename) {
SkPath path;
path.setFillType((SkPath::FillType) 0);
path.moveTo(SkBits2Float(0x42910000), SkBits2Float(0x00000000)); // 72.5f, 0
path.lineTo(SkBits2Float(0x42166668), SkBits2Float(0x00000000)); // 37.6f, 0
path.cubicTo(SkBits2Float(0x42166668), SkBits2Float(0xc1966668), SkBits2Float(0x41c66668), SkBits2Float(0xc20a6666), SkBits2Float(0x40f00010), SkBits2Float(0xc21ccccd)); // 37.6f, -18.8f, 24.8f, -34.6f, 7.50001f, -39.2f
path.lineTo(SkBits2Float(0x41840004), SkBits2Float(0xc291cccd)); // 16.5f, -72.9f
path.lineTo(SkBits2Float(0x42fb6668), SkBits2Float(0x42c73334)); // 125.7f, 99.6f
path.lineTo(SkBits2Float(0x43646668), SkBits2Float(0x43880ccd)); // 228.4f, 272.1f
SkPath path1(path);
path.reset();
path.setFillType((SkPath::FillType) 0);
path.moveTo(SkBits2Float(0x428bf702), SkBits2Float(0xcf223cbf)); // 69.9824f, -2.72189e+09f
path.lineTo(SkBits2Float(0x42112d68), SkBits2Float(0xcf223cbf)); // 36.2943f, -2.72189e+09f
path.cubicTo(SkBits2Float(0x4220d9fc), SkBits2Float(0xcf223cc0), SkBits2Float(0x420ee118), SkBits2Float(0xcf223cc0), SkBits2Float(0x41cef2f8), SkBits2Float(0xcf223cc0)); // 40.2129f, -2.72189e+09f, 35.7198f, -2.72189e+09f, 25.8686f, -2.72189e+09f
path.lineTo(SkBits2Float(0x424a99e0), SkBits2Float(0xcf223cc0)); // 50.6503f, -2.72189e+09f
path.cubicTo(SkBits2Float(0x42266e32), SkBits2Float(0xcf223cc0), SkBits2Float(0x41f0fa20), SkBits2Float(0xcf223cc0), SkBits2Float(0x41872ed4), SkBits2Float(0xcf223cc0)); // 41.6076f, -2.72189e+09f, 30.1221f, -2.72189e+09f, 16.8979f, -2.72189e+09f
path.lineTo(SkBits2Float(0x40f8fbe0), SkBits2Float(0xcf223cc0)); // 7.78075f, -2.72189e+09f
SkPath path2(path);
pathops coincidence and security rewrite Most changes stem from working on an examples bracketed by #if DEBUG_UNDER_DEVELOPMENT // tiger These exposed many problems with coincident curves, as well as errors throughout the code. Fixing these errors also fixed a number of fuzzer-inspired bug reports. * Line/Curve Intersections Check to see if the end of the line nearly intersects the curve. This was a FIXME in the old code. * Performance Use a central chunk allocator. Plumb the allocator into the global variable state so that it can be shared. (Note that 'SkGlobalState' is allocated on the stack and is visible to children functions but not other threads.) * Refactor Let SkOpAngle grow up from a structure to a class. Let SkCoincidentSpans grow up from a structure to a class. Rename enum Alias to AliasMatch. * Coincidence Rewrite Add more debugging to coincidence detection. Parallel debugging routines have read-only logic to report the current coincidence state so that steps through the logic can expose whether things got better or worse. More functions can error-out and cause the pathops engine to non-destructively exit. * Accuracy Remove code that adjusted point locations. Instead, offset the curve part so that sorted curves all use the same origin. Reduce the size (and influence) of magic numbers. * Testing The debug suite with verify and the full release suite ./out/Debug/pathops_unittest -v -V ./out/Release/pathops_unittest -v -V -x expose one error. That error is captured as cubics_d3. This error exists in the checked in code as well. BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 Review-Url: https://codereview.chromium.org/2128633003
2016-07-18 17:01:36 +00:00
testPathOp(reporter, path1, path2, (SkPathOp) 2, filename);
path ops: fix fuzz-found divide by zero TBR=reed@google.com BUG=542391 Review URL: https://codereview.chromium.org/1427913005
2015-11-02 15:32:58 +00:00
}
give up if huge paths have unresolvable coincidence This fuzzy test has enormous curves with coincidence runs that break numerics. If the computed intersections identify that the span of coincidence has been deleted, give up and return that the path op failed. TBR=reed@google.com BUG=597926 GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1854333002 Review URL: https://codereview.chromium.org/1854333002
2016-04-05 14:28:48 +00:00
static void bug597926_0(skiatest::Reporter* reporter, const char* filename) {
SkPath path;
path.setFillType((SkPath::FillType) 0);
path.moveTo(SkBits2Float(0x43b38000), SkBits2Float(0x433e0000)); // 359, 190
path.lineTo(SkBits2Float(0x40c00000), SkBits2Float(0x449ce000)); // 6, 1255
path.cubicTo(SkBits2Float(0x438c0000), SkBits2Float(0x4497a000), SkBits2Float(0x43e40000), SkBits2Float(0x44750000), SkBits2Float(0x41000000), SkBits2Float(0x44aa2000)); // 280, 1213, 456, 980, 8, 1361
path.moveTo(SkBits2Float(0x43290000), SkBits2Float(0x4431c000)); // 169, 711
path.lineTo(SkBits2Float(0xd987d6ba), SkBits2Float(0xd93d0ad4)); // -4.7794e+15f, -3.32567e+15f
path.conicTo(SkBits2Float(0x43cc8000), SkBits2Float(0x445b8000), SkBits2Float(0xd888b096), SkBits2Float(0xd9a1ebfa), SkBits2Float(0x3ebcb199)); // 409, 878, -1.20234e+15f, -5.69712e+15f, 0.368542f
path.cubicTo(SkBits2Float(0x43c00000), SkBits2Float(0x443a8000), SkBits2Float(0x42380000), SkBits2Float(0x4421c000), SkBits2Float(0x42500000), SkBits2Float(0x448ca000)); // 384, 746, 46, 647, 52, 1125
path.quadTo(SkBits2Float(0x43948000), SkBits2Float(0x42ac0000), SkBits2Float(0x43880000), SkBits2Float(0x4487e000)); // 297, 86, 272, 1087
SkPath path1(path);
path.reset();
path.setFillType((SkPath::FillType) 0);
path.moveTo(SkBits2Float(0xc51d735c), SkBits2Float(0xc49db029)); // -2519.21f, -1261.51f
path.cubicTo(SkBits2Float(0xc51d1dbd), SkBits2Float(0xc49d7a3f), SkBits2Float(0xc51c524a), SkBits2Float(0xc49d1610), SkBits2Float(0xc51d1a96), SkBits2Float(0xc49d86a6)); // -2513.86f, -1259.82f, -2501.14f, -1256.69f, -2513.66f, -1260.21f
path.cubicTo(SkBits2Float(0xc51cd471), SkBits2Float(0xc49d54d0), SkBits2Float(0xc51c2e51), SkBits2Float(0xc49d0081), SkBits2Float(0xc51d197b), SkBits2Float(0xc49d7927)); // -2509.28f, -1258.65f, -2498.89f, -1256.02f, -2513.59f, -1259.79f
path.quadTo(SkBits2Float(0xc51bf7eb), SkBits2Float(0xc49cf010), SkBits2Float(0xc51ba866), SkBits2Float(0xc49cb9e6)); // -2495.49f, -1255.5f, -2490.52f, -1253.81f
path.cubicTo(SkBits2Float(0xc51bac0d), SkBits2Float(0xc49cc50e), SkBits2Float(0xc51c29eb), SkBits2Float(0xc49cfb01), SkBits2Float(0xc51c5bca), SkBits2Float(0xc49d1fa6)); // -2490.75f, -1254.16f, -2498.62f, -1255.84f, -2501.74f, -1256.99f
SkPath path2(path);
pathops coincidence and security rewrite Most changes stem from working on an examples bracketed by #if DEBUG_UNDER_DEVELOPMENT // tiger These exposed many problems with coincident curves, as well as errors throughout the code. Fixing these errors also fixed a number of fuzzer-inspired bug reports. * Line/Curve Intersections Check to see if the end of the line nearly intersects the curve. This was a FIXME in the old code. * Performance Use a central chunk allocator. Plumb the allocator into the global variable state so that it can be shared. (Note that 'SkGlobalState' is allocated on the stack and is visible to children functions but not other threads.) * Refactor Let SkOpAngle grow up from a structure to a class. Let SkCoincidentSpans grow up from a structure to a class. Rename enum Alias to AliasMatch. * Coincidence Rewrite Add more debugging to coincidence detection. Parallel debugging routines have read-only logic to report the current coincidence state so that steps through the logic can expose whether things got better or worse. More functions can error-out and cause the pathops engine to non-destructively exit. * Accuracy Remove code that adjusted point locations. Instead, offset the curve part so that sorted curves all use the same origin. Reduce the size (and influence) of magic numbers. * Testing The debug suite with verify and the full release suite ./out/Debug/pathops_unittest -v -V ./out/Release/pathops_unittest -v -V -x expose one error. That error is captured as cubics_d3. This error exists in the checked in code as well. BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 Review-Url: https://codereview.chromium.org/2128633003
2016-07-18 17:01:36 +00:00
testPathOp(reporter, path1, path2, (SkPathOp) 1, filename);
give up if huge paths have unresolvable coincidence This fuzzy test has enormous curves with coincidence runs that break numerics. If the computed intersections identify that the span of coincidence has been deleted, give up and return that the path op failed. TBR=reed@google.com BUG=597926 GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1854333002 Review URL: https://codereview.chromium.org/1854333002
2016-04-05 14:28:48 +00:00
}
fix pathops fuzz bugs Fail out in a couple of new places when the input data is very large and exceeds the limits of the pathops machinery. Most of the change here plumbs in a way to exclude an assert in one of these exceptional cases. The current SkAddIntersection implementation and the inner functions it calls has no way to report an error to the root caller for an early exit, so rather than add that in, exclude the assert when the test that would trigger it runs (allowing the test to otherwise ensure that it properly fails). TBR=reed@google.com BUG=617586,617635 GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2046713003 Review-Url: https://codereview.chromium.org/2046713003
2016-06-08 11:28:19 +00:00
static void fuzz1450_0(skiatest::Reporter* reporter, const char* filename) {
SkPath path;
path.moveTo(SkBits2Float(0x43b40000), SkBits2Float(0xcf000000)); // 360, -2.14748e+09f
path.conicTo(SkBits2Float(0x4e800002), SkBits2Float(0xcf000000), SkBits2Float(0x4e800002), SkBits2Float(0xce7ffffe), SkBits2Float(0x3f3504f4)); // 1.07374e+09f, -2.14748e+09f, 1.07374e+09f, -1.07374e+09f, 0.707107f
path.conicTo(SkBits2Float(0x4e800002), SkBits2Float(0x43800001), SkBits2Float(0x43348000), SkBits2Float(0x43800001), SkBits2Float(0x3f3504f4)); // 1.07374e+09f, 256, 180.5f, 256, 0.707107f
SkPath path1(path);
path.reset();
path.moveTo(SkBits2Float(0x43b40000), SkBits2Float(0x45816000)); // 360, 4140
path.conicTo(SkBits2Float(0x43b40005), SkBits2Float(0x458a945d), SkBits2Float(0x45610000), SkBits2Float(0x458a945d), SkBits2Float(0x3f3504f3)); // 360, 4434.55f, 3600, 4434.55f, 0.707107f
path.conicTo(SkBits2Float(0x45d5bfff), SkBits2Float(0x458a945d), SkBits2Float(0x45d5bfff), SkBits2Float(0x45816000), SkBits2Float(0x3f3504f3)); // 6840, 4434.55f, 6840, 4140, 0.707107f
path.lineTo(SkBits2Float(0x42c80000), SkBits2Float(0x44000000)); // 100, 512
path.lineTo(SkBits2Float(0x42000000), SkBits2Float(0x41800000)); // 32, 16
path.lineTo(SkBits2Float(0x43b40000), SkBits2Float(0x44800000)); // 360, 1024
path.lineTo(SkBits2Float(0x43b40000), SkBits2Float(0x45816000)); // 360, 4140
path.close();
SkPath path2(path);
pathops coincidence and security rewrite Most changes stem from working on an examples bracketed by #if DEBUG_UNDER_DEVELOPMENT // tiger These exposed many problems with coincident curves, as well as errors throughout the code. Fixing these errors also fixed a number of fuzzer-inspired bug reports. * Line/Curve Intersections Check to see if the end of the line nearly intersects the curve. This was a FIXME in the old code. * Performance Use a central chunk allocator. Plumb the allocator into the global variable state so that it can be shared. (Note that 'SkGlobalState' is allocated on the stack and is visible to children functions but not other threads.) * Refactor Let SkOpAngle grow up from a structure to a class. Let SkCoincidentSpans grow up from a structure to a class. Rename enum Alias to AliasMatch. * Coincidence Rewrite Add more debugging to coincidence detection. Parallel debugging routines have read-only logic to report the current coincidence state so that steps through the logic can expose whether things got better or worse. More functions can error-out and cause the pathops engine to non-destructively exit. * Accuracy Remove code that adjusted point locations. Instead, offset the curve part so that sorted curves all use the same origin. Reduce the size (and influence) of magic numbers. * Testing The debug suite with verify and the full release suite ./out/Debug/pathops_unittest -v -V ./out/Release/pathops_unittest -v -V -x expose one error. That error is captured as cubics_d3. This error exists in the checked in code as well. BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 Review-Url: https://codereview.chromium.org/2128633003
2016-07-18 17:01:36 +00:00
testPathOpSkipAssert(reporter, path1, path2, kUnion_SkPathOp, filename);
fix pathops fuzz bugs Fail out in a couple of new places when the input data is very large and exceeds the limits of the pathops machinery. Most of the change here plumbs in a way to exclude an assert in one of these exceptional cases. The current SkAddIntersection implementation and the inner functions it calls has no way to report an error to the root caller for an early exit, so rather than add that in, exclude the assert when the test that would trigger it runs (allowing the test to otherwise ensure that it properly fails). TBR=reed@google.com BUG=617586,617635 GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2046713003 Review-Url: https://codereview.chromium.org/2046713003
2016-06-08 11:28:19 +00:00
}
static void fuzz1450_1(skiatest::Reporter* reporter, const char* filename) {
SkPath path;
path.setFillType(SkPath::kEvenOdd_FillType);
path.moveTo(SkBits2Float(0x4e800002), SkBits2Float(0xce7ffffe)); // 1.07374e+09f, -1.07374e+09f
path.conicTo(SkBits2Float(0x4e800002), SkBits2Float(0xcf000000), SkBits2Float(0x43b40000), SkBits2Float(0xcf000000), SkBits2Float(0x3f3504f4)); // 1.07374e+09f, -2.14748e+09f, 360, -2.14748e+09f, 0.707107f
path.lineTo(SkBits2Float(0x43348000), SkBits2Float(0x43800001)); // 180.5f, 256
path.lineTo(SkBits2Float(0x42000000), SkBits2Float(0x41800000)); // 32, 16
path.lineTo(SkBits2Float(0x42c80000), SkBits2Float(0x44000000)); // 100, 512
path.lineTo(SkBits2Float(0x43553abd), SkBits2Float(0x440f3cbd)); // 213.229f, 572.949f
path.lineTo(SkBits2Float(0x43b40000), SkBits2Float(0x44800000)); // 360, 1024
path.lineTo(SkBits2Float(0x43b40000), SkBits2Float(0x45816000)); // 360, 4140
path.conicTo(SkBits2Float(0x43b40005), SkBits2Float(0x458a945d), SkBits2Float(0x45610000), SkBits2Float(0x458a945d), SkBits2Float(0x3f3504f3)); // 360, 4434.55f, 3600, 4434.55f, 0.707107f
path.conicTo(SkBits2Float(0x45d5bfff), SkBits2Float(0x458a945d), SkBits2Float(0x45d5bfff), SkBits2Float(0x45816000), SkBits2Float(0x3f3504f3)); // 6840, 4434.55f, 6840, 4140, 0.707107f
path.lineTo(SkBits2Float(0x43553abd), SkBits2Float(0x440f3cbd)); // 213.229f, 572.949f
path.lineTo(SkBits2Float(0x43348000), SkBits2Float(0x43800001)); // 180.5f, 256
path.conicTo(SkBits2Float(0x4e800002), SkBits2Float(0x43800001), SkBits2Float(0x4e800002), SkBits2Float(0xce7ffffe), SkBits2Float(0x3f3504f4)); // 1.07374e+09f, 256, 1.07374e+09f, -1.07374e+09f, 0.707107f
path.close();
SkPath path1(path);
path.reset();
path.moveTo(SkBits2Float(0x42fe0000), SkBits2Float(0x43a08000)); // 127, 321
path.lineTo(SkBits2Float(0x45d5c000), SkBits2Float(0x43870000)); // 6840, 270
path.lineTo(SkBits2Float(0xd0a00000), SkBits2Float(0x4cbebc20)); // -2.14748e+10f, 1e+08
path.lineTo(SkBits2Float(0x451f7000), SkBits2Float(0x42800000)); // 2551, 64
path.lineTo(SkBits2Float(0x42fe0000), SkBits2Float(0x43a08000)); // 127, 321
path.close();
SkPath path2(path);
pathops coincidence and security rewrite Most changes stem from working on an examples bracketed by #if DEBUG_UNDER_DEVELOPMENT // tiger These exposed many problems with coincident curves, as well as errors throughout the code. Fixing these errors also fixed a number of fuzzer-inspired bug reports. * Line/Curve Intersections Check to see if the end of the line nearly intersects the curve. This was a FIXME in the old code. * Performance Use a central chunk allocator. Plumb the allocator into the global variable state so that it can be shared. (Note that 'SkGlobalState' is allocated on the stack and is visible to children functions but not other threads.) * Refactor Let SkOpAngle grow up from a structure to a class. Let SkCoincidentSpans grow up from a structure to a class. Rename enum Alias to AliasMatch. * Coincidence Rewrite Add more debugging to coincidence detection. Parallel debugging routines have read-only logic to report the current coincidence state so that steps through the logic can expose whether things got better or worse. More functions can error-out and cause the pathops engine to non-destructively exit. * Accuracy Remove code that adjusted point locations. Instead, offset the curve part so that sorted curves all use the same origin. Reduce the size (and influence) of magic numbers. * Testing The debug suite with verify and the full release suite ./out/Debug/pathops_unittest -v -V ./out/Release/pathops_unittest -v -V -x expose one error. That error is captured as cubics_d3. This error exists in the checked in code as well. BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 Review-Url: https://codereview.chromium.org/2128633003
2016-07-18 17:01:36 +00:00
testPathOpSkipAssert(reporter, path1, path2, kUnion_SkPathOp, filename);
fix pathops fuzz bugs Fail out in a couple of new places when the input data is very large and exceeds the limits of the pathops machinery. Most of the change here plumbs in a way to exclude an assert in one of these exceptional cases. The current SkAddIntersection implementation and the inner functions it calls has no way to report an error to the root caller for an early exit, so rather than add that in, exclude the assert when the test that would trigger it runs (allowing the test to otherwise ensure that it properly fails). TBR=reed@google.com BUG=617586,617635 GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2046713003 Review-Url: https://codereview.chromium.org/2046713003
2016-06-08 11:28:19 +00:00
}
fix fuzz bugs Detect more places where the pathops numerics cause numbers to become nearly identical and subsequently fail. These tests have extreme inputs and cannot succeed. Also remove the expectSuccess parameter from PathOpsDebug and check instead in the test framework. R=mbarbella@chromium.org TBR=reed@google.com BUG=623072,623022 GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2103513002 Review-Url: https://codereview.chromium.org/2103513002
2016-06-28 16:23:57 +00:00
static void fuzz763_9(skiatest::Reporter* reporter, const char* filename) {
SkPath path;
path.setFillType((SkPath::FillType) 1);
SkPath path1(path);
path.reset();
path.setFillType((SkPath::FillType) 0);
path.moveTo(SkBits2Float(0x00000000), SkBits2Float(0x00000000)); // 0, 0
path.conicTo(SkBits2Float(0x2a8c555b), SkBits2Float(0x081f2a21), SkBits2Float(0x7bc00321), SkBits2Float(0xed7a6a4b), SkBits2Float(0x1f212a8c)); // 2.49282e-13f, 4.78968e-34f, 1.99397e+36f, -4.84373e+27f, 3.41283e-20f
path.lineTo(SkBits2Float(0x7bc00321), SkBits2Float(0xed7a6a4b)); // 1.99397e+36f, -4.84373e+27f
path.lineTo(SkBits2Float(0x282a3a21), SkBits2Float(0x3a21df28)); // 9.4495e-15f, 0.000617492f
path.lineTo(SkBits2Float(0x00000000), SkBits2Float(0x00000000)); // 0, 0
path.close();
path.moveTo(SkBits2Float(0x00000000), SkBits2Float(0x00000000)); // 0, 0
path.quadTo(SkBits2Float(0x8a284f9a), SkBits2Float(0x3ac23ab3), SkBits2Float(0x1d2a2928), SkBits2Float(0x63962be6)); // -8.10388e-33f, 0.00148185f, 2.25206e-21f, 5.54035e+21f
path.moveTo(SkBits2Float(0x29272a81), SkBits2Float(0x2ab03a55)); // 3.71183e-14f, 3.13044e-13f
path.quadTo(SkBits2Float(0x2720213b), SkBits2Float(0x3a214729), SkBits2Float(0xdf28282a), SkBits2Float(0x8a2f2121)); // 2.22225e-15f, 0.000615227f, -1.2117e+19f, -8.43217e-33f
path.quadTo(SkBits2Float(0x373b3a27), SkBits2Float(0x201fc4c1), SkBits2Float(0x27576c2a), SkBits2Float(0x5921c25d)); // 1.11596e-05f, 1.35329e-19f, 2.98959e-15f, 2.8457e+15f
path.quadTo(SkBits2Float(0x2720213b), SkBits2Float(0x3a214729), SkBits2Float(0xdf28282a), SkBits2Float(0x3a8a3a21)); // 2.22225e-15f, 0.000615227f, -1.2117e+19f, 0.00105459f
path.cubicTo(SkBits2Float(0x373b3ac5), SkBits2Float(0x201fc422), SkBits2Float(0x523a702a), SkBits2Float(0x27576c51), SkBits2Float(0x5921c25d), SkBits2Float(0x51523a70)); // 1.11598e-05f, 1.35327e-19f, 2.00186e+11f, 2.9896e-15f, 2.8457e+15f, 5.64327e+10f
path.quadTo(SkBits2Float(0xd912102a), SkBits2Float(0x284f9a28), SkBits2Float(0xb38a1f30), SkBits2Float(0x3a3ac23a)); // -2.56957e+15f, 1.15242e-14f, -6.4318e-08f, 0.000712428f
path.lineTo(SkBits2Float(0xc809272a), SkBits2Float(0x29b02829)); // -140445, 7.82294e-14f
SkPath path2(path);
pathops coincidence and security rewrite Most changes stem from working on an examples bracketed by #if DEBUG_UNDER_DEVELOPMENT // tiger These exposed many problems with coincident curves, as well as errors throughout the code. Fixing these errors also fixed a number of fuzzer-inspired bug reports. * Line/Curve Intersections Check to see if the end of the line nearly intersects the curve. This was a FIXME in the old code. * Performance Use a central chunk allocator. Plumb the allocator into the global variable state so that it can be shared. (Note that 'SkGlobalState' is allocated on the stack and is visible to children functions but not other threads.) * Refactor Let SkOpAngle grow up from a structure to a class. Let SkCoincidentSpans grow up from a structure to a class. Rename enum Alias to AliasMatch. * Coincidence Rewrite Add more debugging to coincidence detection. Parallel debugging routines have read-only logic to report the current coincidence state so that steps through the logic can expose whether things got better or worse. More functions can error-out and cause the pathops engine to non-destructively exit. * Accuracy Remove code that adjusted point locations. Instead, offset the curve part so that sorted curves all use the same origin. Reduce the size (and influence) of magic numbers. * Testing The debug suite with verify and the full release suite ./out/Debug/pathops_unittest -v -V ./out/Release/pathops_unittest -v -V -x expose one error. That error is captured as cubics_d3. This error exists in the checked in code as well. BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 Review-Url: https://codereview.chromium.org/2128633003
2016-07-18 17:01:36 +00:00
testPathOp(reporter, path1, path2, (SkPathOp) 1, filename);
fix fuzz bugs Detect more places where the pathops numerics cause numbers to become nearly identical and subsequently fail. These tests have extreme inputs and cannot succeed. Also remove the expectSuccess parameter from PathOpsDebug and check instead in the test framework. R=mbarbella@chromium.org TBR=reed@google.com BUG=623072,623022 GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2103513002 Review-Url: https://codereview.chromium.org/2103513002
2016-06-28 16:23:57 +00:00
}
static void fuzz763_4(skiatest::Reporter* reporter, const char* filename) {
SkPath path;
path.setFillType((SkPath::FillType) 1);
SkPath path1(path);
path.reset();
path.setFillType((SkPath::FillType) 0);
path.moveTo(SkBits2Float(0x00000000), SkBits2Float(0x00000000)); // 0, 0
path.lineTo(SkBits2Float(0x555b3a2d), SkBits2Float(0x2a212a8c)); // 1.50652e+13f, 1.43144e-13f
path.conicTo(SkBits2Float(0xc0032108), SkBits2Float(0x7a6a4b7b), SkBits2Float(0x212a8ced), SkBits2Float(0x0321081f), SkBits2Float(0x6a3a7bc0)); // -2.04889f, 3.04132e+35f, 5.77848e-19f, 4.7323e-37f, 5.63611e+25f
path.conicTo(SkBits2Float(0x3a2147ed), SkBits2Float(0xdf28282a), SkBits2Float(0x3a8a3a21), SkBits2Float(0x8a284f9a), SkBits2Float(0x3ac2b33a)); // 0.000615238f, -1.2117e+19f, 0.00105459f, -8.10388e-33f, 0.00148544f
path.cubicTo(SkBits2Float(0x1d2a2928), SkBits2Float(0x63962be6), SkBits2Float(0x295b2d2a), SkBits2Float(0x68295b2d), SkBits2Float(0x2d296855), SkBits2Float(0x2a8c275b)); // 2.25206e-21f, 5.54035e+21f, 4.86669e-14f, 3.19905e+24f, 9.6297e-12f, 2.48963e-13f
path.lineTo(SkBits2Float(0x00000000), SkBits2Float(0x00000000)); // 0, 0
path.close();
path.moveTo(SkBits2Float(0x55685b1f), SkBits2Float(0x5b2d2968)); // 1.59674e+13f, 4.87407e+16f
path.lineTo(SkBits2Float(0x2a212a8c), SkBits2Float(0x2a21081f)); // 1.43144e-13f, 1.43025e-13f
path.conicTo(SkBits2Float(0xde6a4b7b), SkBits2Float(0x2a8ced7a), SkBits2Float(0x21081f21), SkBits2Float(0x3a7bc003), SkBits2Float(0x47ed7a6a)); // -4.22068e+18f, 2.50338e-13f, 4.61198e-19f, 0.00096035f, 121589
path.lineTo(SkBits2Float(0x55685b1f), SkBits2Float(0x5b2d2968)); // 1.59674e+13f, 4.87407e+16f
path.close();
path.moveTo(SkBits2Float(0x55685b1f), SkBits2Float(0x5b2d2968)); // 1.59674e+13f, 4.87407e+16f
path.quadTo(SkBits2Float(0xdf28282a), SkBits2Float(0x3a8a3a21), SkBits2Float(0x8a284f9a), SkBits2Float(0x3ac23ab3)); // -1.2117e+19f, 0.00105459f, -8.10388e-33f, 0.00148185f
path.lineTo(SkBits2Float(0x2928088c), SkBits2Float(0x2be61d2a)); // 3.73109e-14f, 1.63506e-12f
path.conicTo(SkBits2Float(0x2a812a63), SkBits2Float(0x2d292a27), SkBits2Float(0x5568295b), SkBits2Float(0x5b2d2968), SkBits2Float(0x552d6829)); // 2.29444e-13f, 9.6159e-12f, 1.5954e+13f, 4.87407e+16f, 1.19164e+13f
path.conicTo(SkBits2Float(0x395b2d5b), SkBits2Float(0x68552768), SkBits2Float(0x555b2df0), SkBits2Float(0x1f722a8c), SkBits2Float(0x082a212a)); // 0.000209024f, 4.02636e+24f, 1.50619e+13f, 5.12807e-20f, 5.11965e-34f
path.lineTo(SkBits2Float(0x55685b1f), SkBits2Float(0x5b2d2968)); // 1.59674e+13f, 4.87407e+16f
path.close();
path.moveTo(SkBits2Float(0x212a8c55), SkBits2Float(0x21081f2a)); // 5.7784e-19f, 4.61198e-19f
path.conicTo(SkBits2Float(0x6a4b7bc0), SkBits2Float(0x2147ed7a), SkBits2Float(0x28282a3a), SkBits2Float(0x21df212a), SkBits2Float(0x033a8a3a)); // 6.14991e+25f, 6.77381e-19f, 9.33503e-15f, 1.51198e-18f, 5.48192e-37f
SkPath path2(path);
pathops coincidence and security rewrite Most changes stem from working on an examples bracketed by #if DEBUG_UNDER_DEVELOPMENT // tiger These exposed many problems with coincident curves, as well as errors throughout the code. Fixing these errors also fixed a number of fuzzer-inspired bug reports. * Line/Curve Intersections Check to see if the end of the line nearly intersects the curve. This was a FIXME in the old code. * Performance Use a central chunk allocator. Plumb the allocator into the global variable state so that it can be shared. (Note that 'SkGlobalState' is allocated on the stack and is visible to children functions but not other threads.) * Refactor Let SkOpAngle grow up from a structure to a class. Let SkCoincidentSpans grow up from a structure to a class. Rename enum Alias to AliasMatch. * Coincidence Rewrite Add more debugging to coincidence detection. Parallel debugging routines have read-only logic to report the current coincidence state so that steps through the logic can expose whether things got better or worse. More functions can error-out and cause the pathops engine to non-destructively exit. * Accuracy Remove code that adjusted point locations. Instead, offset the curve part so that sorted curves all use the same origin. Reduce the size (and influence) of magic numbers. * Testing The debug suite with verify and the full release suite ./out/Debug/pathops_unittest -v -V ./out/Release/pathops_unittest -v -V -x expose one error. That error is captured as cubics_d3. This error exists in the checked in code as well. BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 Review-Url: https://codereview.chromium.org/2128633003
2016-07-18 17:01:36 +00:00
testPathOpSkipAssert(reporter, path1, path2, (SkPathOp) 1, filename);
fix fuzz bugs Detect more places where the pathops numerics cause numbers to become nearly identical and subsequently fail. These tests have extreme inputs and cannot succeed. Also remove the expectSuccess parameter from PathOpsDebug and check instead in the test framework. R=mbarbella@chromium.org TBR=reed@google.com BUG=623072,623022 GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2103513002 Review-Url: https://codereview.chromium.org/2103513002
2016-06-28 16:23:57 +00:00
}
fix fuzz test that exceeds numeric limit The extreme values here exceed an internal test that expects computed numbers to be less than MAX_FLT. Use MAX_DBL instead. R=mmoroz@chromium.org TBR=reed@google.com BUG=624346 GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2103903004 Review-Url: https://codereview.chromium.org/2103903004
2016-06-29 15:18:38 +00:00
static void fuzz763_3(skiatest::Reporter* reporter, const char* filename) {
SkPath path;
path.setFillType((SkPath::FillType) 1);
SkPath path1(path);
path.reset();
path.setFillType((SkPath::FillType) 0);
path.moveTo(SkBits2Float(0x00000000), SkBits2Float(0x00000000)); // 0, 0
path.lineTo(SkBits2Float(0x555b292d), SkBits2Float(0x2a212a8c)); // 1.50606e+13f, 1.43144e-13f
path.conicTo(SkBits2Float(0xc0032108), SkBits2Float(0x7a6a4b7b), SkBits2Float(0x212a8ced), SkBits2Float(0x295b2d1f), SkBits2Float(0x29685568)); // -2.04889f, 3.04132e+35f, 5.77848e-19f, 4.86669e-14f, 5.15884e-14f
path.conicTo(SkBits2Float(0x8c28295b), SkBits2Float(0x1f21212a), SkBits2Float(0xc0032a08), SkBits2Float(0x7a6a4b7b), SkBits2Float(0x212a8ced)); // -1.29547e-31f, 3.41205e-20f, -2.04944f, 3.04132e+35f, 5.77848e-19f
path.moveTo(SkBits2Float(0x25682929), SkBits2Float(0x212a8c5b)); // 2.01367e-16f, 5.7784e-19f
path.moveTo(SkBits2Float(0x0321081f), SkBits2Float(0x6a4a7bc0)); // 4.7323e-37f, 6.11969e+25f
path.conicTo(SkBits2Float(0x032108ed), SkBits2Float(0x283a7bc0), SkBits2Float(0x47ed7a6a), SkBits2Float(0x282a3a21), SkBits2Float(0x3a21ff28)); // 4.73239e-37f, 1.03519e-14f, 121589, 9.4495e-15f, 0.000617968f
path.quadTo(SkBits2Float(0x8a284f9a), SkBits2Float(0x3ac23ab3), SkBits2Float(0x2a292827), SkBits2Float(0x962be61d)); // -8.10388e-33f, 0.00148185f, 1.50241e-13f, -1.38859e-25f
path.lineTo(SkBits2Float(0x295b2d2a), SkBits2Float(0x2d296868)); // 4.86669e-14f, 9.62972e-12f
path.moveTo(SkBits2Float(0x212a8c55), SkBits2Float(0x21081f2a)); // 5.7784e-19f, 4.61198e-19f
path.conicTo(SkBits2Float(0x6a4b7bc0), SkBits2Float(0x898ced7a), SkBits2Float(0x21081f21), SkBits2Float(0x3a7bc003), SkBits2Float(0x47ed7a6a)); // 6.14991e+25f, -3.39271e-33f, 4.61198e-19f, 0.00096035f, 121589
path.lineTo(SkBits2Float(0x212a8c55), SkBits2Float(0x21081f2a)); // 5.7784e-19f, 4.61198e-19f
path.close();
path.moveTo(SkBits2Float(0x212a8c55), SkBits2Float(0x21081f2a)); // 5.7784e-19f, 4.61198e-19f
path.quadTo(SkBits2Float(0xdf28282a), SkBits2Float(0x3a8a3a21), SkBits2Float(0xb38a281a), SkBits2Float(0x29283ac2)); // -1.2117e+19f, 0.00105459f, -6.43342e-08f, 3.73545e-14f
path.moveTo(SkBits2Float(0x962be61d), SkBits2Float(0x432a2927)); // -1.38859e-25f, 170.161f
path.conicTo(SkBits2Float(0x3a2a552a), SkBits2Float(0x3b1e2ab0), SkBits2Float(0x29272021), SkBits2Float(0x3b3ac527), SkBits2Float(0x1fc42236)); // 0.000649768f, 0.00241343f, 3.71093e-14f, 0.00284989f, 8.30658e-20f
path.cubicTo(SkBits2Float(0x27576c2a), SkBits2Float(0x5921c25d), SkBits2Float(0x51503a70), SkBits2Float(0x12102a10), SkBits2Float(0x633a28d9), SkBits2Float(0x29c80927)); // 2.98959e-15f, 2.8457e+15f, 5.58959e+10f, 4.54902e-28f, 3.43404e+21f, 8.88337e-14f
path.lineTo(SkBits2Float(0x272927b0), SkBits2Float(0x5b392929)); // 2.3475e-15f, 5.21181e+16f
path.moveTo(SkBits2Float(0x3a1127b4), SkBits2Float(0x2921ee3b)); // 0.000553723f, 3.59558e-14f
path.cubicTo(SkBits2Float(0x5e215d3b), SkBits2Float(0x7828ee3a), SkBits2Float(0x8e28b03b), SkBits2Float(0x50783be8), SkBits2Float(0x9e0b8a3a), SkBits2Float(0x555b2d68)); // 2.90688e+18f, 1.37053e+34f, -2.07925e-30f, 1.66587e+10f, -7.38718e-21f, 1.50618e+13f
path.moveTo(SkBits2Float(0x21081f3f), SkBits2Float(0x9fd4e62a)); // 4.61199e-19f, -9.01663e-20f
path.cubicTo(SkBits2Float(0x3a293a2a), SkBits2Float(0x0e3bf0c5), SkBits2Float(0x3b29d42a), SkBits2Float(0x0f217265), SkBits2Float(0x2d5d2921), SkBits2Float(0x5568295b)); // 0.000645551f, 2.31655e-30f, 0.00259138f, 7.95994e-30f, 1.25715e-11f, 1.5954e+13f
SkPath path2(path);
pathops coincidence and security rewrite Most changes stem from working on an examples bracketed by #if DEBUG_UNDER_DEVELOPMENT // tiger These exposed many problems with coincident curves, as well as errors throughout the code. Fixing these errors also fixed a number of fuzzer-inspired bug reports. * Line/Curve Intersections Check to see if the end of the line nearly intersects the curve. This was a FIXME in the old code. * Performance Use a central chunk allocator. Plumb the allocator into the global variable state so that it can be shared. (Note that 'SkGlobalState' is allocated on the stack and is visible to children functions but not other threads.) * Refactor Let SkOpAngle grow up from a structure to a class. Let SkCoincidentSpans grow up from a structure to a class. Rename enum Alias to AliasMatch. * Coincidence Rewrite Add more debugging to coincidence detection. Parallel debugging routines have read-only logic to report the current coincidence state so that steps through the logic can expose whether things got better or worse. More functions can error-out and cause the pathops engine to non-destructively exit. * Accuracy Remove code that adjusted point locations. Instead, offset the curve part so that sorted curves all use the same origin. Reduce the size (and influence) of magic numbers. * Testing The debug suite with verify and the full release suite ./out/Debug/pathops_unittest -v -V ./out/Release/pathops_unittest -v -V -x expose one error. That error is captured as cubics_d3. This error exists in the checked in code as well. BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 Review-Url: https://codereview.chromium.org/2128633003
2016-07-18 17:01:36 +00:00
testPathOpSkipAssert(reporter, path1, path2, (SkPathOp) 1, filename);
fix fuzz test that exceeds numeric limit The extreme values here exceed an internal test that expects computed numbers to be less than MAX_FLT. Use MAX_DBL instead. R=mmoroz@chromium.org TBR=reed@google.com BUG=624346 GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2103903004 Review-Url: https://codereview.chromium.org/2103903004
2016-06-29 15:18:38 +00:00
}
fix another pathops fuzz bug Tunnel out the error case when the numerics create uncalculable angles. R=mmoroz@chromium.org TBR=reed@google.com BUG=624351 GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2103863005 Review-Url: https://codereview.chromium.org/2103863005
2016-06-29 17:46:08 +00:00
static void fuzz763_5(skiatest::Reporter* reporter, const char* filename) {
SkPath path;
path.setFillType((SkPath::FillType) 1);
SkPath path1(path);
path.reset();
path.setFillType((SkPath::FillType) 0);
path.moveTo(SkBits2Float(0x5b292d55), SkBits2Float(0x2a2a8c55)); // 4.76191e+16f, 1.51477e-13f
path.conicTo(SkBits2Float(0xc0032108), SkBits2Float(0x7a6a4b79), SkBits2Float(0x212a8ced), SkBits2Float(0x0321081f), SkBits2Float(0x6a3a7bc0)); // -2.04889f, 3.04132e+35f, 5.77848e-19f, 4.7323e-37f, 5.63611e+25f
path.conicTo(SkBits2Float(0x3a2147ed), SkBits2Float(0xdf28282a), SkBits2Float(0x3a8a3a21), SkBits2Float(0x8a284f9a), SkBits2Float(0x3ac23ab3)); // 0.000615238f, -1.2117e+19f, 0.00105459f, -8.10388e-33f, 0.00148185f
path.cubicTo(SkBits2Float(0xe62a2928), SkBits2Float(0x2a63962b), SkBits2Float(0x68295b2d), SkBits2Float(0x2d296855), SkBits2Float(0x2a8c555b), SkBits2Float(0x001f2a21)); // -2.0089e+23f, 2.02138e-13f, 3.19905e+24f, 9.6297e-12f, 2.49282e-13f, 2.86201e-39f
path.lineTo(SkBits2Float(0x5b292d55), SkBits2Float(0x2a2a8c55)); // 4.76191e+16f, 1.51477e-13f
path.close();
path.moveTo(SkBits2Float(0x5b292d55), SkBits2Float(0x2a2a8c55)); // 4.76191e+16f, 1.51477e-13f
path.conicTo(SkBits2Float(0x6a4b7bc0), SkBits2Float(0x2a8ced7a), SkBits2Float(0x21081f21), SkBits2Float(0x3a7bc003), SkBits2Float(0x47ed7a6a)); // 6.14991e+25f, 2.50338e-13f, 4.61198e-19f, 0.00096035f, 121589
path.lineTo(SkBits2Float(0x5b292d55), SkBits2Float(0x2a2a8c55)); // 4.76191e+16f, 1.51477e-13f
path.close();
path.moveTo(SkBits2Float(0x5b292d55), SkBits2Float(0x2a2a8c55)); // 4.76191e+16f, 1.51477e-13f
path.quadTo(SkBits2Float(0xdf28282a), SkBits2Float(0x3a8a3b21), SkBits2Float(0x28ee4f9a), SkBits2Float(0x68293b78)); // -1.2117e+19f, 0.00105462f, 2.64578e-14f, 3.19671e+24f
path.lineTo(SkBits2Float(0x5b2d2968), SkBits2Float(0x5b2d8c55)); // 4.87407e+16f, 4.88495e+16f
SkPath path2(path);
pathops coincidence and security rewrite Most changes stem from working on an examples bracketed by #if DEBUG_UNDER_DEVELOPMENT // tiger These exposed many problems with coincident curves, as well as errors throughout the code. Fixing these errors also fixed a number of fuzzer-inspired bug reports. * Line/Curve Intersections Check to see if the end of the line nearly intersects the curve. This was a FIXME in the old code. * Performance Use a central chunk allocator. Plumb the allocator into the global variable state so that it can be shared. (Note that 'SkGlobalState' is allocated on the stack and is visible to children functions but not other threads.) * Refactor Let SkOpAngle grow up from a structure to a class. Let SkCoincidentSpans grow up from a structure to a class. Rename enum Alias to AliasMatch. * Coincidence Rewrite Add more debugging to coincidence detection. Parallel debugging routines have read-only logic to report the current coincidence state so that steps through the logic can expose whether things got better or worse. More functions can error-out and cause the pathops engine to non-destructively exit. * Accuracy Remove code that adjusted point locations. Instead, offset the curve part so that sorted curves all use the same origin. Reduce the size (and influence) of magic numbers. * Testing The debug suite with verify and the full release suite ./out/Debug/pathops_unittest -v -V ./out/Release/pathops_unittest -v -V -x expose one error. That error is captured as cubics_d3. This error exists in the checked in code as well. BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 Review-Url: https://codereview.chromium.org/2128633003
2016-07-18 17:01:36 +00:00
testPathOpSkipAssert(reporter, path1, path2, (SkPathOp) 4, filename);
fix another pathops fuzz bug Tunnel out the error case when the numerics create uncalculable angles. R=mmoroz@chromium.org TBR=reed@google.com BUG=624351 GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2103863005 Review-Url: https://codereview.chromium.org/2103863005
2016-06-29 17:46:08 +00:00
}
static void fuzz763_2(skiatest::Reporter* reporter, const char* filename) {
SkPath path;
path.setFillType((SkPath::FillType) 1);
SkPath path1(path);
path.reset();
path.setFillType((SkPath::FillType) 0);
path.moveTo(SkBits2Float(0x00000000), SkBits2Float(0x00000000)); // 0, 0
path.lineTo(SkBits2Float(0x555b292d), SkBits2Float(0x2a212a8c)); // 1.50606e+13f, 1.43144e-13f
path.conicTo(SkBits2Float(0xc0032108), SkBits2Float(0x7a6a4b7b), SkBits2Float(0x212a8ced), SkBits2Float(0x0321081f), SkBits2Float(0x6a3a7bc0)); // -2.04889f, 3.04132e+35f, 5.77848e-19f, 4.7323e-37f, 5.63611e+25f
path.lineTo(SkBits2Float(0x081f2ad7), SkBits2Float(0x7bc00321)); // 4.78977e-34f, 1.99397e+36f
path.moveTo(SkBits2Float(0x2a3a2147), SkBits2Float(0xdf212828)); // 1.65317e-13f, -1.16126e+19f
path.quadTo(SkBits2Float(0x4f1a3a8a), SkBits2Float(0x3ab38a28), SkBits2Float(0x29283ac2), SkBits2Float(0x962be62a)); // 2.58753e+09f, 0.00136978f, 3.73545e-14f, -1.38859e-25f
path.cubicTo(SkBits2Float(0x272a812a), SkBits2Float(0x3a2a5529), SkBits2Float(0x3b1e2ab0), SkBits2Float(0x29272021), SkBits2Float(0x3b3ac527), SkBits2Float(0x1fc42237)); // 2.36623e-15f, 0.000649768f, 0.00241343f, 3.71093e-14f, 0.00284989f, 8.30658e-20f
path.cubicTo(SkBits2Float(0x27576c2a), SkBits2Float(0x5921c25d), SkBits2Float(0x51523a70), SkBits2Float(0x12102a10), SkBits2Float(0x633a28d9), SkBits2Float(0x29c80927)); // 2.98959e-15f, 2.8457e+15f, 5.64327e+10f, 4.54902e-28f, 3.43404e+21f, 8.88337e-14f
path.lineTo(SkBits2Float(0x29292727), SkBits2Float(0x21475b3b)); // 3.75595e-14f, 6.75446e-19f
path.quadTo(SkBits2Float(0xdf28282a), SkBits2Float(0x3a8a3a21), SkBits2Float(0x8a284f9a), SkBits2Float(0x3ac23ab3)); // -1.2117e+19f, 0.00105459f, -8.10388e-33f, 0.00148185f
path.cubicTo(SkBits2Float(0x682d2928), SkBits2Float(0x555b6829), SkBits2Float(0x555b292d), SkBits2Float(0x2a212a8c), SkBits2Float(0x0321081f), SkBits2Float(0x6a4b7bc0)); // 3.27091e+24f, 1.50775e+13f, 1.50606e+13f, 1.43144e-13f, 4.7323e-37f, 6.14991e+25f
path.conicTo(SkBits2Float(0x295b2ded), SkBits2Float(0x29685568), SkBits2Float(0x8c555b2d), SkBits2Float(0xe61d2a2a), SkBits2Float(0x2a63962b)); // 4.86676e-14f, 5.15884e-14f, -1.64364e-31f, -1.85547e+23f, 2.02138e-13f
path.conicTo(SkBits2Float(0x5568295b), SkBits2Float(0x5b2d2968), SkBits2Float(0x212a8c55), SkBits2Float(0x21081f2a), SkBits2Float(0x4b7bc003)); // 1.5954e+13f, 4.87407e+16f, 5.7784e-19f, 4.61198e-19f, 1.64987e+07f
path.lineTo(SkBits2Float(0x2a8ced7a), SkBits2Float(0x21081f21)); // 2.50338e-13f, 4.61198e-19f
path.conicTo(SkBits2Float(0x6a3a7bc0), SkBits2Float(0x2147ed7a), SkBits2Float(0x28282a3a), SkBits2Float(0x8a3a21df), SkBits2Float(0x27b42a3a)); // 5.63611e+25f, 6.77381e-19f, 9.33503e-15f, -8.96194e-33f, 5.00058e-15f
path.conicTo(SkBits2Float(0x2921217d), SkBits2Float(0x5e3a3b35), SkBits2Float(0x7828ee3a), SkBits2Float(0x8e28b03b), SkBits2Float(0x783be82a)); // 3.57782e-14f, 3.35484e+18f, 1.37053e+34f, -2.07925e-30f, 1.52448e+34f
path.conicTo(SkBits2Float(0x8e0b8a3a), SkBits2Float(0x279fd4e6), SkBits2Float(0x7a293a2a), SkBits2Float(0x2a0ef0c5), SkBits2Float(0x653b29d4)); // -1.71996e-30f, 4.43622e-15f, 2.19669e+35f, 1.26957e-13f, 5.52409e+22f
path.quadTo(SkBits2Float(0x29210f21), SkBits2Float(0x282a085d), SkBits2Float(0xc2ab2127), SkBits2Float(0xa6800028)); // 3.57623e-14f, 9.43871e-15f, -85.5648f, -8.88183e-16f
path.lineTo(SkBits2Float(0x2a3a2147), SkBits2Float(0xdf212828)); // 1.65317e-13f, -1.16126e+19f
path.close();
path.moveTo(SkBits2Float(0x2a3a2147), SkBits2Float(0xdf212828)); // 1.65317e-13f, -1.16126e+19f
path.quadTo(SkBits2Float(0x216a2770), SkBits2Float(0x2ab73b28), SkBits2Float(0x4b28f427), SkBits2Float(0x283b5b28)); // 7.93345e-19f, 3.25484e-13f, 1.10726e+07f, 1.04004e-14f
path.lineTo(SkBits2Float(0x2a3a2147), SkBits2Float(0xdf212828)); // 1.65317e-13f, -1.16126e+19f
path.close();
path.moveTo(SkBits2Float(0x2a3a2147), SkBits2Float(0xdf212828)); // 1.65317e-13f, -1.16126e+19f
path.conicTo(SkBits2Float(0xf86d273b), SkBits2Float(0x27e523e3), SkBits2Float(0x2927e0f5), SkBits2Float(0x2ac0e729), SkBits2Float(0x6b492128)); // -1.92402e+34f, 6.35992e-15f, 3.72766e-14f, 3.42665e-13f, 2.43151e+26f
path.cubicTo(SkBits2Float(0x2f273927), SkBits2Float(0xa83a2c21), SkBits2Float(0xd7122121), SkBits2Float(0x21212921), SkBits2Float(0x3be3db3a), SkBits2Float(0xa9deb63b)); // 1.52089e-10f, -1.03346e-14f, -1.60671e+14f, 5.46034e-19f, 0.00695362f, -9.89039e-14f
SkPath path2(path);
pathops coincidence and security rewrite Most changes stem from working on an examples bracketed by #if DEBUG_UNDER_DEVELOPMENT // tiger These exposed many problems with coincident curves, as well as errors throughout the code. Fixing these errors also fixed a number of fuzzer-inspired bug reports. * Line/Curve Intersections Check to see if the end of the line nearly intersects the curve. This was a FIXME in the old code. * Performance Use a central chunk allocator. Plumb the allocator into the global variable state so that it can be shared. (Note that 'SkGlobalState' is allocated on the stack and is visible to children functions but not other threads.) * Refactor Let SkOpAngle grow up from a structure to a class. Let SkCoincidentSpans grow up from a structure to a class. Rename enum Alias to AliasMatch. * Coincidence Rewrite Add more debugging to coincidence detection. Parallel debugging routines have read-only logic to report the current coincidence state so that steps through the logic can expose whether things got better or worse. More functions can error-out and cause the pathops engine to non-destructively exit. * Accuracy Remove code that adjusted point locations. Instead, offset the curve part so that sorted curves all use the same origin. Reduce the size (and influence) of magic numbers. * Testing The debug suite with verify and the full release suite ./out/Debug/pathops_unittest -v -V ./out/Release/pathops_unittest -v -V -x expose one error. That error is captured as cubics_d3. This error exists in the checked in code as well. BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 Review-Url: https://codereview.chromium.org/2128633003
2016-07-18 17:01:36 +00:00
testPathOpSkipAssert(reporter, path1, path2, (SkPathOp) 1, filename);
}
// crbug.com/626164
static void fuzz763_1c(skiatest::Reporter* reporter, const char* filename) {
SkPath path;
path.setFillType((SkPath::FillType) 0);
SkPath path1(path);
path.reset();
path.setFillType((SkPath::FillType) 0);
path.moveTo(SkBits2Float(0x00000000), SkBits2Float(0x00000000)); // 0, 0
path.cubicTo(SkBits2Float(0x1931204a), SkBits2Float(0x2ba1a14a), SkBits2Float(0x4a4a08ff), SkBits2Float(0x4a4a08ff), SkBits2Float(0x4a4a4a34), SkBits2Float(0x4a4a4a4a)); // 9.15721e-24f, 1.14845e-12f, 3.31014e+06f, 3.31014e+06f, 3.31432e+06f, 3.31432e+06f
path.moveTo(SkBits2Float(0x000010a1), SkBits2Float(0x19312000)); // 5.96533e-42f, 9.15715e-24f
path.cubicTo(SkBits2Float(0x4a4a4a4a), SkBits2Float(0x4a4a4a4a), SkBits2Float(0xa14a4a4a), SkBits2Float(0x08ff2ba1), SkBits2Float(0x08ff4a4a), SkBits2Float(0x4a344a4a)); // 3.31432e+06f, 3.31432e+06f, -6.85386e-19f, 1.53575e-33f, 1.53647e-33f, 2.95387e+06f
path.cubicTo(SkBits2Float(0x4a4a4a4a), SkBits2Float(0x4a4a4a4a), SkBits2Float(0x2ba1a14a), SkBits2Float(0x4e4a08ff), SkBits2Float(0x4a4a4a4a), SkBits2Float(0xa1a181ff)); // 3.31432e+06f, 3.31432e+06f, 1.14845e-12f, 8.47397e+08f, 3.31432e+06f, -1.09442e-18f
SkPath path2(path);
SkPath dummy;
testPathOp(reporter, path1, path2, (SkPathOp)4, filename);
}
// crbug.com/626186
static void fuzz763_1b(skiatest::Reporter* reporter, const char* filename) {
SkPath path;
path.setFillType((SkPath::FillType) 0);
path.moveTo(SkBits2Float(0x00000000), SkBits2Float(0x00000000)); // 0, 0
path.cubicTo(SkBits2Float(0x0000ff07), SkBits2Float(0xf9f9ff00), SkBits2Float(0xfe0ef9f4), SkBits2Float(0xd9b105fb), SkBits2Float(0x000000f9), SkBits2Float(0xfe11f901)); // 9.14866e-41f, -1.62257e+35f, -4.75121e+37f, -6.22846e+15f, 3.48923e-43f, -4.85077e+37f
path.lineTo(SkBits2Float(0xda1905ed), SkBits2Float(0x3c05fbfb)); // -1.0768e+16f, 0.00817775f
path.cubicTo(SkBits2Float(0x3c3c3c3c), SkBits2Float(0x3c3c3c3c), SkBits2Float(0x253c7f00), SkBits2Float(0xfa00d3fa), SkBits2Float(0x250025fe), SkBits2Float(0x00000006)); // 0.011489f, 0.011489f, 1.63494e-16f, -1.67228e+35f, 1.11151e-16f, 8.40779e-45f
SkPath path1(path);
path.reset();
path.setFillType((SkPath::FillType) 0);
path.moveTo(SkBits2Float(0x00000000), SkBits2Float(0x00000000)); // 0, 0
path.quadTo(SkBits2Float(0x3c3c3c3c), SkBits2Float(0xfa253c3c), SkBits2Float(0xfefa00d3), SkBits2Float(0x25fad9df)); // 0.011489f, -2.14488e+35f, -1.66156e+38f, 4.35157e-16f
path.lineTo(SkBits2Float(0x00000000), SkBits2Float(0x00000000)); // 0, 0
path.close();
path.moveTo(SkBits2Float(0x00000000), SkBits2Float(0x00000000)); // 0, 0
path.lineTo(SkBits2Float(0x8dfefa00), SkBits2Float(0xf0f9fad9)); // -1.57141e-30f, -6.1892e+29f
path.cubicTo(SkBits2Float(0x20fe58f9), SkBits2Float(0x0525fbed), SkBits2Float(0x1905ffff), SkBits2Float(0x01f9f9f9), SkBits2Float(0xfbfe0ef9), SkBits2Float(0xfb212fff)); // 4.30882e-19f, 7.80453e-36f, 6.92764e-24f, 9.18268e-38f, -2.63829e+36f, -8.36933e+35f
SkPath path2(path);
testPathOp(reporter, path1, path2, (SkPathOp)2, filename);
}
static void fuzz763_1a(skiatest::Reporter* reporter, const char* filename) {
disable test that fails to compute bounds on nexus 5 TBR=jcgregario@google.com GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2160713003 Review-Url: https://codereview.chromium.org/2160713003
2016-07-18 19:13:23 +00:00
// FIXME: fails on Nexus5 by generating a path with points that are not inside the path bounds
#if DEBUG_UNDER_DEVELOPMENT
return;
#endif
pathops coincidence and security rewrite Most changes stem from working on an examples bracketed by #if DEBUG_UNDER_DEVELOPMENT // tiger These exposed many problems with coincident curves, as well as errors throughout the code. Fixing these errors also fixed a number of fuzzer-inspired bug reports. * Line/Curve Intersections Check to see if the end of the line nearly intersects the curve. This was a FIXME in the old code. * Performance Use a central chunk allocator. Plumb the allocator into the global variable state so that it can be shared. (Note that 'SkGlobalState' is allocated on the stack and is visible to children functions but not other threads.) * Refactor Let SkOpAngle grow up from a structure to a class. Let SkCoincidentSpans grow up from a structure to a class. Rename enum Alias to AliasMatch. * Coincidence Rewrite Add more debugging to coincidence detection. Parallel debugging routines have read-only logic to report the current coincidence state so that steps through the logic can expose whether things got better or worse. More functions can error-out and cause the pathops engine to non-destructively exit. * Accuracy Remove code that adjusted point locations. Instead, offset the curve part so that sorted curves all use the same origin. Reduce the size (and influence) of magic numbers. * Testing The debug suite with verify and the full release suite ./out/Debug/pathops_unittest -v -V ./out/Release/pathops_unittest -v -V -x expose one error. That error is captured as cubics_d3. This error exists in the checked in code as well. BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 Review-Url: https://codereview.chromium.org/2128633003
2016-07-18 17:01:36 +00:00
SkPath path;
path.setFillType((SkPath::FillType) 0);
path.moveTo(SkBits2Float(0x00000000), SkBits2Float(0x00000000)); // 0, 0
path.cubicTo(SkBits2Float(0x154be880), SkBits2Float(0x80000640), SkBits2Float(0x5559a419), SkBits2Float(0x59d55928), SkBits2Float(0x80045959), SkBits2Float(0x40154be8)); // 4.11789e-26f, -2.24208e-42f, 1.49562e+13f, 7.50652e+15f, -3.99394e-40f, 2.33276f
SkPath path1(path);
path.reset();
path.setFillType((SkPath::FillType) 0);
path.moveTo(SkBits2Float(0x00000000), SkBits2Float(0x00000000)); // 0, 0
path.quadTo(SkBits2Float(0x5559a419), SkBits2Float(0x59d55928), SkBits2Float(0xbd595959), SkBits2Float(0x3f3f3f09)); // 1.49562e+13f, 7.50652e+15f, -0.0530637f, 0.747056f
path.moveTo(SkBits2Float(0x3f3f3f3f), SkBits2Float(0x3f3f3f3f)); // 0.747059f, 0.747059f
path.moveTo(SkBits2Float(0x3f3f3f3f), SkBits2Float(0xff3f3f3f)); // 0.747059f, -2.54211e+38f
path.lineTo(SkBits2Float(0x09090909), SkBits2Float(0x3038d509)); // 1.6495e-33f, 6.72416e-10f
path.conicTo(SkBits2Float(0x5947ffff), SkBits2Float(0x40e88004), SkBits2Float(0x00002059), SkBits2Float(0x28555900), SkBits2Float(0x5959d559)); // 3.51844e+15f, 7.26563f, 1.16042e-41f, 1.18432e-14f, 3.83217e+15f
path.lineTo(SkBits2Float(0x3f3f3f3f), SkBits2Float(0xff3f3f3f)); // 0.747059f, -2.54211e+38f
path.close();
path.moveTo(SkBits2Float(0x3f3f3f3f), SkBits2Float(0xff3f3f3f)); // 0.747059f, -2.54211e+38f
path.lineTo(SkBits2Float(0x38d57f4b), SkBits2Float(0x59597f4b)); // 0.000101803f, 3.82625e+15f
path.lineTo(SkBits2Float(0x3f3f3f3f), SkBits2Float(0xff3f3f3f)); // 0.747059f, -2.54211e+38f
path.close();
path.moveTo(SkBits2Float(0x384700ff), SkBits2Float(0x0108804b)); // 4.74462e-05f, 2.50713e-38f
SkPath path2(path);
testPathOp(reporter, path1, path2, (SkPathOp)0, filename);
}
// crbug.com/627780
static void fuzz763_3a(skiatest::Reporter* reporter, const char* filename) {
SkPath path;
path.setFillType((SkPath::FillType) 1);
SkPath path1(path);
path.reset();
path.setFillType((SkPath::FillType) 0);
path.moveTo(SkBits2Float(0x00000000), SkBits2Float(0x00000000)); // 0, 0
path.lineTo(SkBits2Float(0x555b292d), SkBits2Float(0x2a212a8c)); // 1.50606e+13f, 1.43144e-13f
path.conicTo(SkBits2Float(0xc0032108), SkBits2Float(0x7a6a4b7b), SkBits2Float(0x212a8ced), SkBits2Float(0x0321081f), SkBits2Float(0x6a3a7bc0)); // -2.04889f, 3.04132e+35f, 5.77848e-19f, 4.7323e-37f, 5.63611e+25f
path.conicTo(SkBits2Float(0x3a2147ed), SkBits2Float(0xdf28282a), SkBits2Float(0x3a8a3a21), SkBits2Float(0x8a284f9a), SkBits2Float(0x3ac23ab3)); // 0.000615238f, -1.2117e+19f, 0.00105459f, -8.10388e-33f, 0.00148185f
path.cubicTo(SkBits2Float(0x1d2a2928), SkBits2Float(0x63962be6), SkBits2Float(0x272a812a), SkBits2Float(0x295b2d29), SkBits2Float(0x2a685568), SkBits2Float(0x68295b2d)); // 2.25206e-21f, 5.54035e+21f, 2.36623e-15f, 4.86669e-14f, 2.06354e-13f, 3.19905e+24f
path.conicTo(SkBits2Float(0x2a8c555b), SkBits2Float(0x081f2a21), SkBits2Float(0x7bc00321), SkBits2Float(0x7a6a4b77), SkBits2Float(0x3a214726)); // 2.49282e-13f, 4.78968e-34f, 1.99397e+36f, 3.04132e+35f, 0.000615226f
path.moveTo(SkBits2Float(0x8adf2028), SkBits2Float(0x3a219a3a)); // -2.14862e-32f, 0.000616464f
path.quadTo(SkBits2Float(0x3ab38e28), SkBits2Float(0x29283ac2), SkBits2Float(0x2be61d2a), SkBits2Float(0x812a4396)); // 0.0013699f, 3.73545e-14f, 1.63506e-12f, -3.12726e-38f
SkPath path2(path);
testPathOpSkipAssert(reporter, path1, path2, (SkPathOp) 1, filename);
}
// crbug.com/627689
static void fuzz763_5a(skiatest::Reporter* reporter, const char* filename) {
SkPath path;
path.setFillType((SkPath::FillType) 1);
path.moveTo(SkBits2Float(0x38bd8610), SkBits2Float(0x00000000)); // 9.03719e-05f, 0
path.conicTo(SkBits2Float(0x4183d871), SkBits2Float(0x41fea321), SkBits2Float(0xb700ff00), SkBits2Float(0x4240b8b8), SkBits2Float(0x3b058283)); // 16.4807f, 31.8297f, -7.68877e-06f, 48.1804f, 0.0020372f
path.lineTo(SkBits2Float(0x3a3a3ab8), SkBits2Float(0xb8b8b8b8)); // 0.000710409f, -8.80821e-05f
path.conicTo(SkBits2Float(0x3a455ec8), SkBits2Float(0xb8b8b8b3), SkBits2Float(0x38b2418d), SkBits2Float(0xb730d014), SkBits2Float(0x3f7ffff3)); // 0.000752908f, -8.80821e-05f, 8.49991e-05f, -1.05389e-05f, 0.999999f
path.quadTo(SkBits2Float(0x3a51246a), SkBits2Float(0xb6da45a3), SkBits2Float(0x38bc5c3c), SkBits2Float(0x00000000)); // 0.000797814f, -6.50501e-06f, 8.98172e-05f, 0
path.lineTo(SkBits2Float(0x3a3a3ab8), SkBits2Float(0xb8b8b8b8)); // 0.000710409f, -8.80821e-05f
path.quadTo(SkBits2Float(0x39a32d2d), SkBits2Float(0x00000000), SkBits2Float(0xb8a13a00), SkBits2Float(0x00000000)); // 0.000311234f, 0, -7.68788e-05f, 0
path.lineTo(SkBits2Float(0x3a3a3ab8), SkBits2Float(0xb8b8b8b8)); // 0.000710409f, -8.80821e-05f
path.quadTo(SkBits2Float(0x39ba814c), SkBits2Float(0xb838fed2), SkBits2Float(0x00000000), SkBits2Float(0x00000000)); // 0.00035573f, -4.41063e-05f, 0, 0
path.lineTo(SkBits2Float(0x38bd8610), SkBits2Float(0x00000000)); // 9.03719e-05f, 0
path.close();
SkPath path1(path);
path.reset();
path.setFillType((SkPath::FillType) 0);
SkPath path2(path);
testPathOp(reporter, path1, path2, (SkPathOp) 4, filename);
}
// crbug.com/627401
static void fuzz763_2a(skiatest::Reporter* reporter, const char* filename) {
SkPath path;
path.setFillType((SkPath::FillType) 1);
SkPath path1(path);
path.reset();
path.setFillType((SkPath::FillType) 0);
path.moveTo(SkBits2Float(0x00000000), SkBits2Float(0x00000000)); // 0, 0
path.quadTo(SkBits2Float(0x3e484500), SkBits2Float(0x164f3a30), SkBits2Float(0x49484801), SkBits2Float(0x7d0100c8)); // 0.195576f, 1.67397e-25f, 820352, 1.07172e+37f
path.conicTo(SkBits2Float(0xff7f36fd), SkBits2Float(0x3e647d01), SkBits2Float(0x0c00f430), SkBits2Float(0x486b6448), SkBits2Float(0x00484848)); // -3.39239e+38f, 0.223133f, 9.93424e-32f, 241041, 6.63809e-39f
path.lineTo(SkBits2Float(0x4f4f557d), SkBits2Float(0x48480112)); // 3.47849e+09f, 204804
path.lineTo(SkBits2Float(0xf40c01ff), SkBits2Float(0x45008000)); // -4.43702e+31f, 2056
path.moveTo(SkBits2Float(0x4bfffa00), SkBits2Float(0x7d4ac859)); // 3.35514e+07f, 1.68465e+37f
path.conicTo(SkBits2Float(0x7d014f3e), SkBits2Float(0x00f4ff01), SkBits2Float(0x6b64480c), SkBits2Float(0x48484848), SkBits2Float(0x557d0100)); // 1.07426e+37f, 2.24993e-38f, 2.75975e+26f, 205089, 1.73863e+13f
SkPath path2(path);
testPathOpSkipAssert(reporter, path1, path2, (SkPathOp) 0, filename);
}
// crbug.com/627761
static void fuzz763_2b(skiatest::Reporter* reporter, const char* filename) {
SkPath path;
path.setFillType((SkPath::FillType) 1);
SkPath path1(path);
path.reset();
path.setFillType((SkPath::FillType) 0);
path.moveTo(SkBits2Float(0x5b292d55), SkBits2Float(0x212a8c55)); // 4.76191e+16f, 5.7784e-19f
path.moveTo(SkBits2Float(0x3b21081f), SkBits2Float(0x4b7bc003)); // 0.00245715f, 1.64987e+07f
path.lineTo(SkBits2Float(0x2a8ced7a), SkBits2Float(0x21081f21)); // 2.50338e-13f, 4.61198e-19f
path.conicTo(SkBits2Float(0x6a3a7bc0), SkBits2Float(0x4721ed7a), SkBits2Float(0x282a3a21), SkBits2Float(0x3a21df28), SkBits2Float(0x4f9a3a8a)); // 5.63611e+25f, 41453.5f, 9.4495e-15f, 0.000617492f, 5.17506e+09f
path.lineTo(SkBits2Float(0x3b21081f), SkBits2Float(0x4b7bc003)); // 0.00245715f, 1.64987e+07f
path.close();
path.moveTo(SkBits2Float(0x3b21081f), SkBits2Float(0x4b7bc003)); // 0.00245715f, 1.64987e+07f
path.cubicTo(SkBits2Float(0x273ac23a), SkBits2Float(0x1d2a2928), SkBits2Float(0x63962be6), SkBits2Float(0x272a812a), SkBits2Float(0x295b2d29), SkBits2Float(0x29685568)); // 2.5918e-15f, 2.25206e-21f, 5.54035e+21f, 2.36623e-15f, 4.86669e-14f, 5.15884e-14f
path.lineTo(SkBits2Float(0x081f2a21), SkBits2Float(0x7bc00321)); // 4.78968e-34f, 1.99397e+36f
path.lineTo(SkBits2Float(0x282a3a21), SkBits2Float(0x3a21df28)); // 9.4495e-15f, 0.000617492f
path.lineTo(SkBits2Float(0x3b21081f), SkBits2Float(0x4b7bc003)); // 0.00245715f, 1.64987e+07f
path.close();
path.moveTo(SkBits2Float(0x3b21081f), SkBits2Float(0x4b7bc003)); // 0.00245715f, 1.64987e+07f
path.quadTo(SkBits2Float(0x8a4fc29a), SkBits2Float(0x3ab3283a), SkBits2Float(0x1d2a2928), SkBits2Float(0x43962be6)); // -1.00033e-32f, 0.00136686f, 2.25206e-21f, 300.343f
path.moveTo(SkBits2Float(0x5b2d2a81), SkBits2Float(0x29276829)); // 4.87419e+16f, 3.71718e-14f
path.conicTo(SkBits2Float(0x1e2ab03a), SkBits2Float(0x2920213b), SkBits2Float(0x3b3ac527), SkBits2Float(0xc422333b), SkBits2Float(0x6c2a9f1f)); // 9.03617e-21f, 3.5556e-14f, 0.00284989f, -648.8f, 8.25075e+26f
path.quadTo(SkBits2Float(0xc25d2757), SkBits2Float(0x3a705921), SkBits2Float(0x2a105152), SkBits2Float(0x28d91210)); // -55.2884f, 0.000916855f, 1.2818e-13f, 2.40997e-14f
path.quadTo(SkBits2Float(0x68295b2d), SkBits2Float(0x2d296855), SkBits2Float(0x2a8c555b), SkBits2Float(0x081f2a21)); // 3.19905e+24f, 9.6297e-12f, 2.49282e-13f, 4.78968e-34f
path.lineTo(SkBits2Float(0x5b2d2a81), SkBits2Float(0x29276829)); // 4.87419e+16f, 3.71718e-14f
path.close();
path.moveTo(SkBits2Float(0x5b2d2a81), SkBits2Float(0x29276829)); // 4.87419e+16f, 3.71718e-14f
path.conicTo(SkBits2Float(0x6a4b7bc0), SkBits2Float(0x2a8ced7a), SkBits2Float(0x21081f21), SkBits2Float(0xcb7bc003), SkBits2Float(0x47ed7a6a)); // 6.14991e+25f, 2.50338e-13f, 4.61198e-19f, -1.64987e+07f, 121589
path.lineTo(SkBits2Float(0x5b2d2a81), SkBits2Float(0x29276829)); // 4.87419e+16f, 3.71718e-14f
path.close();
path.moveTo(SkBits2Float(0x5b2d2a81), SkBits2Float(0x29276829)); // 4.87419e+16f, 3.71718e-14f
path.quadTo(SkBits2Float(0xdf28282a), SkBits2Float(0x2d8a3a21), SkBits2Float(0x5b682b68), SkBits2Float(0x5b292d55)); // -1.2117e+19f, 1.57146e-11f, 6.53499e+16f, 4.76191e+16f
path.lineTo(SkBits2Float(0x2a212a8c), SkBits2Float(0x0321081f)); // 1.43144e-13f, 4.7323e-37f
path.conicTo(SkBits2Float(0x7a6a4b7b), SkBits2Float(0x212a8ced), SkBits2Float(0x0321081f), SkBits2Float(0x6a3a7bc0), SkBits2Float(0x3a21477a)); // 3.04132e+35f, 5.77848e-19f, 4.7323e-37f, 5.63611e+25f, 0.000615231f
path.moveTo(SkBits2Float(0x21df2828), SkBits2Float(0x9a3a8a3a)); // 1.51217e-18f, -3.85756e-23f
path.quadTo(SkBits2Float(0x3ab38a28), SkBits2Float(0x28273ac2), SkBits2Float(0xe61d2a29), SkBits2Float(0x2a63962b)); // 0.00136978f, 9.2831e-15f, -1.85547e+23f, 2.02138e-13f
path.conicTo(SkBits2Float(0x2d29272a), SkBits2Float(0x5568295b), SkBits2Float(0x5b2d2968), SkBits2Float(0x5b2d6829), SkBits2Float(0x212a8c55)); // 9.61523e-12f, 1.5954e+13f, 4.87407e+16f, 4.88097e+16f, 5.7784e-19f
path.moveTo(SkBits2Float(0x0321081f), SkBits2Float(0x6a4b7bc0)); // 4.7323e-37f, 6.14991e+25f
path.conicTo(SkBits2Float(0x3a2147ed), SkBits2Float(0xdf28282a), SkBits2Float(0x3a8a3a21), SkBits2Float(0x8a284f9a), SkBits2Float(0x3ac23ab3)); // 0.000615238f, -1.2117e+19f, 0.00105459f, -8.10388e-33f, 0.00148185f
path.lineTo(SkBits2Float(0x0321081f), SkBits2Float(0x6a4b7bc0)); // 4.7323e-37f, 6.14991e+25f
path.close();
SkPath path2(path);
testPathOpSkipAssert(reporter, path1, path2, (SkPathOp) 4, filename);
}
static void fuzz763_2c(skiatest::Reporter* reporter, const char* filename) {
SkPath path;
path.setFillType((SkPath::FillType) 1);
path.moveTo(SkBits2Float(0x00000000), SkBits2Float(0x36344a4a)); // 0, 2.68653e-06f
path.cubicTo(SkBits2Float(0x00000000), SkBits2Float(0x00000000), SkBits2Float(0x364a4a4a), SkBits2Float(0x364a4a4a), SkBits2Float(0x00000000), SkBits2Float(0x00000000)); // 0, 0, 3.01436e-06f, 3.01436e-06f, 0, 0
path.lineTo(SkBits2Float(0x364a4a4a), SkBits2Float(0x00000000)); // 3.01436e-06f, 0
path.cubicTo(SkBits2Float(0x364a30f0), SkBits2Float(0x344ac7fb), SkBits2Float(0x3656d432), SkBits2Float(0x34cabb48), SkBits2Float(0x367031a9), SkBits2Float(0x351802f1)); // 3.01288e-06f, 1.88855e-07f, 3.2012e-06f, 3.77617e-07f, 3.57917e-06f, 5.66287e-07f
path.cubicTo(SkBits2Float(0x36a7b150), SkBits2Float(0x35ab09db), SkBits2Float(0x371874ed), SkBits2Float(0x3604f2c7), SkBits2Float(0x3784e0c7), SkBits2Float(0x36344a51)); // 4.99763e-06f, 1.27434e-06f, 9.08713e-06f, 1.98108e-06f, 1.58403e-05f, 2.68653e-06f
path.cubicTo(SkBits2Float(0x3743dc9a), SkBits2Float(0x36344a4f), SkBits2Float(0x36fbef33), SkBits2Float(0x36344a4e), SkBits2Float(0x36604a35), SkBits2Float(0x36344a4c)); // 1.16743e-05f, 2.68653e-06f, 7.50823e-06f, 2.68653e-06f, 3.34218e-06f, 2.68653e-06f
path.cubicTo(SkBits2Float(0x36531715), SkBits2Float(0x36344a4c), SkBits2Float(0x3645e3f5), SkBits2Float(0x36344a4b), SkBits2Float(0x3638b0d4), SkBits2Float(0x36344a4b)); // 3.14549e-06f, 2.68653e-06f, 2.9488e-06f, 2.68653e-06f, 2.75211e-06f, 2.68653e-06f
path.cubicTo(SkBits2Float(0x35f64120), SkBits2Float(0x36344a4b), SkBits2Float(0x35764124), SkBits2Float(0x36344a4a), SkBits2Float(0x00000000), SkBits2Float(0x36344a4a)); // 1.83474e-06f, 2.68653e-06f, 9.17369e-07f, 2.68653e-06f, 0, 2.68653e-06f
path.close();
SkPath path1(path);
path.reset();
path.setFillType((SkPath::FillType) 0);
path.moveTo(SkBits2Float(0x00000000), SkBits2Float(0x00000000)); // 0, 0
path.cubicTo(SkBits2Float(0x1931204a), SkBits2Float(0x2ba1a14a), SkBits2Float(0x4a4a08ff), SkBits2Float(0x4a4a08ff), SkBits2Float(0x4a4a4a34), SkBits2Float(0x4a4a4a4a)); // 9.15721e-24f, 1.14845e-12f, 3.31014e+06f, 3.31014e+06f, 3.31432e+06f, 3.31432e+06f
path.moveTo(SkBits2Float(0x000010a1), SkBits2Float(0x19312000)); // 5.96533e-42f, 9.15715e-24f
path.cubicTo(SkBits2Float(0x4a4a4a4a), SkBits2Float(0x4a4a4a4a), SkBits2Float(0xa14a4a4a), SkBits2Float(0x08ff2ba1), SkBits2Float(0x08ff4a4a), SkBits2Float(0x4a344a4a)); // 3.31432e+06f, 3.31432e+06f, -6.85386e-19f, 1.53575e-33f, 1.53647e-33f, 2.95387e+06f
path.cubicTo(SkBits2Float(0x544a4a4a), SkBits2Float(0x4a4a4a4a), SkBits2Float(0x2ba1a14a), SkBits2Float(0x4e4a08ff), SkBits2Float(0x4a4a4a4a), SkBits2Float(0xa1a181ff)); // 3.47532e+12f, 3.31432e+06f, 1.14845e-12f, 8.47397e+08f, 3.31432e+06f, -1.09442e-18f
SkPath path2(path);
testPathOpCheck(reporter, path1, path2, kReverseDifference_SkPathOp, filename, true);
fix another pathops fuzz bug Tunnel out the error case when the numerics create uncalculable angles. R=mmoroz@chromium.org TBR=reed@google.com BUG=624351 GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2103863005 Review-Url: https://codereview.chromium.org/2103863005
2016-06-29 17:46:08 +00:00
}
fix fuzzer Previous spans always have a valid next pointer. The final span does not. Change the test for a valid link to take into consideration whether the links are chased forwards or backwards. TBR=reed@google.com BUG=629454 GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2166543002 Review-Url: https://codereview.chromium.org/2166543002
2016-07-19 18:29:14 +00:00
static void fuzz763_6(skiatest::Reporter* reporter, const char* filename) {
SkPath path;
path.setFillType((SkPath::FillType) 1);
path.moveTo(SkBits2Float(0x00000000), SkBits2Float(0x6a2a291f)); // 0, 5.14279e+25f
path.cubicTo(SkBits2Float(0x68295b2d), SkBits2Float(0x00000000), SkBits2Float(0x00000000), SkBits2Float(0x00000000), SkBits2Float(0x00000000), SkBits2Float(0x68556829)); // 3.19905e+24f, 0, 0, 0, 0, 4.03114e+24f
path.lineTo(SkBits2Float(0x00000000), SkBits2Float(0x68555b2a)); // 0, 4.03018e+24f
path.cubicTo(SkBits2Float(0x00000000), SkBits2Float(0x67d55b2a), SkBits2Float(0x67296a4b), SkBits2Float(0x67555b2a), SkBits2Float(0x677e1f70), SkBits2Float(0x66d55b2a)); // 0, 2.01509e+24f, 8.00041e+23f, 1.00755e+24f, 1.20006e+24f, 5.03773e+23f
path.cubicTo(SkBits2Float(0x678f0684), SkBits2Float(0x6684f008), SkBits2Float(0x6798f8ea), SkBits2Float(0x6625a942), SkBits2Float(0x67961914), SkBits2Float(0x65ce709a)); // 1.35084e+24f, 3.1389e+23f, 1.44478e+24f, 1.95578e+23f, 1.41764e+24f, 1.21861e+23f
path.cubicTo(SkBits2Float(0x679174f7), SkBits2Float(0x63199132), SkBits2Float(0x6756c79f), SkBits2Float(0x606478de), SkBits2Float(0x65682bcf), SkBits2Float(0x00000000)); // 1.3738e+24f, 2.83281e+21f, 1.01427e+24f, 6.58526e+19f, 6.85248e+22f, 0
path.conicTo(SkBits2Float(0x68295b02), SkBits2Float(0x60f7f28b), SkBits2Float(0x00000000), SkBits2Float(0x6a2a291f), SkBits2Float(0x42784f5a)); // 3.19903e+24f, 1.42932e+20f, 0, 5.14279e+25f, 62.0775f
path.close();
path.moveTo(SkBits2Float(0x654d6d10), SkBits2Float(0x00000000)); // 6.06311e+22f, 0
path.lineTo(SkBits2Float(0x6a4b7bc0), SkBits2Float(0x00000000)); // 6.14991e+25f, 0
path.lineTo(SkBits2Float(0x00000000), SkBits2Float(0x6a4b7bc0)); // 0, 6.14991e+25f
path.lineTo(SkBits2Float(0x00000000), SkBits2Float(0x00000000)); // 0, 0
SkPath path1(path);
path.reset();
path.setFillType((SkPath::FillType) 0);
path.moveTo(SkBits2Float(0x3ac23a55), SkBits2Float(0x2a292827)); // 0.00148184f, 1.50241e-13f
path.lineTo(SkBits2Float(0x63962be6), SkBits2Float(0x272a812a)); // 5.54035e+21f, 2.36623e-15f
SkPath path2(path);
testPathOpFail(reporter, path1, path2, (SkPathOp) 0, filename);
}
fix fuzzer bug Fix another fuzzer bug. Some PathOps asserts only make sense if the incoming data is well-behaved. Well-behaved tests set debugging state to trigger these additional asserts. Formalize this by creating macros similar to SkASSERT that check to see if the assert should be skipped. TBR=reed@google.com BUG=629962 GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2169863002 Review-Url: https://codereview.chromium.org/2169863002
2016-07-21 12:48:43 +00:00
static void fuzz763_7(skiatest::Reporter* reporter, const char* filename) {
SkPath path;
path.setFillType((SkPath::FillType) 0);
SkPath path1(path);
path.reset();
path.setFillType((SkPath::FillType) 0);
path.moveTo(SkBits2Float(0x68556829), SkBits2Float(0x555b2d29)); // 4.03114e+24f, 1.50617e+13f
path.moveTo(SkBits2Float(0x0f2a312a), SkBits2Float(0xc0032108)); // 8.39112e-30f, -2.04889f
path.cubicTo(SkBits2Float(0x68392d55), SkBits2Float(0xf05b684b), SkBits2Float(0x8c55272d), SkBits2Float(0x212a1f2a), SkBits2Float(0x0321082a), SkBits2Float(0x6a4b7bc0)); // 3.4979e+24f, -2.71613e+29f, -1.64207e-31f, 5.76395e-19f, 4.7323e-37f, 6.14991e+25f
path.conicTo(SkBits2Float(0x212a8ced), SkBits2Float(0x0321081f), SkBits2Float(0x6a4b7bc0), SkBits2Float(0x2829ed84), SkBits2Float(0x2d555b2d)); // 5.77848e-19f, 4.7323e-37f, 6.14991e+25f, 9.43289e-15f, 1.21279e-11f
path.moveTo(SkBits2Float(0x68345b2d), SkBits2Float(0xf0682955)); // 3.40683e+24f, -2.87402e+29f
path.conicTo(SkBits2Float(0x212a1f5b), SkBits2Float(0xef2a8c55), SkBits2Float(0x295b2d2a), SkBits2Float(0x08685568), SkBits2Float(0x7bc00321)); // 5.76397e-19f, -5.27821e+28f, 4.86669e-14f, 6.99154e-34f, 1.99397e+36f
path.lineTo(SkBits2Float(0x68345b2d), SkBits2Float(0xf0682955)); // 3.40683e+24f, -2.87402e+29f
path.close();
path.moveTo(SkBits2Float(0x68345b2d), SkBits2Float(0xf0682955)); // 3.40683e+24f, -2.87402e+29f
path.lineTo(SkBits2Float(0x5b2c6829), SkBits2Float(0x212a8c55)); // 4.85282e+16f, 5.7784e-19f
path.moveTo(SkBits2Float(0x0321081f), SkBits2Float(0x6a4b7bc0)); // 4.7323e-37f, 6.14991e+25f
path.lineTo(SkBits2Float(0x3a8a3adf), SkBits2Float(0x8a281a4f)); // 0.00105461f, -8.09385e-33f
path.quadTo(SkBits2Float(0x1d2a2928), SkBits2Float(0x43962be6), SkBits2Float(0x272a812a), SkBits2Float(0x3a2a5529)); // 2.25206e-21f, 300.343f, 2.36623e-15f, 0.000649768f
path.lineTo(SkBits2Float(0x213b1e2a), SkBits2Float(0x27292720)); // 6.3398e-19f, 2.34747e-15f
path.conicTo(SkBits2Float(0xba1f203a), SkBits2Float(0xc422c538), SkBits2Float(0x215d5927), SkBits2Float(0x70ec2ac2), SkBits2Float(0x2a51523a)); // -0.000607017f, -651.082f, 7.49957e-19f, 5.84721e+29f, 1.85915e-13f
path.quadTo(SkBits2Float(0x633ad912), SkBits2Float(0x29c80927), SkBits2Float(0x272927b0), SkBits2Float(0x683a5b2d)); // 3.44674e+21f, 8.88337e-14f, 2.3475e-15f, 3.52017e+24f
path.lineTo(SkBits2Float(0x295b2d68), SkBits2Float(0x29685568)); // 4.86672e-14f, 5.15884e-14f
path.conicTo(SkBits2Float(0xaa8c555b), SkBits2Float(0x081f2a21), SkBits2Float(0x5b2d0321), SkBits2Float(0x68556829), SkBits2Float(0x2a552d29)); // -2.49282e-13f, 4.78968e-34f, 4.86986e+16f, 4.03114e+24f, 1.89339e-13f
path.cubicTo(SkBits2Float(0x21295b2d), SkBits2Float(0x2a688c5b), SkBits2Float(0x68295b2d), SkBits2Float(0x2d296855), SkBits2Float(0x8c08555b), SkBits2Float(0x2a2a29ca)); // 5.73801e-19f, 2.06544e-13f, 3.19905e+24f, 9.6297e-12f, -1.05027e-31f, 1.51135e-13f
path.quadTo(SkBits2Float(0x68295b21), SkBits2Float(0x2d296855), SkBits2Float(0x2a8c555b), SkBits2Float(0x081f2a21)); // 3.19904e+24f, 9.6297e-12f, 2.49282e-13f, 4.78968e-34f
path.lineTo(SkBits2Float(0x0321081f), SkBits2Float(0x6a4b7bc0)); // 4.7323e-37f, 6.14991e+25f
path.close();
path.moveTo(SkBits2Float(0x0321081f), SkBits2Float(0x6a4b7bc0)); // 4.7323e-37f, 6.14991e+25f
path.conicTo(SkBits2Float(0x6a4b7bc0), SkBits2Float(0x5b2d6829), SkBits2Float(0x212a8c55), SkBits2Float(0xed7aba1f), SkBits2Float(0x2a212a8c)); // 6.14991e+25f, 4.88097e+16f, 5.7784e-19f, -4.84977e+27f, 1.43144e-13f
path.moveTo(SkBits2Float(0x2d212d08), SkBits2Float(0x5568295b)); // 9.16179e-12f, 1.5954e+13f
path.moveTo(SkBits2Float(0x5529685b), SkBits2Float(0x11295b68)); // 1.16416e+13f, 1.33599e-28f
path.conicTo(SkBits2Float(0x5b782968), SkBits2Float(0x3a292d55), SkBits2Float(0x2a8c555b), SkBits2Float(0x68295a2d), SkBits2Float(0x2d296855)); // 6.98513e+16f, 0.000645359f, 2.49282e-13f, 3.19897e+24f, 9.6297e-12f
path.moveTo(SkBits2Float(0x555b8c55), SkBits2Float(0x21682929)); // 1.50872e+13f, 7.86591e-19f
path.moveTo(SkBits2Float(0x0321081f), SkBits2Float(0x6a4b7bc0)); // 4.7323e-37f, 6.14991e+25f
path.conicTo(SkBits2Float(0xac2d8ced), SkBits2Float(0x5b682968), SkBits2Float(0x5b292d55), SkBits2Float(0x212a8c55), SkBits2Float(0x081f282a)); // -2.4663e-12f, 6.53477e+16f, 4.76191e+16f, 5.7784e-19f, 4.78945e-34f
path.lineTo(SkBits2Float(0x0321081f), SkBits2Float(0x6a4b7bc0)); // 4.7323e-37f, 6.14991e+25f
path.close();
path.moveTo(SkBits2Float(0x0321081f), SkBits2Float(0x6a4b7bc0)); // 4.7323e-37f, 6.14991e+25f
path.conicTo(SkBits2Float(0x6a4b7bc0), SkBits2Float(0x2a8ced7a), SkBits2Float(0x03081f21), SkBits2Float(0x6a3a7bc0), SkBits2Float(0x2147ed7a)); // 6.14991e+25f, 2.50338e-13f, 4.00025e-37f, 5.63611e+25f, 6.77381e-19f
path.lineTo(SkBits2Float(0x0321081f), SkBits2Float(0x6a4b7bc0)); // 4.7323e-37f, 6.14991e+25f
path.close();
path.moveTo(SkBits2Float(0x0321081f), SkBits2Float(0x6a4b7bc0)); // 4.7323e-37f, 6.14991e+25f
path.quadTo(SkBits2Float(0x2d28282a), SkBits2Float(0x5568295b), SkBits2Float(0x3a21df68), SkBits2Float(0x4f9a3a8a)); // 9.55861e-12f, 1.5954e+13f, 0.000617495f, 5.17506e+09f
path.lineTo(SkBits2Float(0x0321081f), SkBits2Float(0x6a4b7bc0)); // 4.7323e-37f, 6.14991e+25f
path.close();
path.moveTo(SkBits2Float(0x0321081f), SkBits2Float(0x6a4b7bc0)); // 4.7323e-37f, 6.14991e+25f
path.cubicTo(SkBits2Float(0x5568c23a), SkBits2Float(0x5b2d2968), SkBits2Float(0x212a8c55), SkBits2Float(0x21081f2a), SkBits2Float(0x3a7bc003), SkBits2Float(0x294b2827)); // 1.59951e+13f, 4.87407e+16f, 5.7784e-19f, 4.61198e-19f, 0.00096035f, 4.51099e-14f
SkPath path2(path);
testPathOpFail(reporter, path1, path2, (SkPathOp) 0, filename);
}
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
static struct TestDesc failTests[] = {
fix fuzzer bug Fix another fuzzer bug. Some PathOps asserts only make sense if the incoming data is well-behaved. Well-behaved tests set debugging state to trigger these additional asserts. Formalize this by creating macros similar to SkASSERT that check to see if the assert should be skipped. TBR=reed@google.com BUG=629962 GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2169863002 Review-Url: https://codereview.chromium.org/2169863002
2016-07-21 12:48:43 +00:00
TEST(fuzz763_7),
fix fuzzer Previous spans always have a valid next pointer. The final span does not. Change the test for a valid link to take into consideration whether the links are chased forwards or backwards. TBR=reed@google.com BUG=629454 GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2166543002 Review-Url: https://codereview.chromium.org/2166543002
2016-07-19 18:29:14 +00:00
TEST(fuzz763_6),
pathops coincidence and security rewrite Most changes stem from working on an examples bracketed by #if DEBUG_UNDER_DEVELOPMENT // tiger These exposed many problems with coincident curves, as well as errors throughout the code. Fixing these errors also fixed a number of fuzzer-inspired bug reports. * Line/Curve Intersections Check to see if the end of the line nearly intersects the curve. This was a FIXME in the old code. * Performance Use a central chunk allocator. Plumb the allocator into the global variable state so that it can be shared. (Note that 'SkGlobalState' is allocated on the stack and is visible to children functions but not other threads.) * Refactor Let SkOpAngle grow up from a structure to a class. Let SkCoincidentSpans grow up from a structure to a class. Rename enum Alias to AliasMatch. * Coincidence Rewrite Add more debugging to coincidence detection. Parallel debugging routines have read-only logic to report the current coincidence state so that steps through the logic can expose whether things got better or worse. More functions can error-out and cause the pathops engine to non-destructively exit. * Accuracy Remove code that adjusted point locations. Instead, offset the curve part so that sorted curves all use the same origin. Reduce the size (and influence) of magic numbers. * Testing The debug suite with verify and the full release suite ./out/Debug/pathops_unittest -v -V ./out/Release/pathops_unittest -v -V -x expose one error. That error is captured as cubics_d3. This error exists in the checked in code as well. BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 Review-Url: https://codereview.chromium.org/2128633003
2016-07-18 17:01:36 +00:00
TEST(fuzz763_2c),
TEST(fuzz763_2b),
TEST(fuzz763_2a),
TEST(fuzz763_5a),
TEST(fuzz763_3a),
TEST(fuzz763_1a),
TEST(fuzz763_1b),
TEST(fuzz763_1c),
fix another pathops fuzz bug Tunnel out the error case when the numerics create uncalculable angles. R=mmoroz@chromium.org TBR=reed@google.com BUG=624351 GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2103863005 Review-Url: https://codereview.chromium.org/2103863005
2016-06-29 17:46:08 +00:00
TEST(fuzz763_2),
TEST(fuzz763_5),
fix fuzz test that exceeds numeric limit The extreme values here exceed an internal test that expects computed numbers to be less than MAX_FLT. Use MAX_DBL instead. R=mmoroz@chromium.org TBR=reed@google.com BUG=624346 GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2103903004 Review-Url: https://codereview.chromium.org/2103903004
2016-06-29 15:18:38 +00:00
TEST(fuzz763_3),
fix fuzz bugs Detect more places where the pathops numerics cause numbers to become nearly identical and subsequently fail. These tests have extreme inputs and cannot succeed. Also remove the expectSuccess parameter from PathOpsDebug and check instead in the test framework. R=mbarbella@chromium.org TBR=reed@google.com BUG=623072,623022 GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2103513002 Review-Url: https://codereview.chromium.org/2103513002
2016-06-28 16:23:57 +00:00
TEST(fuzz763_4),
TEST(fuzz763_9),
fix pathops fuzz bugs Fail out in a couple of new places when the input data is very large and exceeds the limits of the pathops machinery. Most of the change here plumbs in a way to exclude an assert in one of these exceptional cases. The current SkAddIntersection implementation and the inner functions it calls has no way to report an error to the root caller for an early exit, so rather than add that in, exclude the assert when the test that would trigger it runs (allowing the test to otherwise ensure that it properly fails). TBR=reed@google.com BUG=617586,617635 GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2046713003 Review-Url: https://codereview.chromium.org/2046713003
2016-06-08 11:28:19 +00:00
TEST(fuzz1450_1),
TEST(fuzz1450_0),
give up if huge paths have unresolvable coincidence This fuzzy test has enormous curves with coincidence runs that break numerics. If the computed intersections identify that the span of coincidence has been deleted, give up and return that the path op failed. TBR=reed@google.com BUG=597926 GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1854333002 Review URL: https://codereview.chromium.org/1854333002
2016-04-05 14:28:48 +00:00
TEST(bug597926_0),
fix path ops fuzz bug If one path is empty and the other has extreme values, the intermediate coincident paths cannot be resolved, but triggers an assert that a data structure unexpectedly has zero-length. Tunnel this failure back up to the top and return that the entire path op fails. A future optimization could detect the empty path and avoid this, allowing the op to succeed -- not sure that it's worth the additional logic though. TBR=reed@google.com BUG=535151 GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1730293002 Review URL: https://codereview.chromium.org/1730293002
2016-02-24 17:03:07 +00:00
TEST(fuzz535151),
path ops: fix fuzz-found divide by zero TBR=reed@google.com BUG=542391 Review URL: https://codereview.chromium.org/1427913005
2015-11-02 15:32:58 +00:00
TEST(fuzz753_91),
Enabling clip stack flattening exercises path ops. Iterating through the 903K skps that represent the imagable 1M top web pages triggers a number of bugs, some of which are addressed here. Some web pages trigger intersecting cubic representations of arc with their conic counterparts. This exposed a flaw in coincident detection that caused an infinite loop. The loop alternatively extended the coincident section and, determining the that the bounds of the curve pairs did not overlap, deleted the extension. Track the number of times the coincident detection is called, and if it exceeds an empirically found limit, assume that the curves are coincident and force it to be so. The loop count limit can be determined by enabling DEBUG_T_SECT_LOOP_COUNT and running all tests. The largest count is reported on completion. Another class of bugs was caused by concident detection duplicating nearly identical points that had been merged earlier. To track these bugs, the 'handle coincidence' code was duplicated as a const debug variety that reported if one of a dozen or so irregularities are present; then it is easier to see when a block of code that fixes one irregularity regresses another. Creating the debug const code version exposed some non-debug code that could be const, and some that was experimental and could be removed. Set DEBUG_COINCIDENCE to track coincidence health and handling. For running on Chrome, DEBUG_VERIFY checks the result of pathops against the same operation using SkRegion to verify that the results are nearly the same. When visualizing the pathops work using tools/pathops_visualizer.htm, set DEBUG_DUMP_ALIGNMENT to see the curves after they've been aligned for coincidence. Other bugs fixed include detecting when a section of a pair of curves have devolved into lines and are coincident. TBR=reed@google.com Review URL: https://codereview.chromium.org/1394503003
2015-10-16 16:03:38 +00:00
TEST(fuzz714),
Path ops formerly found the topmost unprocessed edge and determined its angle sort order to initialize the winding. This never worked correctly with cubics and was flaky with paths consisting mostly of vertical edges. This replacement shoots axis-aligned rays through all intersecting edges to find the outermost one either horizontally or vertically. The resulting code is smaller and twice as fast. To support this, most of the horizontal / vertical intersection code was rewritten and standardized, and old code supporting the top-directed winding was deleted. Contours were pointed to by an SkTDArray. Instead, put them in a linked list, and designate the list head with its own class to ensure that methods that take lists of contours start at the top. This change removed a large percentage of memory allocations used by path ops. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1111333002
2015-05-11 14:21:27 +00:00
TEST(fuzz487a),
fix multiple intersection logic When three or more curves intersect at the same point, ensure that each curve records the intersections of the others. This fixes a number of cubic tests. TBR=reed@google.com BUG=skia:3588 Review URL: https://codereview.chromium.org/1105943002
2015-04-24 16:08:57 +00:00
TEST(fuzz433),
abort op early if path isn't parseable TBR= BUG=419649 Review URL: https://codereview.chromium.org/623943002
2014-10-03 12:36:27 +00:00
TEST(fuzz1),
fail early if coincidence can't be resolved Bail out if a very large value causes coincidence resolution to fail. TBR= BUG=415866 Author: caryclark@google.com Review URL: https://codereview.chromium.org/585913002
2014-09-19 13:33:31 +00:00
TEST(fuzz487b),
fail when coincidence is too far apart TBR= BUG=410552 Author: caryclark@google.com Review URL: https://codereview.chromium.org/556433002
2014-09-08 17:25:38 +00:00
TEST(fuzz433b),
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
TEST(bufferOverflow),
};
static const size_t failTestCount = SK_ARRAY_COUNT(failTests);
DEF_TEST(PathOpsFailOp, reporter) {
#if DEBUG_SHOW_TEST_NAME
strncpy(DEBUG_FILENAME_STRING, "", DEBUG_FILENAME_STRING_LENGTH);
#endif
Style Change: NULL->nullptr DOCS_PREVIEW= https://skia.org/?cl=1316233002 Review URL: https://codereview.chromium.org/1316233002
2015-08-27 14:41:13 +00:00
RunTestSet(reporter, failTests, failTestCount, nullptr, nullptr, nullptr, false);
Mike R: please sanity check SkPostConfig.h Mike K: please sanity check Test.cpp and skia_test.cpp Feel free to look at the rest, but I don't expect any in depth review of path ops innards. Path Ops first iteration used QuickSort to order segments radiating from an intersection to compute the winding rule. This revision uses a circular sort instead. Breaking out the circular sort into its own long-lived structure (SkOpAngle) allows doing less work and provides a home for caching additional sorting data. The circle sort is more stable than the former sort, has a robust ordering and fewer exceptions. It finds unsortable ordering less often. It is less reliant on the initial curve tangent, using convex hulls instead whenever it can. Additional debug validation makes sure that the computed structures are self-consistent. A new visualization tool helps verify that the angle ordering is correct. The 70+M tests pass with this change on Windows, Mac, Linux 32 and Linux 64 in debug and release. R=mtklein@google.com, reed@google.com Author: caryclark@google.com Review URL: https://codereview.chromium.org/131103009 git-svn-id: http://skia.googlecode.com/svn/trunk@14183 2bbb7eff-a529-9590-31e7-b0007b416f81
2014-04-14 17:08:59 +00:00
}
pathops coincidence and security rewrite Most changes stem from working on an examples bracketed by #if DEBUG_UNDER_DEVELOPMENT // tiger These exposed many problems with coincident curves, as well as errors throughout the code. Fixing these errors also fixed a number of fuzzer-inspired bug reports. * Line/Curve Intersections Check to see if the end of the line nearly intersects the curve. This was a FIXME in the old code. * Performance Use a central chunk allocator. Plumb the allocator into the global variable state so that it can be shared. (Note that 'SkGlobalState' is allocated on the stack and is visible to children functions but not other threads.) * Refactor Let SkOpAngle grow up from a structure to a class. Let SkCoincidentSpans grow up from a structure to a class. Rename enum Alias to AliasMatch. * Coincidence Rewrite Add more debugging to coincidence detection. Parallel debugging routines have read-only logic to report the current coincidence state so that steps through the logic can expose whether things got better or worse. More functions can error-out and cause the pathops engine to non-destructively exit. * Accuracy Remove code that adjusted point locations. Instead, offset the curve part so that sorted curves all use the same origin. Reduce the size (and influence) of magic numbers. * Testing The debug suite with verify and the full release suite ./out/Debug/pathops_unittest -v -V ./out/Release/pathops_unittest -v -V -x expose one error. That error is captured as cubics_d3. This error exists in the checked in code as well. BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 BUG=skia: GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2128633003 Review-Url: https://codereview.chromium.org/2128633003
2016-07-18 17:01:36 +00:00
static struct TestDesc repTests[] = {
TEST(loops44i),
TEST(loops45i),
TEST(loops46i),
};
DEF_TEST(PathOpsRepOp, reporter) {
for (int index = 0; index < 2; ++index)
RunTestSet(reporter, repTests, SK_ARRAY_COUNT(repTests), nullptr, nullptr, nullptr, false);
}
Reference in New Issue Copy Permalink