skia2/dm/DM.cpp

1595 lines
57 KiB
C++
Raw Normal View History

/*
* Copyright 2013 Google Inc.
*
* Use of this source code is governed by a BSD-style license that can be
* found in the LICENSE file.
*/
#include "dm/DMJsonWriter.h"
#include "dm/DMSrcSink.h"
#include "gm/verifiers/gmverifier.h"
#include "include/codec/SkCodec.h"
#include "include/core/SkBBHFactory.h"
#include "include/core/SkColorPriv.h"
#include "include/core/SkColorSpace.h"
#include "include/core/SkData.h"
#include "include/core/SkDocument.h"
#include "include/core/SkGraphics.h"
#include "include/private/SkChecksum.h"
#include "include/private/SkHalf.h"
#include "include/private/SkSpinlock.h"
#include "include/private/SkTHash.h"
#include "src/core/SkColorSpacePriv.h"
#include "src/core/SkLeanWindows.h"
#include "src/core/SkMD5.h"
#include "src/core/SkOSFile.h"
#include "src/core/SkTaskGroup.h"
#include "src/utils/SkOSPath.h"
#include "tests/Test.h"
#include "tools/AutoreleasePool.h"
#include "tools/HashAndEncode.h"
#include "tools/ProcStats.h"
#include "tools/Resources.h"
#include "tools/ToolUtils.h"
#include "tools/flags/CommonFlags.h"
#include "tools/flags/CommonFlagsConfig.h"
#include "tools/ios_utils.h"
#include "tools/trace/ChromeTracingTracer.h"
#include "tools/trace/EventTracingPriv.h"
#include "tools/trace/SkDebugfTracer.h"
#include <memory>
Add support for multiple frames in SkCodec Add an interface to decode frames beyond the first in SkCodec, and add an implementation for SkGifCodec. Add getFrameData to SkCodec. This method reads ahead in the stream to return a vector containing meta data about each frame in the image. This is not required in order to decode frames beyond the first, but it allows a client to learn extra information: - how long the frame should be displayed - whether a frame should be blended with a prior frame, allowing the client to provide the prior frame to speed up decoding Add a new fields to SkCodec::Options: - fFrameIndex - fHasPriorFrame The API is designed so that SkCodec never caches frames. If a client wants a frame beyond the first, they specify the frame in Options.fFrameIndex. If the client does not have the frame's required frame (the frame that this frame must be blended on top of) cached, they pass false for Options.fHasPriorFrame. Unless the frame is independent, the codec will then recursively decode all frames necessary to decode fFrameIndex. If the client has the required frame cached, they can put it in the dst they pass to the codec, and the codec will only draw fFrameIndex onto it. Replace SkGifCodec's scanline decoding support with progressive decoding, and update the tests accordingly. Implement new APIs in SkGifCodec. Instead of using gif_lib, use GIFImageReader, imported from Chromium (along with its copyright headers) with the following changes: - SkGifCodec is now the client - Replace blink types - Combine GIFColorMap::buildTable and ::getTable into a method that creates and returns an SkColorTable - Input comes from an SkStream, instead of a SegmentReader. Add SkStreamBuffer, which buffers the (potentially partial) stream in order to decode progressively. (FIXME: This requires copying data that previously was read directly from the SegmentReader. Does this hurt performance? If so, can we fix it?) - Remove UMA code - Instead of reporting screen width and height to the client, allow the client to query for it - Fail earlier if the first frame AND screen have size of zero - Compute required previous frame when adding a new one - Move GIFParseQuery from GIFImageDecoder to GIFImageReader - Allow parsing up to a specific frame (to skip parsing the rest of the stream if a client only wants the first frame) - Compute whether the first frame has alpha and supports index 8, to create the SkImageInfo. This happens before reporting that the size has been decoded. Add GIFImageDecoder::haveDecodedRow to SkGifCodec, imported from Chromium (along with its copyright header), with the following changes: - Add support for sampling - Use the swizzler - Keep track of the rows decoded - Do *not* keep track of whether we've seen alpha Remove SkCodec::kOutOfOrder_SkScanlineOrder, which was only used by GIF scanline decoding. Call onRewind even if there is no stream (SkGifCodec needs to clear its decoded state so it will decode from the beginning). Add a method to SkSwizzler to access the offset into the dst, taking subsetting into account. Add a GM that animates a GIF. Add tests for the new APIs. *** Behavior changes: * Previously, we reported that an image with a subset frame and no transparent index was opaque and used the background index (if present) to fill the background. This is necessary in order to support index 8, but it does not match viewers/browsers I have seen. Examples: - Chromium and Gimp render the background transparent - Firefox, Safari, Linux Image Viewer, Safari Preview clip to the frame (for a single frame image) This CL matches Chromium's behavior and renders the background transparent. This allows us to have consistent behavior across products and simplifies the code (relative to what we would have to do to continue the old behavior on Android). It also means that we will no longer support index 8 for some GIFs. * Stop checking for GIFSTAMP - all GIFs should be either 89a or 87a. This matches Chromium. I suspect that bugs would have been reported if valid GIFs started with "GIFVER" instead of "GIF89a" or "GIF87a" (but did not decode in Chromium). *** Future work not included in this CL: * Move some checks out of haveDecodedRow, since they are the same for the entire frame e.g. - intersecting the frameRect with the full image size - whether there is a color table * Change when we write transparent pixels - In some cases, Chromium deemed this unnecessary, but I suspect it is slower than the fallback case. There will continue to be cases where we should *not* write them, but for e.g. the first pass where we have already cleared to transparent (which we may also be able to skip) writing the transparent pixels will not make anything incorrect. * Report color type and alpha type per frame - Depending on alpha values, disposal methods, frame rects, etc, subsequent frames may have different properties than the first. * Skip copies of the encoded data - We copy the encoded data in case the stream is one that cannot be rewound, so we can parse and then decode (possibly not immediately). For some input streams, this is unnecessary. - I was concerned this cause a performance regression, but on average the new code is faster than the old for the images I tested [1]. - It may cause a performance regression for Chromium, though, where we can always move back in the stream, so this should be addressed. Design doc: https://docs.google.com/a/google.com/document/d/12Qhf9T92MWfdWujQwCIjhCO3sw6pTJB5pJBwDM1T7Kc/ [1] https://docs.google.com/a/google.com/spreadsheets/d/19V-t9BfbFw5eiwBTKA1qOBkZbchjlTC5EIz6HFy-6RI/ GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=2045293002 Review-Url: https://codereview.chromium.org/2045293002
2016-10-24 16:03:26 +00:00
#include <vector>
#include <stdlib.h>
#ifndef SK_BUILD_FOR_WIN
#include <unistd.h>
#endif
#if defined(SK_BUILD_FOR_ANDROID_FRAMEWORK) && defined(SK_HAS_HEIF_LIBRARY)
#include <binder/IPCThreadState.h>
#endif
#if defined(SK_BUILD_FOR_MAC)
#include "include/utils/mac/SkCGUtils.h"
#include "src/utils/mac/SkUniqueCFRef.h"
#endif
extern bool gSkForceRasterPipelineBlitter;
extern bool gUseSkVMBlitter;
extern bool gSkVMAllowJIT;
static DEFINE_string(src, "tests gm skp mskp lottie rive svg image colorImage",
"Source types to test.");
static DEFINE_bool(nameByHash, false,
"If true, write to FLAGS_writePath[0]/<hash>.png instead of "
"to FLAGS_writePath[0]/<config>/<sourceType>/<sourceOptions>/<name>.png");
static DEFINE_bool2(pathOpsExtended, x, false, "Run extended pathOps tests.");
static DEFINE_string(matrix, "1 0 0 1",
"2x2 scale+skew matrix to apply or upright when using "
"'matrix' or 'upright' in config.");
static DEFINE_string(skip, "",
"Space-separated config/src/srcOptions/name quadruples to skip. "
"'_' matches anything. '~' negates the match. E.g. \n"
"'--skip gpu skp _ _' will skip all SKPs drawn into the gpu config.\n"
"'--skip gpu skp _ _ 8888 gm _ aarects' will also skip the aarects GM on 8888.\n"
"'--skip ~8888 svg _ svgparse_' blocks non-8888 SVGs that contain \"svgparse_\" in "
"the name.");
static DEFINE_string2(readPath, r, "",
"If set check for equality with golden results in this directory.");
DEFINE_string2(writePath, w, "", "If set, write bitmaps here as .pngs.");
static DEFINE_string(uninterestingHashesFile, "",
"File containing a list of uninteresting hashes. If a result hashes to something in "
"this list, no image is written for that result.");
static DEFINE_int(shards, 1, "We're splitting source data into this many shards.");
static DEFINE_int(shard, 0, "Which shard do I run?");
static DEFINE_string(mskps, "", "Directory to read mskps from, or a single mskp file.");
static DEFINE_bool(forceRasterPipeline, false, "sets gSkForceRasterPipelineBlitter");
static DEFINE_bool(skvm, false, "sets gUseSkVMBlitter");
static DEFINE_bool(jit, true, "sets gSkVMAllowJIT");
static DEFINE_string(bisect, "",
"Pair of: SKP file to bisect, followed by an l/r bisect trail string (e.g., 'lrll'). The "
"l/r trail specifies which half to keep at each step of a binary search through the SKP's "
"paths. An empty string performs no bisect. Only the SkPaths are bisected; all other draws "
"are thrown out. This is useful for finding a reduced repo case for path drawing bugs.");
static DEFINE_bool(ignoreSigInt, false, "ignore SIGINT signals during test execution");
static DEFINE_bool(checkF16, false, "Ensure that F16Norm pixels are clamped.");
static DEFINE_string(colorImages, "",
"List of images and/or directories to decode with color correction. "
"A directory with no images is treated as a fatal error.");
static DEFINE_bool2(veryVerbose, V, false, "tell individual tests to be verbose.");
static DEFINE_bool(cpu, true, "Run CPU-bound work?");
static DEFINE_bool(gpu, true, "Run GPU-bound work?");
static DEFINE_bool(graphite, true, "Run Graphite work?");
static DEFINE_bool(dryRun, false,
"just print the tests that would be run, without actually running them.");
static DEFINE_string(images, "",
"List of images and/or directories to decode. A directory with no images"
" is treated as a fatal error.");
static DEFINE_bool(simpleCodec, false,
"Runs of a subset of the codec tests, "
"with no scaling or subsetting, always using the canvas color type.");
static DEFINE_string2(match, m, nullptr,
"[~][^]substring[$] [...] of name to run.\n"
"Multiple matches may be separated by spaces.\n"
"~ causes a matching name to always be skipped\n"
"^ requires the start of the name to match\n"
"$ requires the end of the name to match\n"
"^ and $ requires an exact match\n"
"If a name does not match any list entry,\n"
"it is skipped unless some list entry starts with ~");
static DEFINE_bool2(quiet, q, false, "if true, don't print status updates.");
static DEFINE_bool2(verbose, v, false, "enable verbose output from the test driver.");
static DEFINE_string(skps, "skps", "Directory to read skps from.");
static DEFINE_string(lotties, "lotties", "Directory to read (Bodymovin) jsons from.");
static DEFINE_string(rives, "rives", "Directory to read Rive/Flare files from.");
static DEFINE_string(svgs, "", "Directory to read SVGs from, or a single SVG file.");
static DEFINE_int_2(threads, j, -1,
"Run threadsafe tests on a threadpool with this many extra threads, "
"defaulting to one extra thread per core.");
static DEFINE_string(key, "",
"Space-separated key/value pairs to add to JSON identifying this builder.");
static DEFINE_string(properties, "",
"Space-separated key/value pairs to add to JSON identifying this run.");
static DEFINE_bool(rasterize_pdf, false, "Rasterize PDFs when possible.");
static DEFINE_bool(runVerifiers, false,
"if true, run SkQP-style verification of GM-produced images.");
#if defined(__MSVC_RUNTIME_CHECKS)
#include <rtcapi.h>
int RuntimeCheckErrorFunc(int errorType, const char* filename, int linenumber,
const char* moduleName, const char* fmt, ...) {
va_list args;
va_start(args, fmt);
vfprintf(stderr, fmt, args);
va_end(args);
SkDebugf("Line #%d\nFile: %s\nModule: %s\n",
linenumber, filename ? filename : "Unknown", moduleName ? moduleName : "Unknwon");
return 1;
}
#endif
using namespace DM;
using sk_gpu_test::GrContextFactory;
[fuzzer] Remove GL from (now-Vulkan) build The fuzzer runs against the Vulkan version of Swiftshader. There are no libGL.so (etc) on the fuzz runtime, so we want to avoid linking against those. The GL code that is #ifdef'd out is still necessary to avoid timeouts on TSAN with our NVIDIA jobs. https://skia-review.googlesource.com/c/skia/+/502638 Procedure for testing this locally (and iterating): 1. In oss-fuzz checkout, run `python infra/helper.py shell skia` to pull up local interactive version of Docker fuzzer build image. 2. Run `compile` in fuzzer shell. Stop after the swiftshader compiles and is copied into /out with Ctrl + C. 3. Comment out the swiftshader compilation part [1] (no need to re-do this when modifying Skia code). `apt-get install nano -y` `nano ../build.sh` 4. Make change to Skia repo using normal methods. 5. Run the following in the Skia repo `git diff origin main > foo.patch` Copy the patch into the Docker shell using Ctrl+C and nano. 6. Apply the patch inside the Docker shell `git apply foo.patch` and re-compile (which should skip right to building the fuzzer libs) `compile` 7. Repeat 4-7 or make small changes directly in the Docker shell via nano. 8. When compilation and link succeeds, run `ldd /out/api_mock_gpu_canvas` to verify GL and friends were not dynamically linked. [1] https://github.com/google/oss-fuzz/pull/7214/files#diff-76f13875e33875cdd372f1f0933206be599cd87952f1bd1eaa57ca928ee9e3e1R49-R53 Change-Id: Idf569820527c1304b0e5a68fd36295be89dfa2a0 Bug: oss-fuzz:44132 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/503016 Reviewed-by: Brian Osman <brianosman@google.com> Commit-Queue: Kevin Lubick <kjlubick@google.com>
2022-02-02 14:55:42 +00:00
using sk_gpu_test::GLTestContext;
Revert "[fuzzer] Remove GL from (now-Vulkan) build" This reverts commit 805acda3f3aa3d9fc85c84bac5aa52c472ec3bb1. Reason for revert: Looks to be breaking the Android roll. See https://android-build.googleplex.com/builds/pending/P29733268/aosp_bramble-userdebug/latest/view/logs/build_error.log Original change's description: > [fuzzer] Remove GL from (now-Vulkan) build > > The fuzzer runs against the Vulkan version of Swiftshader. > There are no libGL.so (etc) on the fuzz runtime, so we > want to avoid linking against those. > > The GL code that is #ifdef'd out is still necessary to > avoid timeouts on TSAN with our NVIDIA jobs. > https://skia-review.googlesource.com/c/skia/+/502638 > > Procedure for testing this locally (and iterating): > 1. In oss-fuzz checkout, run > `python infra/helper.py shell skia` > to pull up local interactive version of Docker > fuzzer build image. > 2. Run `compile` in fuzzer shell. Stop after > the swiftshader compiles and is copied into /out > with Ctrl + C. > 3. Comment out the swiftshader compilation part [1] > (no need to re-do this when modifying Skia code). > `apt-get install nano -y` > `nano ../build.sh` > 4. Make change to Skia repo using normal methods. > 5. Run the following in the Skia repo > `git diff origin main > foo.patch` > Copy the patch into the Docker shell using Ctrl+C > and nano. > 6. Apply the patch inside the Docker shell > `git apply foo.patch` > and re-compile (which should skip right to > building the fuzzer libs) > `compile` > 7. Repeat 4-7 or make small changes directly in > the Docker shell via nano. > 8. When compilation and link succeeds, run > `ldd /out/api_mock_gpu_canvas` > to verify GL and friends were not dynamically linked. > > [1] https://github.com/google/oss-fuzz/pull/7214/files#diff-76f13875e33875cdd372f1f0933206be599cd87952f1bd1eaa57ca928ee9e3e1R49-R53 > > Change-Id: Idf569820527c1304b0e5a68fd36295be89dfa2a0 > Bug: oss-fuzz:44132 > Reviewed-on: https://skia-review.googlesource.com/c/skia/+/503016 > Reviewed-by: Brian Osman <brianosman@google.com> > Commit-Queue: Kevin Lubick <kjlubick@google.com> Bug: oss-fuzz:44132 Change-Id: I3832417c60ff425572717d37dc9609419922b18e No-Presubmit: true No-Tree-Checks: true No-Try: true Reviewed-on: https://skia-review.googlesource.com/c/skia/+/503351 Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com> Reviewed-by: Brian Osman <brianosman@google.com>
2022-02-02 22:03:33 +00:00
using sk_gpu_test::ContextInfo;
/*~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~*/
static FILE* gVLog;
template <typename... Args>
static void vlog(const char* fmt, Args&&... args) {
if (gVLog) {
fprintf(gVLog, fmt, args...);
fflush(gVLog);
}
}
template <typename... Args>
static void info(const char* fmt, Args&&... args) {
vlog(fmt, args...);
if (!FLAGS_quiet) {
printf(fmt, args...);
}
}
static void info(const char* fmt) {
if (!FLAGS_quiet) {
printf("%s", fmt); // Clang warns printf(fmt) is insecure.
}
}
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
static SkTArray<SkString>* gFailures = new SkTArray<SkString>;
static void fail(const SkString& err) {
static SkSpinlock mutex;
SkAutoSpinlock lock(mutex);
SkDebugf("\n\nFAILURE: %s\n\n", err.c_str());
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
gFailures->push_back(err);
}
struct Running {
SkString id;
SkThreadID thread;
void dump() const {
info("\t%s\n", id.c_str());
}
};
static void dump_json() {
if (!FLAGS_writePath.isEmpty()) {
JsonWriter::DumpJson(FLAGS_writePath[0], FLAGS_key, FLAGS_properties);
}
}
// We use a spinlock to make locking this in a signal handler _somewhat_ safe.
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
static SkSpinlock* gMutex = new SkSpinlock;
static int gPending;
static SkTArray<Running>* gRunning = new SkTArray<Running>;
static void done(const char* config, const char* src, const char* srcOptions, const char* name) {
SkString id = SkStringPrintf("%s %s %s %s", config, src, srcOptions, name);
vlog("done %s\n", id.c_str());
int pending;
{
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
SkAutoSpinlock lock(*gMutex);
for (int i = 0; i < gRunning->count(); i++) {
if (gRunning->at(i).id == id) {
gRunning->removeShuffle(i);
break;
}
}
pending = --gPending;
}
// We write out dm.json file and print out a progress update every once in a while.
// Notice this also handles the final dm.json and progress update when pending == 0.
if (pending % 500 == 0) {
dump_json();
int curr = sk_tools::getCurrResidentSetSizeMB(),
peak = sk_tools::getMaxResidentSetSizeMB();
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
SkAutoSpinlock lock(*gMutex);
info("\n%dMB RAM, %dMB peak, %d queued, %d active:\n",
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
curr, peak, gPending - gRunning->count(), gRunning->count());
for (auto& task : *gRunning) {
task.dump();
}
}
}
static void start(const char* config, const char* src, const char* srcOptions, const char* name) {
SkString id = SkStringPrintf("%s %s %s %s", config, src, srcOptions, name);
vlog("start %s\n", id.c_str());
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
SkAutoSpinlock lock(*gMutex);
gRunning->push_back({id,SkGetThreadID()});
}
static void find_culprit() {
// Assumes gMutex is locked.
SkThreadID thisThread = SkGetThreadID();
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
for (auto& task : *gRunning) {
if (task.thread == thisThread) {
info("Likely culprit:\n");
task.dump();
}
}
}
#if defined(SK_BUILD_FOR_WIN)
static LONG WINAPI crash_handler(EXCEPTION_POINTERS* e) {
static const struct {
const char* name;
DWORD code;
} kExceptions[] = {
#define _(E) {#E, E}
_(EXCEPTION_ACCESS_VIOLATION),
_(EXCEPTION_BREAKPOINT),
_(EXCEPTION_INT_DIVIDE_BY_ZERO),
_(EXCEPTION_STACK_OVERFLOW),
// TODO: more?
#undef _
};
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
SkAutoSpinlock lock(*gMutex);
const DWORD code = e->ExceptionRecord->ExceptionCode;
info("\nCaught exception %u", code);
for (const auto& exception : kExceptions) {
if (exception.code == code) {
info(" %s", exception.name);
}
}
info(", was running:\n");
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
for (auto& task : *gRunning) {
task.dump();
}
find_culprit();
fflush(stdout);
// Execute default exception handler... hopefully, exit.
return EXCEPTION_EXECUTE_HANDLER;
}
static void setup_crash_handler() {
SetUnhandledExceptionFilter(crash_handler);
}
#else
#include <signal.h>
#if !defined(SK_BUILD_FOR_ANDROID)
#include <execinfo.h>
#endif
static constexpr int max_of() { return 0; }
template <typename... Rest>
static constexpr int max_of(int x, Rest... rest) {
return x > max_of(rest...) ? x : max_of(rest...);
}
static void (*previous_handler[max_of(SIGABRT,SIGBUS,SIGFPE,SIGILL,SIGSEGV,SIGTERM)+1])(int);
static void crash_handler(int sig) {
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
SkAutoSpinlock lock(*gMutex);
info("\nCaught signal %d [%s] (%dMB RAM, peak %dMB), was running:\n",
sig, strsignal(sig),
sk_tools::getCurrResidentSetSizeMB(), sk_tools::getMaxResidentSetSizeMB());
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
for (auto& task : *gRunning) {
task.dump();
}
find_culprit();
#if !defined(SK_BUILD_FOR_ANDROID)
void* stack[128];
int count = backtrace(stack, SK_ARRAY_COUNT(stack));
char** symbols = backtrace_symbols(stack, count);
info("\nStack trace:\n");
for (int i = 0; i < count; i++) {
info(" %s\n", symbols[i]);
}
#endif
fflush(stdout);
if (sig == SIGINT && FLAGS_ignoreSigInt) {
info("Ignoring signal %d because of --ignoreSigInt.\n"
"This is probably a sign the bot is overloaded with work.\n", sig);
} else {
signal(sig, previous_handler[sig]);
raise(sig);
}
}
static void setup_crash_handler() {
const int kSignals[] = { SIGABRT, SIGBUS, SIGFPE, SIGILL, SIGINT, SIGSEGV, SIGTERM };
for (int sig : kSignals) {
previous_handler[sig] = signal(sig, crash_handler);
}
}
#endif
/*~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~*/
struct Gold : public SkString {
Gold() : SkString("") {}
Gold(const SkString& sink, const SkString& src,
const SkString& srcOptions, const SkString& name,
const SkString& md5)
: SkString("") {
this->append(sink);
this->append(src);
this->append(srcOptions);
this->append(name);
this->append(md5);
}
struct Hash {
uint32_t operator()(const Gold& g) const {
return SkGoodHash()((const SkString&)g);
}
};
};
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
static SkTHashSet<Gold, Gold::Hash>* gGold = new SkTHashSet<Gold, Gold::Hash>;
static void add_gold(JsonWriter::BitmapResult r) {
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
gGold->add(Gold(r.config, r.sourceType, r.sourceOptions, r.name, r.md5));
}
static void gather_gold() {
if (!FLAGS_readPath.isEmpty()) {
SkString path(FLAGS_readPath[0]);
path.append("/dm.json");
if (!JsonWriter::ReadJson(path.c_str(), add_gold)) {
fail(SkStringPrintf("Couldn't read %s for golden results.", path.c_str()));
}
}
}
/*~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~*/
#if defined(SK_BUILD_FOR_WIN)
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
static constexpr char kNewline[] = "\r\n";
#else
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
static constexpr char kNewline[] = "\n";
#endif
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
static SkTHashSet<SkString>* gUninterestingHashes = new SkTHashSet<SkString>;
static void gather_uninteresting_hashes() {
if (!FLAGS_uninterestingHashesFile.isEmpty()) {
sk_sp<SkData> data(SkData::MakeFromFileName(FLAGS_uninterestingHashesFile[0]));
if (!data) {
info("WARNING: unable to read uninteresting hashes from %s\n",
FLAGS_uninterestingHashesFile[0]);
return;
}
// Copy to a string to make sure SkStrSplit has a terminating \0 to find.
SkString contents((const char*)data->data(), data->size());
SkTArray<SkString> hashes;
SkStrSplit(contents.c_str(), kNewline, &hashes);
for (const SkString& hash : hashes) {
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
gUninterestingHashes->add(hash);
}
info("FYI: loaded %d distinct uninteresting hashes from %d lines\n",
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
gUninterestingHashes->count(), hashes.count());
}
}
/*~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~*/
struct TaggedSrc : public std::unique_ptr<Src> {
SkString tag;
SkString options;
};
struct TaggedSink : public std::unique_ptr<Sink> {
Add config options to run different GPU APIs to dm and nanobench Add extended config specification form that can be used to run different gpu backend with different APIs. The configs can be specified with the form: gpu(api=string,dit=bool,nvpr=bool,samples=int) This replaces and removes the --gpuAPI flag. All existing configs should still work. Adds following documentation: out/Debug/dm --help config Flags: --config: type: string default: 565 8888 gpu nonrendering Options: 565 8888 debug gpu gpudebug gpudft gpunull msaa16 msaa4 nonrendering null nullgpu nvprmsaa16 nvprmsaa4 pdf pdf_poppler skp svg xps or use extended form 'backend(option=value,...)'. Extended form: 'backend(option=value,...)' Possible backends and options: gpu(api=string,dit=bool,nvpr=bool,samples=int) GPU backend api type: string default: native. Select graphics API to use with gpu backend. Options: native Use platform default OpenGL or OpenGL ES backend. gl Use OpenGL. gles Use OpenGL ES. debug Use debug OpenGL. null Use null OpenGL. dit type: bool default: false. Use device independent text. nvpr type: bool default: false. Use NV_path_rendering OpenGL and OpenGL ES extension. samples type: int default: 0. Use multisampling with N samples. Predefined configs: gpu = gpu() msaa4 = gpu(samples=4) msaa16 = gpu(samples=16) nvprmsaa4 = gpu(nvpr=true,samples=4) nvprmsaa16 = gpu(nvpr=true,samples=16) gpudft = gpu(dit=true) gpudebug = gpu(api=debug) gpunull = gpu(api=null) debug = gpu(api=debug) nullgpu = gpu(api=null) BUG=skia:2992 Committed: https://skia.googlesource.com/skia/+/e13ca329fca4c28cf4e078561f591ab27b743d23 GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1490113005 Committed: https://skia.googlesource.com/skia/+/c8b4336444e7b90382e04e33665fb3b8490b825b Committed: https://skia.googlesource.com/skia/+/9ebc3f0ee6db215dde461dc4777d85988cf272dd Review URL: https://codereview.chromium.org/1490113005
2015-12-23 09:33:00 +00:00
SkString tag;
};
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
static constexpr bool kMemcpyOK = true;
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
static SkTArray<TaggedSrc, kMemcpyOK>* gSrcs = new SkTArray<TaggedSrc, kMemcpyOK>;
static SkTArray<TaggedSink, kMemcpyOK>* gSinks = new SkTArray<TaggedSink, kMemcpyOK>;
static bool in_shard() {
static int N = 0;
return N++ % FLAGS_shards == FLAGS_shard;
}
static void push_src(const char* tag, ImplicitString options, Src* inSrc) {
std::unique_ptr<Src> src(inSrc);
if (in_shard() && FLAGS_src.contains(tag) &&
!CommandLineFlags::ShouldSkip(FLAGS_match, src->name().c_str())) {
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
TaggedSrc& s = gSrcs->push_back();
s.reset(src.release());
s.tag = tag;
s.options = options;
}
}
static void push_codec_src(Path path, CodecSrc::Mode mode, CodecSrc::DstColorType dstColorType,
SkAlphaType dstAlphaType, float scale) {
if (FLAGS_simpleCodec) {
Add support for multiple frames in SkCodec Add an interface to decode frames beyond the first in SkCodec, and add an implementation for SkGifCodec. Add getFrameData to SkCodec. This method reads ahead in the stream to return a vector containing meta data about each frame in the image. This is not required in order to decode frames beyond the first, but it allows a client to learn extra information: - how long the frame should be displayed - whether a frame should be blended with a prior frame, allowing the client to provide the prior frame to speed up decoding Add a new fields to SkCodec::Options: - fFrameIndex - fHasPriorFrame The API is designed so that SkCodec never caches frames. If a client wants a frame beyond the first, they specify the frame in Options.fFrameIndex. If the client does not have the frame's required frame (the frame that this frame must be blended on top of) cached, they pass false for Options.fHasPriorFrame. Unless the frame is independent, the codec will then recursively decode all frames necessary to decode fFrameIndex. If the client has the required frame cached, they can put it in the dst they pass to the codec, and the codec will only draw fFrameIndex onto it. Replace SkGifCodec's scanline decoding support with progressive decoding, and update the tests accordingly. Implement new APIs in SkGifCodec. Instead of using gif_lib, use GIFImageReader, imported from Chromium (along with its copyright headers) with the following changes: - SkGifCodec is now the client - Replace blink types - Combine GIFColorMap::buildTable and ::getTable into a method that creates and returns an SkColorTable - Input comes from an SkStream, instead of a SegmentReader. Add SkStreamBuffer, which buffers the (potentially partial) stream in order to decode progressively. (FIXME: This requires copying data that previously was read directly from the SegmentReader. Does this hurt performance? If so, can we fix it?) - Remove UMA code - Instead of reporting screen width and height to the client, allow the client to query for it - Fail earlier if the first frame AND screen have size of zero - Compute required previous frame when adding a new one - Move GIFParseQuery from GIFImageDecoder to GIFImageReader - Allow parsing up to a specific frame (to skip parsing the rest of the stream if a client only wants the first frame) - Compute whether the first frame has alpha and supports index 8, to create the SkImageInfo. This happens before reporting that the size has been decoded. Add GIFImageDecoder::haveDecodedRow to SkGifCodec, imported from Chromium (along with its copyright header), with the following changes: - Add support for sampling - Use the swizzler - Keep track of the rows decoded - Do *not* keep track of whether we've seen alpha Remove SkCodec::kOutOfOrder_SkScanlineOrder, which was only used by GIF scanline decoding. Call onRewind even if there is no stream (SkGifCodec needs to clear its decoded state so it will decode from the beginning). Add a method to SkSwizzler to access the offset into the dst, taking subsetting into account. Add a GM that animates a GIF. Add tests for the new APIs. *** Behavior changes: * Previously, we reported that an image with a subset frame and no transparent index was opaque and used the background index (if present) to fill the background. This is necessary in order to support index 8, but it does not match viewers/browsers I have seen. Examples: - Chromium and Gimp render the background transparent - Firefox, Safari, Linux Image Viewer, Safari Preview clip to the frame (for a single frame image) This CL matches Chromium's behavior and renders the background transparent. This allows us to have consistent behavior across products and simplifies the code (relative to what we would have to do to continue the old behavior on Android). It also means that we will no longer support index 8 for some GIFs. * Stop checking for GIFSTAMP - all GIFs should be either 89a or 87a. This matches Chromium. I suspect that bugs would have been reported if valid GIFs started with "GIFVER" instead of "GIF89a" or "GIF87a" (but did not decode in Chromium). *** Future work not included in this CL: * Move some checks out of haveDecodedRow, since they are the same for the entire frame e.g. - intersecting the frameRect with the full image size - whether there is a color table * Change when we write transparent pixels - In some cases, Chromium deemed this unnecessary, but I suspect it is slower than the fallback case. There will continue to be cases where we should *not* write them, but for e.g. the first pass where we have already cleared to transparent (which we may also be able to skip) writing the transparent pixels will not make anything incorrect. * Report color type and alpha type per frame - Depending on alpha values, disposal methods, frame rects, etc, subsequent frames may have different properties than the first. * Skip copies of the encoded data - We copy the encoded data in case the stream is one that cannot be rewound, so we can parse and then decode (possibly not immediately). For some input streams, this is unnecessary. - I was concerned this cause a performance regression, but on average the new code is faster than the old for the images I tested [1]. - It may cause a performance regression for Chromium, though, where we can always move back in the stream, so this should be addressed. Design doc: https://docs.google.com/a/google.com/document/d/12Qhf9T92MWfdWujQwCIjhCO3sw6pTJB5pJBwDM1T7Kc/ [1] https://docs.google.com/a/google.com/spreadsheets/d/19V-t9BfbFw5eiwBTKA1qOBkZbchjlTC5EIz6HFy-6RI/ GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=2045293002 Review-Url: https://codereview.chromium.org/2045293002
2016-10-24 16:03:26 +00:00
const bool simple = CodecSrc::kCodec_Mode == mode || CodecSrc::kAnimated_Mode == mode;
if (!simple || dstColorType != CodecSrc::kGetFromCanvas_DstColorType || scale != 1.0f) {
// Only decode in the simple case.
return;
Add support for multiple frames in SkCodec Add an interface to decode frames beyond the first in SkCodec, and add an implementation for SkGifCodec. Add getFrameData to SkCodec. This method reads ahead in the stream to return a vector containing meta data about each frame in the image. This is not required in order to decode frames beyond the first, but it allows a client to learn extra information: - how long the frame should be displayed - whether a frame should be blended with a prior frame, allowing the client to provide the prior frame to speed up decoding Add a new fields to SkCodec::Options: - fFrameIndex - fHasPriorFrame The API is designed so that SkCodec never caches frames. If a client wants a frame beyond the first, they specify the frame in Options.fFrameIndex. If the client does not have the frame's required frame (the frame that this frame must be blended on top of) cached, they pass false for Options.fHasPriorFrame. Unless the frame is independent, the codec will then recursively decode all frames necessary to decode fFrameIndex. If the client has the required frame cached, they can put it in the dst they pass to the codec, and the codec will only draw fFrameIndex onto it. Replace SkGifCodec's scanline decoding support with progressive decoding, and update the tests accordingly. Implement new APIs in SkGifCodec. Instead of using gif_lib, use GIFImageReader, imported from Chromium (along with its copyright headers) with the following changes: - SkGifCodec is now the client - Replace blink types - Combine GIFColorMap::buildTable and ::getTable into a method that creates and returns an SkColorTable - Input comes from an SkStream, instead of a SegmentReader. Add SkStreamBuffer, which buffers the (potentially partial) stream in order to decode progressively. (FIXME: This requires copying data that previously was read directly from the SegmentReader. Does this hurt performance? If so, can we fix it?) - Remove UMA code - Instead of reporting screen width and height to the client, allow the client to query for it - Fail earlier if the first frame AND screen have size of zero - Compute required previous frame when adding a new one - Move GIFParseQuery from GIFImageDecoder to GIFImageReader - Allow parsing up to a specific frame (to skip parsing the rest of the stream if a client only wants the first frame) - Compute whether the first frame has alpha and supports index 8, to create the SkImageInfo. This happens before reporting that the size has been decoded. Add GIFImageDecoder::haveDecodedRow to SkGifCodec, imported from Chromium (along with its copyright header), with the following changes: - Add support for sampling - Use the swizzler - Keep track of the rows decoded - Do *not* keep track of whether we've seen alpha Remove SkCodec::kOutOfOrder_SkScanlineOrder, which was only used by GIF scanline decoding. Call onRewind even if there is no stream (SkGifCodec needs to clear its decoded state so it will decode from the beginning). Add a method to SkSwizzler to access the offset into the dst, taking subsetting into account. Add a GM that animates a GIF. Add tests for the new APIs. *** Behavior changes: * Previously, we reported that an image with a subset frame and no transparent index was opaque and used the background index (if present) to fill the background. This is necessary in order to support index 8, but it does not match viewers/browsers I have seen. Examples: - Chromium and Gimp render the background transparent - Firefox, Safari, Linux Image Viewer, Safari Preview clip to the frame (for a single frame image) This CL matches Chromium's behavior and renders the background transparent. This allows us to have consistent behavior across products and simplifies the code (relative to what we would have to do to continue the old behavior on Android). It also means that we will no longer support index 8 for some GIFs. * Stop checking for GIFSTAMP - all GIFs should be either 89a or 87a. This matches Chromium. I suspect that bugs would have been reported if valid GIFs started with "GIFVER" instead of "GIF89a" or "GIF87a" (but did not decode in Chromium). *** Future work not included in this CL: * Move some checks out of haveDecodedRow, since they are the same for the entire frame e.g. - intersecting the frameRect with the full image size - whether there is a color table * Change when we write transparent pixels - In some cases, Chromium deemed this unnecessary, but I suspect it is slower than the fallback case. There will continue to be cases where we should *not* write them, but for e.g. the first pass where we have already cleared to transparent (which we may also be able to skip) writing the transparent pixels will not make anything incorrect. * Report color type and alpha type per frame - Depending on alpha values, disposal methods, frame rects, etc, subsequent frames may have different properties than the first. * Skip copies of the encoded data - We copy the encoded data in case the stream is one that cannot be rewound, so we can parse and then decode (possibly not immediately). For some input streams, this is unnecessary. - I was concerned this cause a performance regression, but on average the new code is faster than the old for the images I tested [1]. - It may cause a performance regression for Chromium, though, where we can always move back in the stream, so this should be addressed. Design doc: https://docs.google.com/a/google.com/document/d/12Qhf9T92MWfdWujQwCIjhCO3sw6pTJB5pJBwDM1T7Kc/ [1] https://docs.google.com/a/google.com/spreadsheets/d/19V-t9BfbFw5eiwBTKA1qOBkZbchjlTC5EIz6HFy-6RI/ GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=2045293002 Review-Url: https://codereview.chromium.org/2045293002
2016-10-24 16:03:26 +00:00
}
}
SkString folder;
switch (mode) {
case CodecSrc::kCodec_Mode:
folder.append("codec");
break;
case CodecSrc::kCodecZeroInit_Mode:
folder.append("codec_zero_init");
break;
case CodecSrc::kScanline_Mode:
folder.append("scanline");
break;
case CodecSrc::kStripe_Mode:
folder.append("stripe");
break;
case CodecSrc::kCroppedScanline_Mode:
folder.append("crop");
break;
case CodecSrc::kSubset_Mode:
folder.append("codec_subset");
break;
Add support for multiple frames in SkCodec Add an interface to decode frames beyond the first in SkCodec, and add an implementation for SkGifCodec. Add getFrameData to SkCodec. This method reads ahead in the stream to return a vector containing meta data about each frame in the image. This is not required in order to decode frames beyond the first, but it allows a client to learn extra information: - how long the frame should be displayed - whether a frame should be blended with a prior frame, allowing the client to provide the prior frame to speed up decoding Add a new fields to SkCodec::Options: - fFrameIndex - fHasPriorFrame The API is designed so that SkCodec never caches frames. If a client wants a frame beyond the first, they specify the frame in Options.fFrameIndex. If the client does not have the frame's required frame (the frame that this frame must be blended on top of) cached, they pass false for Options.fHasPriorFrame. Unless the frame is independent, the codec will then recursively decode all frames necessary to decode fFrameIndex. If the client has the required frame cached, they can put it in the dst they pass to the codec, and the codec will only draw fFrameIndex onto it. Replace SkGifCodec's scanline decoding support with progressive decoding, and update the tests accordingly. Implement new APIs in SkGifCodec. Instead of using gif_lib, use GIFImageReader, imported from Chromium (along with its copyright headers) with the following changes: - SkGifCodec is now the client - Replace blink types - Combine GIFColorMap::buildTable and ::getTable into a method that creates and returns an SkColorTable - Input comes from an SkStream, instead of a SegmentReader. Add SkStreamBuffer, which buffers the (potentially partial) stream in order to decode progressively. (FIXME: This requires copying data that previously was read directly from the SegmentReader. Does this hurt performance? If so, can we fix it?) - Remove UMA code - Instead of reporting screen width and height to the client, allow the client to query for it - Fail earlier if the first frame AND screen have size of zero - Compute required previous frame when adding a new one - Move GIFParseQuery from GIFImageDecoder to GIFImageReader - Allow parsing up to a specific frame (to skip parsing the rest of the stream if a client only wants the first frame) - Compute whether the first frame has alpha and supports index 8, to create the SkImageInfo. This happens before reporting that the size has been decoded. Add GIFImageDecoder::haveDecodedRow to SkGifCodec, imported from Chromium (along with its copyright header), with the following changes: - Add support for sampling - Use the swizzler - Keep track of the rows decoded - Do *not* keep track of whether we've seen alpha Remove SkCodec::kOutOfOrder_SkScanlineOrder, which was only used by GIF scanline decoding. Call onRewind even if there is no stream (SkGifCodec needs to clear its decoded state so it will decode from the beginning). Add a method to SkSwizzler to access the offset into the dst, taking subsetting into account. Add a GM that animates a GIF. Add tests for the new APIs. *** Behavior changes: * Previously, we reported that an image with a subset frame and no transparent index was opaque and used the background index (if present) to fill the background. This is necessary in order to support index 8, but it does not match viewers/browsers I have seen. Examples: - Chromium and Gimp render the background transparent - Firefox, Safari, Linux Image Viewer, Safari Preview clip to the frame (for a single frame image) This CL matches Chromium's behavior and renders the background transparent. This allows us to have consistent behavior across products and simplifies the code (relative to what we would have to do to continue the old behavior on Android). It also means that we will no longer support index 8 for some GIFs. * Stop checking for GIFSTAMP - all GIFs should be either 89a or 87a. This matches Chromium. I suspect that bugs would have been reported if valid GIFs started with "GIFVER" instead of "GIF89a" or "GIF87a" (but did not decode in Chromium). *** Future work not included in this CL: * Move some checks out of haveDecodedRow, since they are the same for the entire frame e.g. - intersecting the frameRect with the full image size - whether there is a color table * Change when we write transparent pixels - In some cases, Chromium deemed this unnecessary, but I suspect it is slower than the fallback case. There will continue to be cases where we should *not* write them, but for e.g. the first pass where we have already cleared to transparent (which we may also be able to skip) writing the transparent pixels will not make anything incorrect. * Report color type and alpha type per frame - Depending on alpha values, disposal methods, frame rects, etc, subsequent frames may have different properties than the first. * Skip copies of the encoded data - We copy the encoded data in case the stream is one that cannot be rewound, so we can parse and then decode (possibly not immediately). For some input streams, this is unnecessary. - I was concerned this cause a performance regression, but on average the new code is faster than the old for the images I tested [1]. - It may cause a performance regression for Chromium, though, where we can always move back in the stream, so this should be addressed. Design doc: https://docs.google.com/a/google.com/document/d/12Qhf9T92MWfdWujQwCIjhCO3sw6pTJB5pJBwDM1T7Kc/ [1] https://docs.google.com/a/google.com/spreadsheets/d/19V-t9BfbFw5eiwBTKA1qOBkZbchjlTC5EIz6HFy-6RI/ GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=2045293002 Review-Url: https://codereview.chromium.org/2045293002
2016-10-24 16:03:26 +00:00
case CodecSrc::kAnimated_Mode:
folder.append("codec_animated");
break;
}
switch (dstColorType) {
case CodecSrc::kGrayscale_Always_DstColorType:
folder.append("_kGray8");
break;
case CodecSrc::kNonNative8888_Always_DstColorType:
folder.append("_kNonNative");
break;
default:
break;
}
switch (dstAlphaType) {
case kPremul_SkAlphaType:
folder.append("_premul");
break;
case kUnpremul_SkAlphaType:
folder.append("_unpremul");
break;
default:
break;
}
if (1.0f != scale) {
folder.appendf("_%.3f", scale);
}
CodecSrc* src = new CodecSrc(path, mode, dstColorType, dstAlphaType, scale);
push_src("image", folder, src);
}
static void push_android_codec_src(Path path, CodecSrc::DstColorType dstColorType,
SkAlphaType dstAlphaType, int sampleSize) {
SkString folder;
folder.append("scaled_codec");
switch (dstColorType) {
case CodecSrc::kGrayscale_Always_DstColorType:
folder.append("_kGray8");
break;
case CodecSrc::kNonNative8888_Always_DstColorType:
folder.append("_kNonNative");
break;
default:
break;
}
switch (dstAlphaType) {
case kPremul_SkAlphaType:
folder.append("_premul");
break;
case kUnpremul_SkAlphaType:
folder.append("_unpremul");
break;
default:
break;
}
if (1 != sampleSize) {
folder.appendf("_%.3f", 1.0f / (float) sampleSize);
}
AndroidCodecSrc* src = new AndroidCodecSrc(path, dstColorType, dstAlphaType, sampleSize);
push_src("image", folder, src);
}
static void push_image_gen_src(Path path, ImageGenSrc::Mode mode, SkAlphaType alphaType, bool isGpu)
{
SkString folder;
switch (mode) {
case ImageGenSrc::kCodec_Mode:
folder.append("gen_codec");
break;
case ImageGenSrc::kPlatform_Mode:
folder.append("gen_platform");
break;
}
if (isGpu) {
folder.append("_gpu");
} else {
switch (alphaType) {
case kOpaque_SkAlphaType:
folder.append("_opaque");
break;
case kPremul_SkAlphaType:
folder.append("_premul");
break;
case kUnpremul_SkAlphaType:
folder.append("_unpremul");
break;
default:
break;
}
}
ImageGenSrc* src = new ImageGenSrc(path, mode, alphaType, isGpu);
push_src("image", folder, src);
}
#ifdef SK_ENABLE_ANDROID_UTILS
static void push_brd_src(Path path, CodecSrc::DstColorType dstColorType, BRDSrc::Mode mode,
uint32_t sampleSize) {
SkString folder("brd_android_codec");
switch (mode) {
case BRDSrc::kFullImage_Mode:
break;
case BRDSrc::kDivisor_Mode:
folder.append("_divisor");
break;
default:
SkASSERT(false);
return;
}
switch (dstColorType) {
case CodecSrc::kGetFromCanvas_DstColorType:
break;
case CodecSrc::kGrayscale_Always_DstColorType:
folder.append("_kGray");
break;
default:
SkASSERT(false);
return;
}
if (1 != sampleSize) {
folder.appendf("_%.3f", 1.0f / (float) sampleSize);
}
BRDSrc* src = new BRDSrc(path, mode, dstColorType, sampleSize);
push_src("image", folder, src);
}
static void push_brd_srcs(Path path, bool gray) {
if (gray) {
// Only run grayscale to one sampleSize and Mode. Though interesting
// to test grayscale, it should not reveal anything across various
// sampleSizes and Modes
// Arbitrarily choose Mode and sampleSize.
push_brd_src(path, CodecSrc::kGrayscale_Always_DstColorType,
BRDSrc::kFullImage_Mode, 2);
}
// Test on a variety of sampleSizes, making sure to include:
// - 2, 4, and 8, which are natively supported by jpeg
// - multiples of 2 which are not divisible by 4 (analogous for 4)
// - larger powers of two, since BRD clients generally use powers of 2
// We will only produce output for the larger sizes on large images.
const uint32_t sampleSizes[] = { 1, 2, 3, 4, 5, 6, 7, 8, 12, 16, 24, 32, 64 };
const BRDSrc::Mode modes[] = { BRDSrc::kFullImage_Mode, BRDSrc::kDivisor_Mode, };
for (uint32_t sampleSize : sampleSizes) {
for (BRDSrc::Mode mode : modes) {
push_brd_src(path, CodecSrc::kGetFromCanvas_DstColorType, mode, sampleSize);
}
}
}
#endif // SK_ENABLE_ANDROID_UTILS
static void push_codec_srcs(Path path) {
sk_sp<SkData> encoded(SkData::MakeFromFileName(path.c_str()));
if (!encoded) {
info("Couldn't read %s.", path.c_str());
return;
}
std::unique_ptr<SkCodec> codec = SkCodec::MakeFromData(encoded);
if (nullptr == codec) {
info("Couldn't create codec for %s.", path.c_str());
return;
}
// native scaling is only supported by WEBP and JPEG
bool supportsNativeScaling = false;
SkTArray<CodecSrc::Mode> nativeModes;
nativeModes.push_back(CodecSrc::kCodec_Mode);
nativeModes.push_back(CodecSrc::kCodecZeroInit_Mode);
switch (codec->getEncodedFormat()) {
case SkEncodedImageFormat::kJPEG:
nativeModes.push_back(CodecSrc::kScanline_Mode);
nativeModes.push_back(CodecSrc::kStripe_Mode);
nativeModes.push_back(CodecSrc::kCroppedScanline_Mode);
supportsNativeScaling = true;
break;
case SkEncodedImageFormat::kWEBP:
nativeModes.push_back(CodecSrc::kSubset_Mode);
supportsNativeScaling = true;
break;
case SkEncodedImageFormat::kDNG:
break;
default:
nativeModes.push_back(CodecSrc::kScanline_Mode);
break;
}
SkTArray<CodecSrc::DstColorType> colorTypes;
colorTypes.push_back(CodecSrc::kGetFromCanvas_DstColorType);
colorTypes.push_back(CodecSrc::kNonNative8888_Always_DstColorType);
switch (codec->getInfo().colorType()) {
case kGray_8_SkColorType:
colorTypes.push_back(CodecSrc::kGrayscale_Always_DstColorType);
break;
default:
break;
}
SkTArray<SkAlphaType> alphaModes;
alphaModes.push_back(kPremul_SkAlphaType);
if (codec->getInfo().alphaType() != kOpaque_SkAlphaType) {
alphaModes.push_back(kUnpremul_SkAlphaType);
}
for (CodecSrc::Mode mode : nativeModes) {
for (CodecSrc::DstColorType colorType : colorTypes) {
for (SkAlphaType alphaType : alphaModes) {
// Only test kCroppedScanline_Mode when the alpha type is premul. The test is
// slow and won't be interestingly different with different alpha types.
if (CodecSrc::kCroppedScanline_Mode == mode &&
kPremul_SkAlphaType != alphaType) {
continue;
}
push_codec_src(path, mode, colorType, alphaType, 1.0f);
// Skip kNonNative on different native scales. It won't be interestingly
// different.
if (supportsNativeScaling &&
CodecSrc::kNonNative8888_Always_DstColorType == colorType) {
// Native Scales
// SkJpegCodec natively supports scaling to the following:
for (auto scale : { 0.125f, 0.25f, 0.375f, 0.5f, 0.625f, 0.750f, 0.875f }) {
push_codec_src(path, mode, colorType, alphaType, scale);
}
}
}
}
}
Add support for multiple frames in SkCodec Add an interface to decode frames beyond the first in SkCodec, and add an implementation for SkGifCodec. Add getFrameData to SkCodec. This method reads ahead in the stream to return a vector containing meta data about each frame in the image. This is not required in order to decode frames beyond the first, but it allows a client to learn extra information: - how long the frame should be displayed - whether a frame should be blended with a prior frame, allowing the client to provide the prior frame to speed up decoding Add a new fields to SkCodec::Options: - fFrameIndex - fHasPriorFrame The API is designed so that SkCodec never caches frames. If a client wants a frame beyond the first, they specify the frame in Options.fFrameIndex. If the client does not have the frame's required frame (the frame that this frame must be blended on top of) cached, they pass false for Options.fHasPriorFrame. Unless the frame is independent, the codec will then recursively decode all frames necessary to decode fFrameIndex. If the client has the required frame cached, they can put it in the dst they pass to the codec, and the codec will only draw fFrameIndex onto it. Replace SkGifCodec's scanline decoding support with progressive decoding, and update the tests accordingly. Implement new APIs in SkGifCodec. Instead of using gif_lib, use GIFImageReader, imported from Chromium (along with its copyright headers) with the following changes: - SkGifCodec is now the client - Replace blink types - Combine GIFColorMap::buildTable and ::getTable into a method that creates and returns an SkColorTable - Input comes from an SkStream, instead of a SegmentReader. Add SkStreamBuffer, which buffers the (potentially partial) stream in order to decode progressively. (FIXME: This requires copying data that previously was read directly from the SegmentReader. Does this hurt performance? If so, can we fix it?) - Remove UMA code - Instead of reporting screen width and height to the client, allow the client to query for it - Fail earlier if the first frame AND screen have size of zero - Compute required previous frame when adding a new one - Move GIFParseQuery from GIFImageDecoder to GIFImageReader - Allow parsing up to a specific frame (to skip parsing the rest of the stream if a client only wants the first frame) - Compute whether the first frame has alpha and supports index 8, to create the SkImageInfo. This happens before reporting that the size has been decoded. Add GIFImageDecoder::haveDecodedRow to SkGifCodec, imported from Chromium (along with its copyright header), with the following changes: - Add support for sampling - Use the swizzler - Keep track of the rows decoded - Do *not* keep track of whether we've seen alpha Remove SkCodec::kOutOfOrder_SkScanlineOrder, which was only used by GIF scanline decoding. Call onRewind even if there is no stream (SkGifCodec needs to clear its decoded state so it will decode from the beginning). Add a method to SkSwizzler to access the offset into the dst, taking subsetting into account. Add a GM that animates a GIF. Add tests for the new APIs. *** Behavior changes: * Previously, we reported that an image with a subset frame and no transparent index was opaque and used the background index (if present) to fill the background. This is necessary in order to support index 8, but it does not match viewers/browsers I have seen. Examples: - Chromium and Gimp render the background transparent - Firefox, Safari, Linux Image Viewer, Safari Preview clip to the frame (for a single frame image) This CL matches Chromium's behavior and renders the background transparent. This allows us to have consistent behavior across products and simplifies the code (relative to what we would have to do to continue the old behavior on Android). It also means that we will no longer support index 8 for some GIFs. * Stop checking for GIFSTAMP - all GIFs should be either 89a or 87a. This matches Chromium. I suspect that bugs would have been reported if valid GIFs started with "GIFVER" instead of "GIF89a" or "GIF87a" (but did not decode in Chromium). *** Future work not included in this CL: * Move some checks out of haveDecodedRow, since they are the same for the entire frame e.g. - intersecting the frameRect with the full image size - whether there is a color table * Change when we write transparent pixels - In some cases, Chromium deemed this unnecessary, but I suspect it is slower than the fallback case. There will continue to be cases where we should *not* write them, but for e.g. the first pass where we have already cleared to transparent (which we may also be able to skip) writing the transparent pixels will not make anything incorrect. * Report color type and alpha type per frame - Depending on alpha values, disposal methods, frame rects, etc, subsequent frames may have different properties than the first. * Skip copies of the encoded data - We copy the encoded data in case the stream is one that cannot be rewound, so we can parse and then decode (possibly not immediately). For some input streams, this is unnecessary. - I was concerned this cause a performance regression, but on average the new code is faster than the old for the images I tested [1]. - It may cause a performance regression for Chromium, though, where we can always move back in the stream, so this should be addressed. Design doc: https://docs.google.com/a/google.com/document/d/12Qhf9T92MWfdWujQwCIjhCO3sw6pTJB5pJBwDM1T7Kc/ [1] https://docs.google.com/a/google.com/spreadsheets/d/19V-t9BfbFw5eiwBTKA1qOBkZbchjlTC5EIz6HFy-6RI/ GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=2045293002 Review-Url: https://codereview.chromium.org/2045293002
2016-10-24 16:03:26 +00:00
{
std::vector<SkCodec::FrameInfo> frameInfos = codec->getFrameInfo();
if (frameInfos.size() > 1) {
Add animation support to SkWebpCodec TBR=reed@google.com (No change to the public API, but changed a header file) SkWebpCodec: - Implement onGetFrameCount, onGetFrameInfo, and onGetRepetitionCount - Respect the alpha reported by libwebp. Although the spec states that it is only a hint, the libwebp encoder uses it properly. Respecting allows us to draw opaque images faster and decode them to 565. This also matches other SkCodecs (and Chromium). - onGetPixels: - Decode the frame requested, recursively decoding required frame if necessary - When blending with a prior frame, use SkRasterPipeline SkCodec: - Move check for negative index to getFrameInfo - Reset the colorXform if one is not needed SkCodecAnimation: - Add new blend enum, for WebP's (and APNG's) non-blending option SkFrameHolder: - New base classes for frames and the owner of the frames, allowing code sharing between SkWebpCodec and SkGifCodec (particularly for determining whether a frame has alpha and what frame it depends on) - When moving items from SkGIFFrameContext, use Skia conventions (i.e. int instead of unsigned) - Rename "delay time" to "duration", to match e.g. SkFrameInfo:: fDuration SkGifImageReader: - Move pieces to SkFrameHolder, and adapt to changes made in the process - Make setAlphaAndRequiredFrame (now on the base class SkFrameHolder) more general to support webp, and add support for frames that do not blend - Change SkGIFFrameContext from a struct to a class, to match how we use the distinction elsewhere (i.e. struct is a small object with public fields) - Rework hasTransparentPixel (now hasTransparency, since it returns true in some cases where there is not a transparent pixel) to better fit with the modified setAlphaAndRequiredFrame. Also be more consistent when there is no transparent pixel but no color map. - Simplify an if condition that was previously simplified in 2d61e717 but accidentally got reverted in a4db9be6 CodecAnimTest: - Test new animated webp files - Rearrange the test to more cleanly print alpha type mismatches for the first frame resources: - webp-animated.webp - animated webp from Chromium - blendBG.webp - new webp file using bits of webp-animated-semitransparent4.webp from Chromium - tests required frame and alpha when using the non-blending mode - frames have the following properties: - Frame 0: no alpha, fills screen - Frame 1: alpha, fills screen - Frame 2: no alpha, fills screen - Frame 3: alpha, fills screen, blendBG - Frame 4: no alpha, fills screen, blendBG - Frame 5: alpha, blendBG - Frame 6: covers 4, has alpha, blendBG - also used to test decoding to 565 if the new frame data has alpha but blends onto an opaque frame DM.cpp: - Test animated images to non-native 8888 and unpremul DMSrcSink.cpp: - Do not test non-native 8888 decodes to f16 dst - Test unpremul decodes to f16 - Copy a frame of an animated image prior to drawing, since in unpremul mode, the DM code will premultiply first. Bug: skia: 3315 Change-Id: I4e55ae2ee5bc095b37a743bdcfac644be603b980 Reviewed-on: https://skia-review.googlesource.com/16707 Commit-Queue: Mike Reed <reed@google.com> Reviewed-by: Mike Reed <reed@google.com> Reviewed-by: Leon Scroggins <scroggo@google.com> Reviewed-by: Matt Sarett <msarett@google.com>
2017-05-23 13:37:21 +00:00
for (auto dstCT : { CodecSrc::kNonNative8888_Always_DstColorType,
CodecSrc::kGetFromCanvas_DstColorType }) {
for (auto at : { kUnpremul_SkAlphaType, kPremul_SkAlphaType }) {
push_codec_src(path, CodecSrc::kAnimated_Mode, dstCT, at, 1.0f);
}
}
Reland "SkAndroidCodec: Support decoding all frames" This is a reland of fc4fdc5b25f448dd9c2cd4e445561a840ce8514b Original change's description: > SkAndroidCodec: Support decoding all frames > > Bug: b/160984428 > Bug: b/163595585 > > Add support to SkAndroidCodec for decoding all frames with an > fSampleSize, so that an entire animation can be decoded to a smaller > size. > > dm/: > - Test scaled + animated decodes > > SkAndroidCodec: > - Make AndroidOptions inherit from SkCodec::Options. This allows > SkAndroidCodec to use fFrameIndex. (It also combines the two versions > of fSubset, which is now const for both.) > - Respect fFrameIndex, and call SkCodec::handleFrameIndex to decode > the required frame. > - Disallow decoding with kRespect + fFrameIndex > 0 if there is a > non-default orientation. As currently written (except without > disabling this combination), SkPixmapPriv::Orient would draw the new > portion of the frame on top of uninitialized pixels, instead of the > prior frame. This could be fixed by > - If SkAndroidCodec needs to decode the required frame, it could do so > without applying the orientation, then decode fFrameIndex, and then > apply the orientation. > - If the client provided the required frame, SkAndroidCodec would need > to un-apply the orientation to get the proper starting state, then > decode and apply. > I think it is simpler to force the client to handle the orientation > externally. > > SkCodec: > - Allow SkAndroidCodec to call its private method handleFrameIndex. This > method handles decoding a required frame, if necessary. When called by > SkAndroidCodec, it now uses the SkAndroidCodec to check for/decode the > required frame, so that it will scale properly. > - Call rewindIfNeeded inside handleFrameIndex. handleFrameIndex calls a > virtual method which may set some state (e.g. in SkJpegCodec). Without > this change, that state would be reset by rewindIfNeeded. > - Simplify handling a kRestoreBGColor frame. Whether provided or not, > take the same path to calling zero_rect. > - Updates to zero_rect: > - Intersect after scaling, which will also check for empty. > - Round out instead of in - this ensures we don't under-erase > - Use kFill_ScaleToFit, which better matches the intent. > > Change-Id: Ibe1951980a0dca8f5b7b1f20192432d395681683 > Reviewed-on: https://skia-review.googlesource.com/c/skia/+/333225 > Commit-Queue: Leon Scroggins <scroggo@google.com> > Reviewed-by: Derek Sollenberger <djsollen@google.com> Bug: b/160984428 Bug: b/163595585 Change-Id: I7c1e79e0f92c75b4840eef65c8fc2b8497189e81 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/334842 Auto-Submit: Leon Scroggins <scroggo@google.com> Commit-Queue: Derek Sollenberger <djsollen@google.com> Reviewed-by: Derek Sollenberger <djsollen@google.com>
2020-11-09 19:18:12 +00:00
for (float scale : { .5f, .33f }) {
push_codec_src(path, CodecSrc::kAnimated_Mode, CodecSrc::kGetFromCanvas_DstColorType,
kPremul_SkAlphaType, scale);
}
Add support for multiple frames in SkCodec Add an interface to decode frames beyond the first in SkCodec, and add an implementation for SkGifCodec. Add getFrameData to SkCodec. This method reads ahead in the stream to return a vector containing meta data about each frame in the image. This is not required in order to decode frames beyond the first, but it allows a client to learn extra information: - how long the frame should be displayed - whether a frame should be blended with a prior frame, allowing the client to provide the prior frame to speed up decoding Add a new fields to SkCodec::Options: - fFrameIndex - fHasPriorFrame The API is designed so that SkCodec never caches frames. If a client wants a frame beyond the first, they specify the frame in Options.fFrameIndex. If the client does not have the frame's required frame (the frame that this frame must be blended on top of) cached, they pass false for Options.fHasPriorFrame. Unless the frame is independent, the codec will then recursively decode all frames necessary to decode fFrameIndex. If the client has the required frame cached, they can put it in the dst they pass to the codec, and the codec will only draw fFrameIndex onto it. Replace SkGifCodec's scanline decoding support with progressive decoding, and update the tests accordingly. Implement new APIs in SkGifCodec. Instead of using gif_lib, use GIFImageReader, imported from Chromium (along with its copyright headers) with the following changes: - SkGifCodec is now the client - Replace blink types - Combine GIFColorMap::buildTable and ::getTable into a method that creates and returns an SkColorTable - Input comes from an SkStream, instead of a SegmentReader. Add SkStreamBuffer, which buffers the (potentially partial) stream in order to decode progressively. (FIXME: This requires copying data that previously was read directly from the SegmentReader. Does this hurt performance? If so, can we fix it?) - Remove UMA code - Instead of reporting screen width and height to the client, allow the client to query for it - Fail earlier if the first frame AND screen have size of zero - Compute required previous frame when adding a new one - Move GIFParseQuery from GIFImageDecoder to GIFImageReader - Allow parsing up to a specific frame (to skip parsing the rest of the stream if a client only wants the first frame) - Compute whether the first frame has alpha and supports index 8, to create the SkImageInfo. This happens before reporting that the size has been decoded. Add GIFImageDecoder::haveDecodedRow to SkGifCodec, imported from Chromium (along with its copyright header), with the following changes: - Add support for sampling - Use the swizzler - Keep track of the rows decoded - Do *not* keep track of whether we've seen alpha Remove SkCodec::kOutOfOrder_SkScanlineOrder, which was only used by GIF scanline decoding. Call onRewind even if there is no stream (SkGifCodec needs to clear its decoded state so it will decode from the beginning). Add a method to SkSwizzler to access the offset into the dst, taking subsetting into account. Add a GM that animates a GIF. Add tests for the new APIs. *** Behavior changes: * Previously, we reported that an image with a subset frame and no transparent index was opaque and used the background index (if present) to fill the background. This is necessary in order to support index 8, but it does not match viewers/browsers I have seen. Examples: - Chromium and Gimp render the background transparent - Firefox, Safari, Linux Image Viewer, Safari Preview clip to the frame (for a single frame image) This CL matches Chromium's behavior and renders the background transparent. This allows us to have consistent behavior across products and simplifies the code (relative to what we would have to do to continue the old behavior on Android). It also means that we will no longer support index 8 for some GIFs. * Stop checking for GIFSTAMP - all GIFs should be either 89a or 87a. This matches Chromium. I suspect that bugs would have been reported if valid GIFs started with "GIFVER" instead of "GIF89a" or "GIF87a" (but did not decode in Chromium). *** Future work not included in this CL: * Move some checks out of haveDecodedRow, since they are the same for the entire frame e.g. - intersecting the frameRect with the full image size - whether there is a color table * Change when we write transparent pixels - In some cases, Chromium deemed this unnecessary, but I suspect it is slower than the fallback case. There will continue to be cases where we should *not* write them, but for e.g. the first pass where we have already cleared to transparent (which we may also be able to skip) writing the transparent pixels will not make anything incorrect. * Report color type and alpha type per frame - Depending on alpha values, disposal methods, frame rects, etc, subsequent frames may have different properties than the first. * Skip copies of the encoded data - We copy the encoded data in case the stream is one that cannot be rewound, so we can parse and then decode (possibly not immediately). For some input streams, this is unnecessary. - I was concerned this cause a performance regression, but on average the new code is faster than the old for the images I tested [1]. - It may cause a performance regression for Chromium, though, where we can always move back in the stream, so this should be addressed. Design doc: https://docs.google.com/a/google.com/document/d/12Qhf9T92MWfdWujQwCIjhCO3sw6pTJB5pJBwDM1T7Kc/ [1] https://docs.google.com/a/google.com/spreadsheets/d/19V-t9BfbFw5eiwBTKA1qOBkZbchjlTC5EIz6HFy-6RI/ GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=2045293002 Review-Url: https://codereview.chromium.org/2045293002
2016-10-24 16:03:26 +00:00
}
}
if (FLAGS_simpleCodec) {
return;
}
const int sampleSizes[] = { 1, 2, 3, 4, 5, 6, 7, 8 };
for (int sampleSize : sampleSizes) {
for (CodecSrc::DstColorType colorType : colorTypes) {
for (SkAlphaType alphaType : alphaModes) {
// We can exercise all of the kNonNative support code in the swizzler with just a
// few sample sizes. Skip the rest.
if (CodecSrc::kNonNative8888_Always_DstColorType == colorType && sampleSize > 3) {
continue;
}
push_android_codec_src(path, colorType, alphaType, sampleSize);
}
}
}
const char* ext = strrchr(path.c_str(), '.');
if (ext) {
ext++;
static const char* const rawExts[] = {
"arw", "cr2", "dng", "nef", "nrw", "orf", "raf", "rw2", "pef", "srw",
"ARW", "CR2", "DNG", "NEF", "NRW", "ORF", "RAF", "RW2", "PEF", "SRW",
};
for (const char* rawExt : rawExts) {
if (0 == strcmp(rawExt, ext)) {
// RAW is not supported by image generator (skbug.com/5079) or BRD.
return;
}
}
#ifdef SK_ENABLE_ANDROID_UTILS
static const char* const brdExts[] = {
"jpg", "jpeg", "png", "webp",
"JPG", "JPEG", "PNG", "WEBP",
};
for (const char* brdExt : brdExts) {
if (0 == strcmp(brdExt, ext)) {
bool gray = codec->getInfo().colorType() == kGray_8_SkColorType;
push_brd_srcs(path, gray);
break;
}
}
#endif
}
// Push image generator GPU test.
push_image_gen_src(path, ImageGenSrc::kCodec_Mode, codec->getInfo().alphaType(), true);
// Push image generator CPU tests.
for (SkAlphaType alphaType : alphaModes) {
push_image_gen_src(path, ImageGenSrc::kCodec_Mode, alphaType, false);
#if defined(SK_BUILD_FOR_MAC) || defined(SK_BUILD_FOR_IOS)
if (SkEncodedImageFormat::kWEBP != codec->getEncodedFormat() &&
SkEncodedImageFormat::kWBMP != codec->getEncodedFormat() &&
kUnpremul_SkAlphaType != alphaType)
{
push_image_gen_src(path, ImageGenSrc::kPlatform_Mode, alphaType, false);
}
#elif defined(SK_BUILD_FOR_WIN)
if (SkEncodedImageFormat::kWEBP != codec->getEncodedFormat() &&
SkEncodedImageFormat::kWBMP != codec->getEncodedFormat())
{
push_image_gen_src(path, ImageGenSrc::kPlatform_Mode, alphaType, false);
}
#elif defined(SK_ENABLE_NDK_IMAGES)
push_image_gen_src(path, ImageGenSrc::kPlatform_Mode, alphaType, false);
#endif
}
}
template <typename T>
void gather_file_srcs(const CommandLineFlags::StringArray& flags,
const char* ext,
const char* src_name = nullptr) {
if (!src_name) {
// With the exception of Lottie files, the source name is the extension.
src_name = ext;
}
for (int i = 0; i < flags.count(); i++) {
const char* path = flags[i];
if (sk_isdir(path)) {
SkOSFile::Iter it(path, ext);
for (SkString file; it.next(&file); ) {
push_src(src_name, "", new T(SkOSPath::Join(path, file.c_str())));
}
} else {
push_src(src_name, "", new T(path));
}
}
}
static bool gather_srcs() {
for (skiagm::GMFactory f : skiagm::GMRegistry::Range()) {
push_src("gm", "", new GMSrc(f));
}
gather_file_srcs<SKPSrc>(FLAGS_skps, "skp");
gather_file_srcs<MSKPSrc>(FLAGS_mskps, "mskp");
#if defined(SK_ENABLE_SKOTTIE)
gather_file_srcs<SkottieSrc>(FLAGS_lotties, "json", "lottie");
#endif
#if defined(SK_ENABLE_SKRIVE)
gather_file_srcs<SkRiveSrc>(FLAGS_rives, "flr", "rive");
#endif
#if defined(SK_ENABLE_SVG)
gather_file_srcs<SVGSrc>(FLAGS_svgs, "svg");
#endif
if (!FLAGS_bisect.isEmpty()) {
// An empty l/r trail string will draw all the paths.
push_src("bisect", "",
new BisectSrc(FLAGS_bisect[0], FLAGS_bisect.count() > 1 ? FLAGS_bisect[1] : ""));
}
SkTArray<SkString> images;
if (!CommonFlags::CollectImages(FLAGS_images, &images)) {
return false;
}
for (const SkString& image : images) {
push_codec_srcs(image);
}
SkTArray<SkString> colorImages;
if (!CommonFlags::CollectImages(FLAGS_colorImages, &colorImages)) {
return false;
}
for (const SkString& colorImage : colorImages) {
push_src("colorImage", "decode_native", new ColorCodecSrc(colorImage, false));
push_src("colorImage", "decode_to_dst", new ColorCodecSrc(colorImage, true));
}
return true;
}
Add config options to run different GPU APIs to dm and nanobench Add extended config specification form that can be used to run different gpu backend with different APIs. The configs can be specified with the form: gpu(api=string,dit=bool,nvpr=bool,samples=int) This replaces and removes the --gpuAPI flag. All existing configs should still work. Adds following documentation: out/Debug/dm --help config Flags: --config: type: string default: 565 8888 gpu nonrendering Options: 565 8888 debug gpu gpudebug gpudft gpunull msaa16 msaa4 nonrendering null nullgpu nvprmsaa16 nvprmsaa4 pdf pdf_poppler skp svg xps or use extended form 'backend(option=value,...)'. Extended form: 'backend(option=value,...)' Possible backends and options: gpu(api=string,dit=bool,nvpr=bool,samples=int) GPU backend api type: string default: native. Select graphics API to use with gpu backend. Options: native Use platform default OpenGL or OpenGL ES backend. gl Use OpenGL. gles Use OpenGL ES. debug Use debug OpenGL. null Use null OpenGL. dit type: bool default: false. Use device independent text. nvpr type: bool default: false. Use NV_path_rendering OpenGL and OpenGL ES extension. samples type: int default: 0. Use multisampling with N samples. Predefined configs: gpu = gpu() msaa4 = gpu(samples=4) msaa16 = gpu(samples=16) nvprmsaa4 = gpu(nvpr=true,samples=4) nvprmsaa16 = gpu(nvpr=true,samples=16) gpudft = gpu(dit=true) gpudebug = gpu(api=debug) gpunull = gpu(api=null) debug = gpu(api=debug) nullgpu = gpu(api=null) BUG=skia:2992 Committed: https://skia.googlesource.com/skia/+/e13ca329fca4c28cf4e078561f591ab27b743d23 GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1490113005 Committed: https://skia.googlesource.com/skia/+/c8b4336444e7b90382e04e33665fb3b8490b825b Committed: https://skia.googlesource.com/skia/+/9ebc3f0ee6db215dde461dc4777d85988cf272dd Review URL: https://codereview.chromium.org/1490113005
2015-12-23 09:33:00 +00:00
static void push_sink(const SkCommandLineConfig& config, Sink* s) {
std::unique_ptr<Sink> sink(s);
Add config options to run different GPU APIs to dm and nanobench Add extended config specification form that can be used to run different gpu backend with different APIs. The configs can be specified with the form: gpu(api=string,dit=bool,nvpr=bool,samples=int) This replaces and removes the --gpuAPI flag. All existing configs should still work. Adds following documentation: out/Debug/dm --help config Flags: --config: type: string default: 565 8888 gpu nonrendering Options: 565 8888 debug gpu gpudebug gpudft gpunull msaa16 msaa4 nonrendering null nullgpu nvprmsaa16 nvprmsaa4 pdf pdf_poppler skp svg xps or use extended form 'backend(option=value,...)'. Extended form: 'backend(option=value,...)' Possible backends and options: gpu(api=string,dit=bool,nvpr=bool,samples=int) GPU backend api type: string default: native. Select graphics API to use with gpu backend. Options: native Use platform default OpenGL or OpenGL ES backend. gl Use OpenGL. gles Use OpenGL ES. debug Use debug OpenGL. null Use null OpenGL. dit type: bool default: false. Use device independent text. nvpr type: bool default: false. Use NV_path_rendering OpenGL and OpenGL ES extension. samples type: int default: 0. Use multisampling with N samples. Predefined configs: gpu = gpu() msaa4 = gpu(samples=4) msaa16 = gpu(samples=16) nvprmsaa4 = gpu(nvpr=true,samples=4) nvprmsaa16 = gpu(nvpr=true,samples=16) gpudft = gpu(dit=true) gpudebug = gpu(api=debug) gpunull = gpu(api=null) debug = gpu(api=debug) nullgpu = gpu(api=null) BUG=skia:2992 Committed: https://skia.googlesource.com/skia/+/e13ca329fca4c28cf4e078561f591ab27b743d23 GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1490113005 Committed: https://skia.googlesource.com/skia/+/c8b4336444e7b90382e04e33665fb3b8490b825b Committed: https://skia.googlesource.com/skia/+/9ebc3f0ee6db215dde461dc4777d85988cf272dd Review URL: https://codereview.chromium.org/1490113005
2015-12-23 09:33:00 +00:00
// Try a simple Src as a canary. If it fails, skip this sink.
struct : public Src {
Result draw(GrDirectContext*, SkCanvas* c) const override {
c->drawRect(SkRect::MakeWH(1,1), SkPaint());
Replace DM:Error with DM::Result. Initially, Error was written with the intent that an empty string meant Ok and anything else meant Fatal. This made things simple with implicit constructors from strings. With the introduction of Nonfatal the state was now tied up with an additional boolean. Now the empty string meant Ok and causes the new boolean to be ignored, or at least that is the way it was used since Error didn't actually enforce that itself. This leads to GMs which return kSkip but don't set the message to not be skipped. This could be fixed in several ways. The first would be for the GMSrc to notice that a GM had returned kSkip with an empty message and create the Error::Nonfatal with a non-empty message. This has the downside of being some extra unexpected complexity and doesn't prevent similar issues from arising in the future. The second would be to change how DM interprets the Error, and if the non-fatal bit is set treat that as a sign to skip, even if the message is empty. This fixes the stated issue, but doesn't fix the issue where a GM can return kFail but also leave the message empty. This could again be fixed by either modifying GMSrc::draw or GM::drawContent, but this also seems a bit brittle in not preventing this from happening again in the future. So this replaces Error with Result, which makes the status orthogonal to the message. It does lose the automatic conversion from string, but by being able to wrap the many uses of SkStringPrintf the explicit nature doesn't add much additional noise to the more complex failure reports. Change-Id: Ibf48b67faa09a91a4a9d792d204bd9810b441c6c Reviewed-on: https://skia-review.googlesource.com/c/skia/+/270362 Commit-Queue: Ben Wagner <bungeman@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-02-12 16:18:46 +00:00
return Result::Ok();
}
SkISize size() const override { return SkISize::Make(16, 16); }
Name name() const override { return "justOneRect"; }
} justOneRect;
SkBitmap bitmap;
SkDynamicMemoryWStream stream;
SkString log;
Replace DM:Error with DM::Result. Initially, Error was written with the intent that an empty string meant Ok and anything else meant Fatal. This made things simple with implicit constructors from strings. With the introduction of Nonfatal the state was now tied up with an additional boolean. Now the empty string meant Ok and causes the new boolean to be ignored, or at least that is the way it was used since Error didn't actually enforce that itself. This leads to GMs which return kSkip but don't set the message to not be skipped. This could be fixed in several ways. The first would be for the GMSrc to notice that a GM had returned kSkip with an empty message and create the Error::Nonfatal with a non-empty message. This has the downside of being some extra unexpected complexity and doesn't prevent similar issues from arising in the future. The second would be to change how DM interprets the Error, and if the non-fatal bit is set treat that as a sign to skip, even if the message is empty. This fixes the stated issue, but doesn't fix the issue where a GM can return kFail but also leave the message empty. This could again be fixed by either modifying GMSrc::draw or GM::drawContent, but this also seems a bit brittle in not preventing this from happening again in the future. So this replaces Error with Result, which makes the status orthogonal to the message. It does lose the automatic conversion from string, but by being able to wrap the many uses of SkStringPrintf the explicit nature doesn't add much additional noise to the more complex failure reports. Change-Id: Ibf48b67faa09a91a4a9d792d204bd9810b441c6c Reviewed-on: https://skia-review.googlesource.com/c/skia/+/270362 Commit-Queue: Ben Wagner <bungeman@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-02-12 16:18:46 +00:00
Result result = sink->draw(justOneRect, &bitmap, &stream, &log);
if (result.isFatal()) {
info("Could not run %s: %s\n", config.getTag().c_str(), result.c_str());
exit(1);
}
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
TaggedSink& ts = gSinks->push_back();
ts.reset(sink.release());
Add config options to run different GPU APIs to dm and nanobench Add extended config specification form that can be used to run different gpu backend with different APIs. The configs can be specified with the form: gpu(api=string,dit=bool,nvpr=bool,samples=int) This replaces and removes the --gpuAPI flag. All existing configs should still work. Adds following documentation: out/Debug/dm --help config Flags: --config: type: string default: 565 8888 gpu nonrendering Options: 565 8888 debug gpu gpudebug gpudft gpunull msaa16 msaa4 nonrendering null nullgpu nvprmsaa16 nvprmsaa4 pdf pdf_poppler skp svg xps or use extended form 'backend(option=value,...)'. Extended form: 'backend(option=value,...)' Possible backends and options: gpu(api=string,dit=bool,nvpr=bool,samples=int) GPU backend api type: string default: native. Select graphics API to use with gpu backend. Options: native Use platform default OpenGL or OpenGL ES backend. gl Use OpenGL. gles Use OpenGL ES. debug Use debug OpenGL. null Use null OpenGL. dit type: bool default: false. Use device independent text. nvpr type: bool default: false. Use NV_path_rendering OpenGL and OpenGL ES extension. samples type: int default: 0. Use multisampling with N samples. Predefined configs: gpu = gpu() msaa4 = gpu(samples=4) msaa16 = gpu(samples=16) nvprmsaa4 = gpu(nvpr=true,samples=4) nvprmsaa16 = gpu(nvpr=true,samples=16) gpudft = gpu(dit=true) gpudebug = gpu(api=debug) gpunull = gpu(api=null) debug = gpu(api=debug) nullgpu = gpu(api=null) BUG=skia:2992 Committed: https://skia.googlesource.com/skia/+/e13ca329fca4c28cf4e078561f591ab27b743d23 GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1490113005 Committed: https://skia.googlesource.com/skia/+/c8b4336444e7b90382e04e33665fb3b8490b825b Committed: https://skia.googlesource.com/skia/+/9ebc3f0ee6db215dde461dc4777d85988cf272dd Review URL: https://codereview.chromium.org/1490113005
2015-12-23 09:33:00 +00:00
ts.tag = config.getTag();
}
static Sink* create_sink(const GrContextOptions& grCtxOptions, const SkCommandLineConfig* config) {
if (FLAGS_gpu) {
Add config options to run different GPU APIs to dm and nanobench Add extended config specification form that can be used to run different gpu backend with different APIs. The configs can be specified with the form: gpu(api=string,dit=bool,nvpr=bool,samples=int) This replaces and removes the --gpuAPI flag. All existing configs should still work. Adds following documentation: out/Debug/dm --help config Flags: --config: type: string default: 565 8888 gpu nonrendering Options: 565 8888 debug gpu gpudebug gpudft gpunull msaa16 msaa4 nonrendering null nullgpu nvprmsaa16 nvprmsaa4 pdf pdf_poppler skp svg xps or use extended form 'backend(option=value,...)'. Extended form: 'backend(option=value,...)' Possible backends and options: gpu(api=string,dit=bool,nvpr=bool,samples=int) GPU backend api type: string default: native. Select graphics API to use with gpu backend. Options: native Use platform default OpenGL or OpenGL ES backend. gl Use OpenGL. gles Use OpenGL ES. debug Use debug OpenGL. null Use null OpenGL. dit type: bool default: false. Use device independent text. nvpr type: bool default: false. Use NV_path_rendering OpenGL and OpenGL ES extension. samples type: int default: 0. Use multisampling with N samples. Predefined configs: gpu = gpu() msaa4 = gpu(samples=4) msaa16 = gpu(samples=16) nvprmsaa4 = gpu(nvpr=true,samples=4) nvprmsaa16 = gpu(nvpr=true,samples=16) gpudft = gpu(dit=true) gpudebug = gpu(api=debug) gpunull = gpu(api=null) debug = gpu(api=debug) nullgpu = gpu(api=null) BUG=skia:2992 Committed: https://skia.googlesource.com/skia/+/e13ca329fca4c28cf4e078561f591ab27b743d23 GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1490113005 Committed: https://skia.googlesource.com/skia/+/c8b4336444e7b90382e04e33665fb3b8490b825b Committed: https://skia.googlesource.com/skia/+/9ebc3f0ee6db215dde461dc4777d85988cf272dd Review URL: https://codereview.chromium.org/1490113005
2015-12-23 09:33:00 +00:00
if (const SkCommandLineConfigGpu* gpuConfig = config->asConfigGpu()) {
GrContextFactory testFactory(grCtxOptions);
if (!testFactory.get(gpuConfig->getContextType(), gpuConfig->getContextOverrides())) {
info("WARNING: can not create GPU context for config '%s'. "
"GM tests will be skipped.\n", gpuConfig->getTag().c_str());
Add config options to run different GPU APIs to dm and nanobench Add extended config specification form that can be used to run different gpu backend with different APIs. The configs can be specified with the form: gpu(api=string,dit=bool,nvpr=bool,samples=int) This replaces and removes the --gpuAPI flag. All existing configs should still work. Adds following documentation: out/Debug/dm --help config Flags: --config: type: string default: 565 8888 gpu nonrendering Options: 565 8888 debug gpu gpudebug gpudft gpunull msaa16 msaa4 nonrendering null nullgpu nvprmsaa16 nvprmsaa4 pdf pdf_poppler skp svg xps or use extended form 'backend(option=value,...)'. Extended form: 'backend(option=value,...)' Possible backends and options: gpu(api=string,dit=bool,nvpr=bool,samples=int) GPU backend api type: string default: native. Select graphics API to use with gpu backend. Options: native Use platform default OpenGL or OpenGL ES backend. gl Use OpenGL. gles Use OpenGL ES. debug Use debug OpenGL. null Use null OpenGL. dit type: bool default: false. Use device independent text. nvpr type: bool default: false. Use NV_path_rendering OpenGL and OpenGL ES extension. samples type: int default: 0. Use multisampling with N samples. Predefined configs: gpu = gpu() msaa4 = gpu(samples=4) msaa16 = gpu(samples=16) nvprmsaa4 = gpu(nvpr=true,samples=4) nvprmsaa16 = gpu(nvpr=true,samples=16) gpudft = gpu(dit=true) gpudebug = gpu(api=debug) gpunull = gpu(api=null) debug = gpu(api=debug) nullgpu = gpu(api=null) BUG=skia:2992 Committed: https://skia.googlesource.com/skia/+/e13ca329fca4c28cf4e078561f591ab27b743d23 GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1490113005 Committed: https://skia.googlesource.com/skia/+/c8b4336444e7b90382e04e33665fb3b8490b825b Committed: https://skia.googlesource.com/skia/+/9ebc3f0ee6db215dde461dc4777d85988cf272dd Review URL: https://codereview.chromium.org/1490113005
2015-12-23 09:33:00 +00:00
return nullptr;
}
if (gpuConfig->getTestThreading()) {
SkASSERT(!gpuConfig->getTestPersistentCache());
return new GPUThreadTestingSink(gpuConfig, grCtxOptions);
} else if (gpuConfig->getTestPersistentCache()) {
return new GPUPersistentCacheTestingSink(gpuConfig, grCtxOptions);
Add support for pre-compiling cached SkSL shaders The client can do a test run of their application with a persistent cache set to SkSL mode. They store the key and data blobs that are produced. Ship those blobs with the application. At startup, call GrContext::precompileShader for each key/data pair. This compiles the shaders, and stores the GL program ID, plus a small amount of metadata in our runtime program cache. Caveats: * Currently only implemented for the GL backend. Other backends will require more metadata to do any useful amount of work. Metal may need a more drastic workflow change, involving offline compilation of the shaders. * Currently only implemented for cached SkSL (not GLSL or program binaries). Supporting other formats again requires more metadata, and the cached shaders become increasingly specialized to GPU and driver versions. * Reusing the cached SkSL on different hardware is not supported. Many driver workarounds are implemented in the SkSL -> GLSL transformation, but some are higher level. Limiting device variance by artificially hiding extensions may help, but there are no guarantees. * The 'gltestprecompile' DM config exercises this code similarly to 'gltestpersistentcache', ensuring that results are visually identical when precompiling, and that no cache misses occur after precompiling. Change-Id: Id314c5d5f5a58fe503a0505a613bd4a540cc3589 Reviewed-on: https://skia-review.googlesource.com/c/skia/+/239438 Reviewed-by: Greg Daniel <egdaniel@google.com> Reviewed-by: Brian Salomon <bsalomon@google.com> Commit-Queue: Brian Osman <brianosman@google.com>
2019-09-06 18:42:43 +00:00
} else if (gpuConfig->getTestPrecompile()) {
return new GPUPrecompileTestingSink(gpuConfig, grCtxOptions);
} else if (gpuConfig->getUseDDLSink()) {
return new GPUDDLSink(gpuConfig, grCtxOptions);
} else if (gpuConfig->getOOPRish()) {
return new GPUOOPRSink(gpuConfig, grCtxOptions);
} else {
return new GPUSink(gpuConfig, grCtxOptions);
}
Add config options to run different GPU APIs to dm and nanobench Add extended config specification form that can be used to run different gpu backend with different APIs. The configs can be specified with the form: gpu(api=string,dit=bool,nvpr=bool,samples=int) This replaces and removes the --gpuAPI flag. All existing configs should still work. Adds following documentation: out/Debug/dm --help config Flags: --config: type: string default: 565 8888 gpu nonrendering Options: 565 8888 debug gpu gpudebug gpudft gpunull msaa16 msaa4 nonrendering null nullgpu nvprmsaa16 nvprmsaa4 pdf pdf_poppler skp svg xps or use extended form 'backend(option=value,...)'. Extended form: 'backend(option=value,...)' Possible backends and options: gpu(api=string,dit=bool,nvpr=bool,samples=int) GPU backend api type: string default: native. Select graphics API to use with gpu backend. Options: native Use platform default OpenGL or OpenGL ES backend. gl Use OpenGL. gles Use OpenGL ES. debug Use debug OpenGL. null Use null OpenGL. dit type: bool default: false. Use device independent text. nvpr type: bool default: false. Use NV_path_rendering OpenGL and OpenGL ES extension. samples type: int default: 0. Use multisampling with N samples. Predefined configs: gpu = gpu() msaa4 = gpu(samples=4) msaa16 = gpu(samples=16) nvprmsaa4 = gpu(nvpr=true,samples=4) nvprmsaa16 = gpu(nvpr=true,samples=16) gpudft = gpu(dit=true) gpudebug = gpu(api=debug) gpunull = gpu(api=null) debug = gpu(api=debug) nullgpu = gpu(api=null) BUG=skia:2992 Committed: https://skia.googlesource.com/skia/+/e13ca329fca4c28cf4e078561f591ab27b743d23 GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1490113005 Committed: https://skia.googlesource.com/skia/+/c8b4336444e7b90382e04e33665fb3b8490b825b Committed: https://skia.googlesource.com/skia/+/9ebc3f0ee6db215dde461dc4777d85988cf272dd Review URL: https://codereview.chromium.org/1490113005
2015-12-23 09:33:00 +00:00
}
Revert of Add config options to run different GPU APIs to dm and nanobench (patchset #21 id:400001 of https://codereview.chromium.org/1490113005/ ) Reason for revert: The Test-Win8-MSVC-ShuttleB-GPU-HD4600-x86_64-Debug builder fails after this CL. Links to specific builds: http://build.chromium.org/p/client.skia/builders/Test-Win8-MSVC-ShuttleB-GPU-HD4600-x86_64-Debug/builds/1689 http://build.chromium.org/p/client.skia/builders/Test-Win8-MSVC-ShuttleB-GPU-HD4600-x86_64-Debug/builds/1690 http://build.chromium.org/p/client.skia/builders/Test-Win8-MSVC-ShuttleB-GPU-HD4600-x86_64-Debug/builds/1691 Original issue's description: > Add config options to run different GPU APIs to dm and nanobench > > Add extended config specification form that can be used to run different > gpu backend with different APIs. > > The configs can be specified with the form: > gpu(api=string,dit=bool,nvpr=bool,samples=int) > > This replaces and removes the --gpuAPI flag. > > All existing configs should still work. > > Adds following documentation: > > out/Debug/dm --help config > > Flags: > --config: type: string default: 565 8888 gpu nonrendering > Options: 565 8888 debug gpu gpudebug gpudft gpunull msaa16 msaa4 > nonrendering null nullgpu nvprmsaa16 nvprmsaa4 pdf pdf_poppler skp svg > xps or use extended form 'backend(option=value,...)'. > > Extended form: 'backend(option=value,...)' > > Possible backends and options: > > gpu(api=string,dit=bool,nvpr=bool,samples=int) GPU backend > api type: string default: native. > Select graphics API to use with gpu backend. > Options: > native Use platform default OpenGL or OpenGL ES backend. > gl Use OpenGL. > gles Use OpenGL ES. > debug Use debug OpenGL. > null Use null OpenGL. > dit type: bool default: false. > Use device independent text. > nvpr type: bool default: false. > Use NV_path_rendering OpenGL and OpenGL ES extension. > samples type: int default: 0. > Use multisampling with N samples. > > Predefined configs: > > gpu = gpu() > msaa4 = gpu(samples=4) > msaa16 = gpu(samples=16) > nvprmsaa4 = gpu(nvpr=true,samples=4) > nvprmsaa16 = gpu(nvpr=true,samples=16) > gpudft = gpu(dit=true) > gpudebug = gpu(api=debug) > gpunull = gpu(api=null) > debug = gpu(api=debug) > nullgpu = gpu(api=null) > > BUG=skia:2992 > > Committed: https://skia.googlesource.com/skia/+/e13ca329fca4c28cf4e078561f591ab27b743d23 > GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1490113005 > > Committed: https://skia.googlesource.com/skia/+/c8b4336444e7b90382e04e33665fb3b8490b825b > > Committed: https://skia.googlesource.com/skia/+/9ebc3f0ee6db215dde461dc4777d85988cf272dd TBR=mtklein@google.com,bsalomon@google.com,joshualitt@google.com,scroggo@google.com,kkinnunen@nvidia.com NOPRESUBMIT=true NOTREECHECKS=true NOTRY=true BUG=skia:2992 Review URL: https://codereview.chromium.org/1548683002
2015-12-22 18:22:26 +00:00
}
#ifdef SK_GRAPHITE_ENABLED
if (FLAGS_graphite) {
if (const SkCommandLineConfigGraphite *graphiteConfig = config->asConfigGraphite()) {
return new GraphiteSink(graphiteConfig);
}
}
#endif
if (const SkCommandLineConfigSvg* svgConfig = config->asConfigSvg()) {
int pageIndex = svgConfig->getPageIndex();
return new SVGSink(pageIndex);
}
Add config options to run different GPU APIs to dm and nanobench Add extended config specification form that can be used to run different gpu backend with different APIs. The configs can be specified with the form: gpu(api=string,dit=bool,nvpr=bool,samples=int) This replaces and removes the --gpuAPI flag. All existing configs should still work. Adds following documentation: out/Debug/dm --help config Flags: --config: type: string default: 565 8888 gpu nonrendering Options: 565 8888 debug gpu gpudebug gpudft gpunull msaa16 msaa4 nonrendering null nullgpu nvprmsaa16 nvprmsaa4 pdf pdf_poppler skp svg xps or use extended form 'backend(option=value,...)'. Extended form: 'backend(option=value,...)' Possible backends and options: gpu(api=string,dit=bool,nvpr=bool,samples=int) GPU backend api type: string default: native. Select graphics API to use with gpu backend. Options: native Use platform default OpenGL or OpenGL ES backend. gl Use OpenGL. gles Use OpenGL ES. debug Use debug OpenGL. null Use null OpenGL. dit type: bool default: false. Use device independent text. nvpr type: bool default: false. Use NV_path_rendering OpenGL and OpenGL ES extension. samples type: int default: 0. Use multisampling with N samples. Predefined configs: gpu = gpu() msaa4 = gpu(samples=4) msaa16 = gpu(samples=16) nvprmsaa4 = gpu(nvpr=true,samples=4) nvprmsaa16 = gpu(nvpr=true,samples=16) gpudft = gpu(dit=true) gpudebug = gpu(api=debug) gpunull = gpu(api=null) debug = gpu(api=debug) nullgpu = gpu(api=null) BUG=skia:2992 Committed: https://skia.googlesource.com/skia/+/e13ca329fca4c28cf4e078561f591ab27b743d23 GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1490113005 Committed: https://skia.googlesource.com/skia/+/c8b4336444e7b90382e04e33665fb3b8490b825b Committed: https://skia.googlesource.com/skia/+/9ebc3f0ee6db215dde461dc4777d85988cf272dd Review URL: https://codereview.chromium.org/1490113005
2015-12-23 09:33:00 +00:00
#define SINK(t, sink, ...) if (config->getBackend().equals(t)) return new sink(__VA_ARGS__)
Add config options to run different GPU APIs to dm and nanobench Add extended config specification form that can be used to run different gpu backend with different APIs. The configs can be specified with the form: gpu(api=string,dit=bool,nvpr=bool,samples=int) This replaces and removes the --gpuAPI flag. All existing configs should still work. Adds following documentation: out/Debug/dm --help config Flags: --config: type: string default: 565 8888 gpu nonrendering Options: 565 8888 debug gpu gpudebug gpudft gpunull msaa16 msaa4 nonrendering null nullgpu nvprmsaa16 nvprmsaa4 pdf pdf_poppler skp svg xps or use extended form 'backend(option=value,...)'. Extended form: 'backend(option=value,...)' Possible backends and options: gpu(api=string,dit=bool,nvpr=bool,samples=int) GPU backend api type: string default: native. Select graphics API to use with gpu backend. Options: native Use platform default OpenGL or OpenGL ES backend. gl Use OpenGL. gles Use OpenGL ES. debug Use debug OpenGL. null Use null OpenGL. dit type: bool default: false. Use device independent text. nvpr type: bool default: false. Use NV_path_rendering OpenGL and OpenGL ES extension. samples type: int default: 0. Use multisampling with N samples. Predefined configs: gpu = gpu() msaa4 = gpu(samples=4) msaa16 = gpu(samples=16) nvprmsaa4 = gpu(nvpr=true,samples=4) nvprmsaa16 = gpu(nvpr=true,samples=16) gpudft = gpu(dit=true) gpudebug = gpu(api=debug) gpunull = gpu(api=null) debug = gpu(api=debug) nullgpu = gpu(api=null) BUG=skia:2992 Committed: https://skia.googlesource.com/skia/+/e13ca329fca4c28cf4e078561f591ab27b743d23 GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1490113005 Committed: https://skia.googlesource.com/skia/+/c8b4336444e7b90382e04e33665fb3b8490b825b Review URL: https://codereview.chromium.org/1490113005
2015-12-22 07:48:13 +00:00
if (FLAGS_cpu) {
SINK("g8", RasterSink, kGray_8_SkColorType);
SINK("565", RasterSink, kRGB_565_SkColorType);
SINK("4444", RasterSink, kARGB_4444_SkColorType);
SINK("8888", RasterSink, kN32_SkColorType);
SINK("rgba", RasterSink, kRGBA_8888_SkColorType);
SINK("bgra", RasterSink, kBGRA_8888_SkColorType);
SINK("rgbx", RasterSink, kRGB_888x_SkColorType);
SINK("1010102", RasterSink, kRGBA_1010102_SkColorType);
SINK("101010x", RasterSink, kRGB_101010x_SkColorType);
SINK("bgra1010102", RasterSink, kBGRA_1010102_SkColorType);
SINK("bgr101010x", RasterSink, kBGR_101010x_SkColorType);
SINK("f16", RasterSink, kRGBA_F16_SkColorType);
SINK("f16norm", RasterSink, kRGBA_F16Norm_SkColorType);
SINK("f32", RasterSink, kRGBA_F32_SkColorType);
SINK("srgba", RasterSink, kSRGBA_8888_SkColorType);
SINK("pdf", PDFSink, false, SK_ScalarDefaultRasterDPI);
SINK("skp", SKPSink);
SINK("svg", SVGSink);
SINK("null", NullSink);
SINK("xps", XPSSink);
SINK("pdfa", PDFSink, true, SK_ScalarDefaultRasterDPI);
SINK("pdf300", PDFSink, false, 300);
SINK("jsdebug", DebugSink);
}
#undef SINK
return nullptr;
}
Add config options to run different GPU APIs to dm and nanobench Add extended config specification form that can be used to run different gpu backend with different APIs. The configs can be specified with the form: gpu(api=string,dit=bool,nvpr=bool,samples=int) This replaces and removes the --gpuAPI flag. All existing configs should still work. Adds following documentation: out/Debug/dm --help config Flags: --config: type: string default: 565 8888 gpu nonrendering Options: 565 8888 debug gpu gpudebug gpudft gpunull msaa16 msaa4 nonrendering null nullgpu nvprmsaa16 nvprmsaa4 pdf pdf_poppler skp svg xps or use extended form 'backend(option=value,...)'. Extended form: 'backend(option=value,...)' Possible backends and options: gpu(api=string,dit=bool,nvpr=bool,samples=int) GPU backend api type: string default: native. Select graphics API to use with gpu backend. Options: native Use platform default OpenGL or OpenGL ES backend. gl Use OpenGL. gles Use OpenGL ES. debug Use debug OpenGL. null Use null OpenGL. dit type: bool default: false. Use device independent text. nvpr type: bool default: false. Use NV_path_rendering OpenGL and OpenGL ES extension. samples type: int default: 0. Use multisampling with N samples. Predefined configs: gpu = gpu() msaa4 = gpu(samples=4) msaa16 = gpu(samples=16) nvprmsaa4 = gpu(nvpr=true,samples=4) nvprmsaa16 = gpu(nvpr=true,samples=16) gpudft = gpu(dit=true) gpudebug = gpu(api=debug) gpunull = gpu(api=null) debug = gpu(api=debug) nullgpu = gpu(api=null) BUG=skia:2992 Committed: https://skia.googlesource.com/skia/+/e13ca329fca4c28cf4e078561f591ab27b743d23 GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1490113005 Committed: https://skia.googlesource.com/skia/+/c8b4336444e7b90382e04e33665fb3b8490b825b Committed: https://skia.googlesource.com/skia/+/9ebc3f0ee6db215dde461dc4777d85988cf272dd Review URL: https://codereview.chromium.org/1490113005
2015-12-23 09:33:00 +00:00
static Sink* create_via(const SkString& tag, Sink* wrapped) {
#define VIA(t, via, ...) if (tag.equals(t)) return new via(__VA_ARGS__)
#ifdef TEST_VIA_SVG
VIA("svg", ViaSVG, wrapped);
#endif
VIA("serialize", ViaSerialization, wrapped);
VIA("pic", ViaPicture, wrapped);
VIA("rtblend", ViaRuntimeBlend, wrapped);
if (FLAGS_matrix.count() == 4) {
SkMatrix m;
m.reset();
m.setScaleX((SkScalar)atof(FLAGS_matrix[0]));
m.setSkewX ((SkScalar)atof(FLAGS_matrix[1]));
m.setSkewY ((SkScalar)atof(FLAGS_matrix[2]));
m.setScaleY((SkScalar)atof(FLAGS_matrix[3]));
VIA("matrix", ViaMatrix, m, wrapped);
VIA("upright", ViaUpright, m, wrapped);
}
#undef VIA
return nullptr;
}
static bool gather_sinks(const GrContextOptions& grCtxOptions, bool defaultConfigs) {
Add config options to run different GPU APIs to dm and nanobench Add extended config specification form that can be used to run different gpu backend with different APIs. The configs can be specified with the form: gpu(api=string,dit=bool,nvpr=bool,samples=int) This replaces and removes the --gpuAPI flag. All existing configs should still work. Adds following documentation: out/Debug/dm --help config Flags: --config: type: string default: 565 8888 gpu nonrendering Options: 565 8888 debug gpu gpudebug gpudft gpunull msaa16 msaa4 nonrendering null nullgpu nvprmsaa16 nvprmsaa4 pdf pdf_poppler skp svg xps or use extended form 'backend(option=value,...)'. Extended form: 'backend(option=value,...)' Possible backends and options: gpu(api=string,dit=bool,nvpr=bool,samples=int) GPU backend api type: string default: native. Select graphics API to use with gpu backend. Options: native Use platform default OpenGL or OpenGL ES backend. gl Use OpenGL. gles Use OpenGL ES. debug Use debug OpenGL. null Use null OpenGL. dit type: bool default: false. Use device independent text. nvpr type: bool default: false. Use NV_path_rendering OpenGL and OpenGL ES extension. samples type: int default: 0. Use multisampling with N samples. Predefined configs: gpu = gpu() msaa4 = gpu(samples=4) msaa16 = gpu(samples=16) nvprmsaa4 = gpu(nvpr=true,samples=4) nvprmsaa16 = gpu(nvpr=true,samples=16) gpudft = gpu(dit=true) gpudebug = gpu(api=debug) gpunull = gpu(api=null) debug = gpu(api=debug) nullgpu = gpu(api=null) BUG=skia:2992 Committed: https://skia.googlesource.com/skia/+/e13ca329fca4c28cf4e078561f591ab27b743d23 GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1490113005 Committed: https://skia.googlesource.com/skia/+/c8b4336444e7b90382e04e33665fb3b8490b825b Committed: https://skia.googlesource.com/skia/+/9ebc3f0ee6db215dde461dc4777d85988cf272dd Review URL: https://codereview.chromium.org/1490113005
2015-12-23 09:33:00 +00:00
SkCommandLineConfigArray configs;
ParseConfigs(FLAGS_config, &configs);
AutoreleasePool pool;
Add config options to run different GPU APIs to dm and nanobench Add extended config specification form that can be used to run different gpu backend with different APIs. The configs can be specified with the form: gpu(api=string,dit=bool,nvpr=bool,samples=int) This replaces and removes the --gpuAPI flag. All existing configs should still work. Adds following documentation: out/Debug/dm --help config Flags: --config: type: string default: 565 8888 gpu nonrendering Options: 565 8888 debug gpu gpudebug gpudft gpunull msaa16 msaa4 nonrendering null nullgpu nvprmsaa16 nvprmsaa4 pdf pdf_poppler skp svg xps or use extended form 'backend(option=value,...)'. Extended form: 'backend(option=value,...)' Possible backends and options: gpu(api=string,dit=bool,nvpr=bool,samples=int) GPU backend api type: string default: native. Select graphics API to use with gpu backend. Options: native Use platform default OpenGL or OpenGL ES backend. gl Use OpenGL. gles Use OpenGL ES. debug Use debug OpenGL. null Use null OpenGL. dit type: bool default: false. Use device independent text. nvpr type: bool default: false. Use NV_path_rendering OpenGL and OpenGL ES extension. samples type: int default: 0. Use multisampling with N samples. Predefined configs: gpu = gpu() msaa4 = gpu(samples=4) msaa16 = gpu(samples=16) nvprmsaa4 = gpu(nvpr=true,samples=4) nvprmsaa16 = gpu(nvpr=true,samples=16) gpudft = gpu(dit=true) gpudebug = gpu(api=debug) gpunull = gpu(api=null) debug = gpu(api=debug) nullgpu = gpu(api=null) BUG=skia:2992 Committed: https://skia.googlesource.com/skia/+/e13ca329fca4c28cf4e078561f591ab27b743d23 GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1490113005 Committed: https://skia.googlesource.com/skia/+/c8b4336444e7b90382e04e33665fb3b8490b825b Committed: https://skia.googlesource.com/skia/+/9ebc3f0ee6db215dde461dc4777d85988cf272dd Review URL: https://codereview.chromium.org/1490113005
2015-12-23 09:33:00 +00:00
for (int i = 0; i < configs.count(); i++) {
const SkCommandLineConfig& config = *configs[i];
Sink* sink = create_sink(grCtxOptions, &config);
Add config options to run different GPU APIs to dm and nanobench Add extended config specification form that can be used to run different gpu backend with different APIs. The configs can be specified with the form: gpu(api=string,dit=bool,nvpr=bool,samples=int) This replaces and removes the --gpuAPI flag. All existing configs should still work. Adds following documentation: out/Debug/dm --help config Flags: --config: type: string default: 565 8888 gpu nonrendering Options: 565 8888 debug gpu gpudebug gpudft gpunull msaa16 msaa4 nonrendering null nullgpu nvprmsaa16 nvprmsaa4 pdf pdf_poppler skp svg xps or use extended form 'backend(option=value,...)'. Extended form: 'backend(option=value,...)' Possible backends and options: gpu(api=string,dit=bool,nvpr=bool,samples=int) GPU backend api type: string default: native. Select graphics API to use with gpu backend. Options: native Use platform default OpenGL or OpenGL ES backend. gl Use OpenGL. gles Use OpenGL ES. debug Use debug OpenGL. null Use null OpenGL. dit type: bool default: false. Use device independent text. nvpr type: bool default: false. Use NV_path_rendering OpenGL and OpenGL ES extension. samples type: int default: 0. Use multisampling with N samples. Predefined configs: gpu = gpu() msaa4 = gpu(samples=4) msaa16 = gpu(samples=16) nvprmsaa4 = gpu(nvpr=true,samples=4) nvprmsaa16 = gpu(nvpr=true,samples=16) gpudft = gpu(dit=true) gpudebug = gpu(api=debug) gpunull = gpu(api=null) debug = gpu(api=debug) nullgpu = gpu(api=null) BUG=skia:2992 Committed: https://skia.googlesource.com/skia/+/e13ca329fca4c28cf4e078561f591ab27b743d23 GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1490113005 Committed: https://skia.googlesource.com/skia/+/c8b4336444e7b90382e04e33665fb3b8490b825b Committed: https://skia.googlesource.com/skia/+/9ebc3f0ee6db215dde461dc4777d85988cf272dd Review URL: https://codereview.chromium.org/1490113005
2015-12-23 09:33:00 +00:00
if (sink == nullptr) {
info("Skipping config %s: Don't understand '%s'.\n", config.getTag().c_str(),
config.getTag().c_str());
Add config options to run different GPU APIs to dm and nanobench Add extended config specification form that can be used to run different gpu backend with different APIs. The configs can be specified with the form: gpu(api=string,dit=bool,nvpr=bool,samples=int) This replaces and removes the --gpuAPI flag. All existing configs should still work. Adds following documentation: out/Debug/dm --help config Flags: --config: type: string default: 565 8888 gpu nonrendering Options: 565 8888 debug gpu gpudebug gpudft gpunull msaa16 msaa4 nonrendering null nullgpu nvprmsaa16 nvprmsaa4 pdf pdf_poppler skp svg xps or use extended form 'backend(option=value,...)'. Extended form: 'backend(option=value,...)' Possible backends and options: gpu(api=string,dit=bool,nvpr=bool,samples=int) GPU backend api type: string default: native. Select graphics API to use with gpu backend. Options: native Use platform default OpenGL or OpenGL ES backend. gl Use OpenGL. gles Use OpenGL ES. debug Use debug OpenGL. null Use null OpenGL. dit type: bool default: false. Use device independent text. nvpr type: bool default: false. Use NV_path_rendering OpenGL and OpenGL ES extension. samples type: int default: 0. Use multisampling with N samples. Predefined configs: gpu = gpu() msaa4 = gpu(samples=4) msaa16 = gpu(samples=16) nvprmsaa4 = gpu(nvpr=true,samples=4) nvprmsaa16 = gpu(nvpr=true,samples=16) gpudft = gpu(dit=true) gpudebug = gpu(api=debug) gpunull = gpu(api=null) debug = gpu(api=debug) nullgpu = gpu(api=null) BUG=skia:2992 Committed: https://skia.googlesource.com/skia/+/e13ca329fca4c28cf4e078561f591ab27b743d23 GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1490113005 Committed: https://skia.googlesource.com/skia/+/c8b4336444e7b90382e04e33665fb3b8490b825b Committed: https://skia.googlesource.com/skia/+/9ebc3f0ee6db215dde461dc4777d85988cf272dd Review URL: https://codereview.chromium.org/1490113005
2015-12-23 09:33:00 +00:00
continue;
}
// The command line config already parsed out the via-style color space. Apply it here.
sink->setColorSpace(config.refColorSpace());
Add config options to run different GPU APIs to dm and nanobench Add extended config specification form that can be used to run different gpu backend with different APIs. The configs can be specified with the form: gpu(api=string,dit=bool,nvpr=bool,samples=int) This replaces and removes the --gpuAPI flag. All existing configs should still work. Adds following documentation: out/Debug/dm --help config Flags: --config: type: string default: 565 8888 gpu nonrendering Options: 565 8888 debug gpu gpudebug gpudft gpunull msaa16 msaa4 nonrendering null nullgpu nvprmsaa16 nvprmsaa4 pdf pdf_poppler skp svg xps or use extended form 'backend(option=value,...)'. Extended form: 'backend(option=value,...)' Possible backends and options: gpu(api=string,dit=bool,nvpr=bool,samples=int) GPU backend api type: string default: native. Select graphics API to use with gpu backend. Options: native Use platform default OpenGL or OpenGL ES backend. gl Use OpenGL. gles Use OpenGL ES. debug Use debug OpenGL. null Use null OpenGL. dit type: bool default: false. Use device independent text. nvpr type: bool default: false. Use NV_path_rendering OpenGL and OpenGL ES extension. samples type: int default: 0. Use multisampling with N samples. Predefined configs: gpu = gpu() msaa4 = gpu(samples=4) msaa16 = gpu(samples=16) nvprmsaa4 = gpu(nvpr=true,samples=4) nvprmsaa16 = gpu(nvpr=true,samples=16) gpudft = gpu(dit=true) gpudebug = gpu(api=debug) gpunull = gpu(api=null) debug = gpu(api=debug) nullgpu = gpu(api=null) BUG=skia:2992 Committed: https://skia.googlesource.com/skia/+/e13ca329fca4c28cf4e078561f591ab27b743d23 GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1490113005 Committed: https://skia.googlesource.com/skia/+/c8b4336444e7b90382e04e33665fb3b8490b825b Committed: https://skia.googlesource.com/skia/+/9ebc3f0ee6db215dde461dc4777d85988cf272dd Review URL: https://codereview.chromium.org/1490113005
2015-12-23 09:33:00 +00:00
const SkTArray<SkString>& parts = config.getViaParts();
for (int j = parts.count(); j-- > 0;) {
const SkString& part = parts[j];
Sink* next = create_via(part, sink);
if (next == nullptr) {
info("Skipping config %s: Don't understand '%s'.\n", config.getTag().c_str(),
part.c_str());
delete sink;
sink = nullptr;
break;
}
sink = next;
}
if (sink) {
push_sink(config, sink);
}
}
// If no configs were requested (just running tests, perhaps?), then we're okay.
if (configs.count() == 0 ||
// If we're using the default configs, we're okay.
defaultConfigs ||
// Otherwise, make sure that all specified configs have become sinks.
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
configs.count() == gSinks->count()) {
return true;
}
return false;
}
static bool match(const char* needle, const char* haystack) {
if ('~' == needle[0]) {
return !match(needle + 1, haystack);
}
if (0 == strcmp("_", needle)) {
return true;
}
return nullptr != strstr(haystack, needle);
}
static bool should_skip(const char* sink, const char* src,
const char* srcOptions, const char* name) {
for (int i = 0; i < FLAGS_skip.count() - 3; i += 4) {
if (match(FLAGS_skip[i+0], sink) &&
match(FLAGS_skip[i+1], src) &&
match(FLAGS_skip[i+2], srcOptions) &&
match(FLAGS_skip[i+3], name)) {
return true;
}
}
return false;
}
// Even when a Task Sink reports to be non-threadsafe (e.g. GPU), we know things like
// .png encoding are definitely thread safe. This lets us offload that work to CPU threads.
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
static SkTaskGroup* gDefinitelyThreadSafeWork = new SkTaskGroup;
// The finest-grained unit of work we can run: draw a single Src into a single Sink,
// report any errors, and perhaps write out the output: a .png of the bitmap, or a raw stream.
struct Task {
Task(const TaggedSrc& src, const TaggedSink& sink) : src(src), sink(sink) {}
const TaggedSrc& src;
const TaggedSink& sink;
static void Run(const Task& task) {
AutoreleasePool pool;
SkString name = task.src->name();
SkString log;
if (!FLAGS_dryRun) {
SkBitmap bitmap;
SkDynamicMemoryWStream stream;
start(task.sink.tag.c_str(), task.src.tag.c_str(),
task.src.options.c_str(), name.c_str());
Replace DM:Error with DM::Result. Initially, Error was written with the intent that an empty string meant Ok and anything else meant Fatal. This made things simple with implicit constructors from strings. With the introduction of Nonfatal the state was now tied up with an additional boolean. Now the empty string meant Ok and causes the new boolean to be ignored, or at least that is the way it was used since Error didn't actually enforce that itself. This leads to GMs which return kSkip but don't set the message to not be skipped. This could be fixed in several ways. The first would be for the GMSrc to notice that a GM had returned kSkip with an empty message and create the Error::Nonfatal with a non-empty message. This has the downside of being some extra unexpected complexity and doesn't prevent similar issues from arising in the future. The second would be to change how DM interprets the Error, and if the non-fatal bit is set treat that as a sign to skip, even if the message is empty. This fixes the stated issue, but doesn't fix the issue where a GM can return kFail but also leave the message empty. This could again be fixed by either modifying GMSrc::draw or GM::drawContent, but this also seems a bit brittle in not preventing this from happening again in the future. So this replaces Error with Result, which makes the status orthogonal to the message. It does lose the automatic conversion from string, but by being able to wrap the many uses of SkStringPrintf the explicit nature doesn't add much additional noise to the more complex failure reports. Change-Id: Ibf48b67faa09a91a4a9d792d204bd9810b441c6c Reviewed-on: https://skia-review.googlesource.com/c/skia/+/270362 Commit-Queue: Ben Wagner <bungeman@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-02-12 16:18:46 +00:00
Result result = task.sink->draw(*task.src, &bitmap, &stream, &log);
if (!log.isEmpty()) {
info("%s %s %s %s:\n%s\n", task.sink.tag.c_str()
, task.src.tag.c_str()
, task.src.options.c_str()
, name.c_str()
, log.c_str());
}
Replace DM:Error with DM::Result. Initially, Error was written with the intent that an empty string meant Ok and anything else meant Fatal. This made things simple with implicit constructors from strings. With the introduction of Nonfatal the state was now tied up with an additional boolean. Now the empty string meant Ok and causes the new boolean to be ignored, or at least that is the way it was used since Error didn't actually enforce that itself. This leads to GMs which return kSkip but don't set the message to not be skipped. This could be fixed in several ways. The first would be for the GMSrc to notice that a GM had returned kSkip with an empty message and create the Error::Nonfatal with a non-empty message. This has the downside of being some extra unexpected complexity and doesn't prevent similar issues from arising in the future. The second would be to change how DM interprets the Error, and if the non-fatal bit is set treat that as a sign to skip, even if the message is empty. This fixes the stated issue, but doesn't fix the issue where a GM can return kFail but also leave the message empty. This could again be fixed by either modifying GMSrc::draw or GM::drawContent, but this also seems a bit brittle in not preventing this from happening again in the future. So this replaces Error with Result, which makes the status orthogonal to the message. It does lose the automatic conversion from string, but by being able to wrap the many uses of SkStringPrintf the explicit nature doesn't add much additional noise to the more complex failure reports. Change-Id: Ibf48b67faa09a91a4a9d792d204bd9810b441c6c Reviewed-on: https://skia-review.googlesource.com/c/skia/+/270362 Commit-Queue: Ben Wagner <bungeman@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-02-12 16:18:46 +00:00
if (result.isSkip()) {
done(task.sink.tag.c_str(), task.src.tag.c_str(),
task.src.options.c_str(), name.c_str());
return;
}
if (result.isFatal()) {
fail(SkStringPrintf("%s %s %s %s: %s",
task.sink.tag.c_str(),
task.src.tag.c_str(),
task.src.options.c_str(),
name.c_str(),
result.c_str()));
}
// Run verifiers if specified
if (FLAGS_runVerifiers) {
RunGMVerifiers(task, bitmap);
}
// We're likely switching threads here, so we must capture by value, [=] or [foo,bar].
SkStreamAsset* data = stream.detachAsStream().release();
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
gDefinitelyThreadSafeWork->add([task,name,bitmap,data]{
std::unique_ptr<SkStreamAsset> ownedData(data);
std::unique_ptr<HashAndEncode> hashAndEncode;
SkString md5;
if (!FLAGS_writePath.isEmpty() || !FLAGS_readPath.isEmpty()) {
SkMD5 hash;
if (data->getLength()) {
hash.writeStream(data, data->getLength());
data->rewind();
} else {
hashAndEncode = std::make_unique<HashAndEncode>(bitmap);
hashAndEncode->feedHash(&hash);
}
SkMD5::Digest digest = hash.finish();
for (int i = 0; i < 16; i++) {
md5.appendf("%02x", digest.data[i]);
}
}
if (!FLAGS_readPath.isEmpty() &&
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
!gGold->contains(Gold(task.sink.tag, task.src.tag,
task.src.options, name, md5))) {
fail(SkStringPrintf("%s not found for %s %s %s %s in %s",
md5.c_str(),
task.sink.tag.c_str(),
task.src.tag.c_str(),
task.src.options.c_str(),
name.c_str(),
FLAGS_readPath[0]));
}
// Tests sometimes use a nullptr ext to indicate no image should be uploaded.
const char* ext = task.sink->fileExtension();
if (ext && !FLAGS_writePath.isEmpty()) {
#if defined(SK_BUILD_FOR_MAC)
if (FLAGS_rasterize_pdf && SkString("pdf").equals(ext)) {
SkASSERT(data->getLength() > 0);
sk_sp<SkData> blob = SkData::MakeFromStream(data, data->getLength());
SkUniqueCFRef<CGDataProviderRef> provider{
CGDataProviderCreateWithData(nullptr,
blob->data(),
blob->size(),
nullptr)};
SkUniqueCFRef<CGPDFDocumentRef> pdf{
CGPDFDocumentCreateWithProvider(provider.get())};
CGPDFPageRef page = CGPDFDocumentGetPage(pdf.get(), 1);
CGRect bounds = CGPDFPageGetBoxRect(page, kCGPDFMediaBox);
const int w = (int)CGRectGetWidth (bounds),
h = (int)CGRectGetHeight(bounds);
SkBitmap rasterized;
rasterized.allocPixels(SkImageInfo::Make(
w, h, kRGBA_8888_SkColorType, kPremul_SkAlphaType));
rasterized.eraseColor(SK_ColorWHITE);
SkUniqueCFRef<CGColorSpaceRef> cs{CGColorSpaceCreateDeviceRGB()};
CGBitmapInfo info = kCGBitmapByteOrder32Big |
(CGBitmapInfo)kCGImageAlphaPremultipliedLast;
SkUniqueCFRef<CGContextRef> ctx{CGBitmapContextCreate(
rasterized.getPixels(), w,h,8, rasterized.rowBytes(), cs.get(), info)};
CGContextDrawPDFPage(ctx.get(), page);
// Skip calling hashAndEncode->feedHash(SkMD5*)... we want the .pdf's hash.
hashAndEncode = std::make_unique<HashAndEncode>(rasterized);
WriteToDisk(task, md5, "png", nullptr,0, &rasterized, hashAndEncode.get());
} else
#endif
if (data->getLength()) {
WriteToDisk(task, md5, ext, data, data->getLength(), nullptr, nullptr);
SkASSERT(bitmap.drawsNothing());
} else if (!bitmap.drawsNothing()) {
WriteToDisk(task, md5, ext, nullptr, 0, &bitmap, hashAndEncode.get());
}
}
SkPixmap pm;
if (FLAGS_checkF16 && bitmap.colorType() == kRGBA_F16Norm_SkColorType &&
bitmap.peekPixels(&pm)) {
bool unclamped = false;
for (int y = 0; y < pm.height() && !unclamped; ++y)
for (int x = 0; x < pm.width() && !unclamped; ++x) {
Sk4f rgba = SkHalfToFloat_finite_ftz(*pm.addr64(x, y));
float a = rgba[3];
if (a > 1.0f || (rgba < 0.0f).anyTrue() || (rgba > a).anyTrue()) {
SkDebugf("[%s] F16Norm pixel [%d, %d] unclamped: (%g, %g, %g, %g)\n",
name.c_str(), x, y, rgba[0], rgba[1], rgba[2], rgba[3]);
unclamped = true;
}
}
}
});
}
done(task.sink.tag.c_str(), task.src.tag.c_str(), task.src.options.c_str(), name.c_str());
}
static SkString identify_gamut(SkColorSpace* cs) {
if (!cs) {
return SkString("untagged");
}
skcms_Matrix3x3 gamut;
if (cs->toXYZD50(&gamut)) {
auto eq = [](skcms_Matrix3x3 x, skcms_Matrix3x3 y) {
for (int i = 0; i < 3; i++)
for (int j = 0; j < 3; j++) {
if (x.vals[i][j] != y.vals[i][j]) { return false; }
}
return true;
};
if (eq(gamut, SkNamedGamut::kSRGB )) { return SkString("sRGB"); }
if (eq(gamut, SkNamedGamut::kAdobeRGB )) { return SkString("Adobe"); }
if (eq(gamut, SkNamedGamut::kDisplayP3)) { return SkString("P3"); }
if (eq(gamut, SkNamedGamut::kRec2020 )) { return SkString("2020"); }
if (eq(gamut, SkNamedGamut::kXYZ )) { return SkString("XYZ"); }
if (eq(gamut, gNarrow_toXYZD50 )) { return SkString("narrow"); }
return SkString("other");
}
return SkString("non-XYZ");
}
static SkString identify_transfer_fn(SkColorSpace* cs) {
if (!cs) {
return SkString("untagged");
}
auto eq = [](skcms_TransferFunction x, skcms_TransferFunction y) {
return x.g == y.g
&& x.a == y.a
&& x.b == y.b
&& x.c == y.c
&& x.d == y.d
&& x.e == y.e
&& x.f == y.f;
};
skcms_TransferFunction tf;
cs->transferFn(&tf);
switch (classify_transfer_fn(tf)) {
case sRGBish_TF:
if (tf.a == 1 && tf.b == 0 && tf.c == 0 && tf.d == 0 && tf.e == 0 && tf.f == 0) {
return SkStringPrintf("gamma %.3g", tf.g);
}
if (eq(tf, SkNamedTransferFn::kSRGB)) { return SkString("sRGB"); }
if (eq(tf, SkNamedTransferFn::kRec2020)) { return SkString("2020"); }
return SkStringPrintf("%.3g %.3g %.3g %.3g %.3g %.3g %.3g",
tf.g, tf.a, tf.b, tf.c, tf.d, tf.e, tf.f);
case PQish_TF:
if (eq(tf, SkNamedTransferFn::kPQ)) { return SkString("PQ"); }
return SkStringPrintf("PQish %.3g %.3g %.3g %.3g %.3g %.3g",
tf.a, tf.b, tf.c, tf.d, tf.e, tf.f);
case HLGish_TF:
if (eq(tf, SkNamedTransferFn::kHLG)) { return SkString("HLG"); }
return SkStringPrintf("HLGish %.3g %.3g %.3g %.3g %.3g (%.3g)",
tf.a, tf.b, tf.c, tf.d, tf.e, tf.f+1);
case HLGinvish_TF: break;
case Bad_TF: break;
}
return SkString("non-numeric");
}
static void WriteToDisk(const Task& task,
SkString md5,
const char* ext,
SkStream* data, size_t len,
const SkBitmap* bitmap,
const HashAndEncode* hashAndEncode) {
JsonWriter::BitmapResult result;
result.name = task.src->name();
result.config = task.sink.tag;
result.sourceType = task.src.tag;
result.sourceOptions = task.src.options;
result.ext = ext;
result.md5 = md5;
if (bitmap) {
result.gamut = identify_gamut (bitmap->colorSpace());
result.transferFn = identify_transfer_fn (bitmap->colorSpace());
result.colorType = ToolUtils::colortype_name (bitmap->colorType());
result.alphaType = ToolUtils::alphatype_name (bitmap->alphaType());
result.colorDepth = ToolUtils::colortype_depth(bitmap->colorType());
}
JsonWriter::AddBitmapResult(result);
// If an MD5 is uninteresting, we want it noted in the JSON file,
// but don't want to dump it out as a .png (or whatever ext is).
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
if (gUninterestingHashes->contains(md5)) {
return;
}
const char* dir = FLAGS_writePath[0];
SkString resources = GetResourcePath();
if (0 == strcmp(dir, "@")) { // Needed for iOS.
dir = resources.c_str();
}
sk_mkdir(dir);
SkString path;
if (FLAGS_nameByHash) {
path = SkOSPath::Join(dir, result.md5.c_str());
path.append(".");
path.append(ext);
if (sk_exists(path.c_str())) {
return; // Content-addressed. If it exists already, we're done.
}
} else {
Add config options to run different GPU APIs to dm and nanobench Add extended config specification form that can be used to run different gpu backend with different APIs. The configs can be specified with the form: gpu(api=string,dit=bool,nvpr=bool,samples=int) This replaces and removes the --gpuAPI flag. All existing configs should still work. Adds following documentation: out/Debug/dm --help config Flags: --config: type: string default: 565 8888 gpu nonrendering Options: 565 8888 debug gpu gpudebug gpudft gpunull msaa16 msaa4 nonrendering null nullgpu nvprmsaa16 nvprmsaa4 pdf pdf_poppler skp svg xps or use extended form 'backend(option=value,...)'. Extended form: 'backend(option=value,...)' Possible backends and options: gpu(api=string,dit=bool,nvpr=bool,samples=int) GPU backend api type: string default: native. Select graphics API to use with gpu backend. Options: native Use platform default OpenGL or OpenGL ES backend. gl Use OpenGL. gles Use OpenGL ES. debug Use debug OpenGL. null Use null OpenGL. dit type: bool default: false. Use device independent text. nvpr type: bool default: false. Use NV_path_rendering OpenGL and OpenGL ES extension. samples type: int default: 0. Use multisampling with N samples. Predefined configs: gpu = gpu() msaa4 = gpu(samples=4) msaa16 = gpu(samples=16) nvprmsaa4 = gpu(nvpr=true,samples=4) nvprmsaa16 = gpu(nvpr=true,samples=16) gpudft = gpu(dit=true) gpudebug = gpu(api=debug) gpunull = gpu(api=null) debug = gpu(api=debug) nullgpu = gpu(api=null) BUG=skia:2992 Committed: https://skia.googlesource.com/skia/+/e13ca329fca4c28cf4e078561f591ab27b743d23 GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1490113005 Committed: https://skia.googlesource.com/skia/+/c8b4336444e7b90382e04e33665fb3b8490b825b Committed: https://skia.googlesource.com/skia/+/9ebc3f0ee6db215dde461dc4777d85988cf272dd Review URL: https://codereview.chromium.org/1490113005
2015-12-23 09:33:00 +00:00
path = SkOSPath::Join(dir, task.sink.tag.c_str());
sk_mkdir(path.c_str());
path = SkOSPath::Join(path.c_str(), task.src.tag.c_str());
sk_mkdir(path.c_str());
if (0 != strcmp(task.src.options.c_str(), "")) {
path = SkOSPath::Join(path.c_str(), task.src.options.c_str());
sk_mkdir(path.c_str());
}
path = SkOSPath::Join(path.c_str(), task.src->name().c_str());
path.append(".");
path.append(ext);
}
SkFILEWStream file(path.c_str());
if (!file.isValid()) {
fail(SkStringPrintf("Can't open %s for writing.\n", path.c_str()));
return;
}
if (bitmap) {
SkASSERT(hashAndEncode);
if (!hashAndEncode->encodePNG(&file,
result.md5.c_str(),
FLAGS_key,
FLAGS_properties)) {
fail(SkStringPrintf("Can't encode PNG to %s.\n", path.c_str()));
return;
}
} else {
if (!file.writeStream(data, len)) {
fail(SkStringPrintf("Can't write to %s.\n", path.c_str()));
return;
}
}
}
static void RunGMVerifiers(const Task& task, const SkBitmap& actualBitmap) {
const SkString name = task.src->name();
auto verifierList = task.src->getVerifiers();
if (verifierList == nullptr) {
return;
}
skiagm::verifiers::VerifierResult
res = verifierList->verifyAll(task.sink->colorInfo(), actualBitmap);
if (!res.ok()) {
fail(
SkStringPrintf(
"%s %s %s %s: verifier failed: %s", task.sink.tag.c_str(), task.src.tag.c_str(),
task.src.options.c_str(), name.c_str(), res.message().c_str()));
}
}
};
/*~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~*/
// Unit tests don't fit so well into the Src/Sink model, so we give them special treatment.
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
static SkTDArray<skiatest::Test>* gParallelTests = new SkTDArray<skiatest::Test>;
static SkTDArray<skiatest::Test>* gSerialTests = new SkTDArray<skiatest::Test>;
static void gather_tests() {
if (!FLAGS_src.contains("tests")) {
return;
}
for (const skiatest::Test& test : skiatest::TestRegistry::Range()) {
if (!in_shard()) {
continue;
}
if (CommandLineFlags::ShouldSkip(FLAGS_match, test.fName)) {
continue;
}
if (test.fNeedsGpu && FLAGS_gpu) {
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
gSerialTests->push_back(test);
} else if (test.fNeedsGraphite && FLAGS_graphite) {
gSerialTests->push_back(test);
} else if (!test.fNeedsGpu && !test.fNeedsGraphite && FLAGS_cpu) {
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
gParallelTests->push_back(test);
}
}
}
static void run_test(skiatest::Test test, const GrContextOptions& grCtxOptions) {
struct : public skiatest::Reporter {
void reportFailed(const skiatest::Failure& failure) override {
fail(failure.toString());
}
bool allowExtendedTest() const override {
return FLAGS_pathOpsExtended;
}
bool verbose() const override { return FLAGS_veryVerbose; }
} reporter;
if (!FLAGS_dryRun && !should_skip("_", "tests", "_", test.fName)) {
AutoreleasePool pool;
GrContextOptions options = grCtxOptions;
test.modifyGrContextOptions(&options);
skiatest::ReporterContext ctx(&reporter, SkString(test.fName));
start("unit", "test", "", test.fName);
test.run(&reporter, options);
}
done("unit", "test", "", test.fName);
}
/*~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~*/
int main(int argc, char** argv) {
#if defined(__MSVC_RUNTIME_CHECKS)
_RTC_SetErrorFunc(RuntimeCheckErrorFunc);
#endif
#if defined(SK_BUILD_FOR_ANDROID_FRAMEWORK) && defined(SK_HAS_HEIF_LIBRARY)
android::ProcessState::self()->startThreadPool();
#endif
CommandLineFlags::Parse(argc, argv);
initializeEventTracingForTools();
#if !defined(SK_BUILD_FOR_GOOGLE3) && defined(SK_BUILD_FOR_IOS)
cd_Documents();
#endif
setbuf(stdout, nullptr);
setup_crash_handler();
CommonFlags::SetDefaultFontMgr();
CommonFlags::SetAnalyticAA();
gSkForceRasterPipelineBlitter = FLAGS_forceRasterPipeline;
gUseSkVMBlitter = FLAGS_skvm;
gSkVMAllowJIT = FLAGS_jit;
// The bots like having a verbose.log to upload, so always touch the file even if --verbose.
if (!FLAGS_writePath.isEmpty()) {
sk_mkdir(FLAGS_writePath[0]);
gVLog = fopen(SkOSPath::Join(FLAGS_writePath[0], "verbose.log").c_str(), "w");
}
if (FLAGS_verbose) {
gVLog = stderr;
}
GrContextOptions grCtxOptions;
CommonFlags::SetCtxOptions(&grCtxOptions);
dump_json(); // It's handy for the bots to assume this is ~never missing.
SkAutoGraphics ag;
SkTaskGroup::Enabler enabled(FLAGS_threads);
Revert "Revert "resources: remove most uses of GetResourcePath()"" This reverts commit cca230055921d2df8708ed6f9abcc2d43468dc7f. Reason for revert: think I guessed wrong about g32 -- unreverting Original change's description: > Revert "resources: remove most uses of GetResourcePath()" > > This reverts commit 5093a539def3ae09df324018f2343827009b2e05. > > Reason for revert: google3 seems broken > > Original change's description: > > resources: remove most uses of GetResourcePath() > > > > Going forward, we will standardize on GetResourceAsData(), which will > > make it easier to run tests in environments without access to the > > filesystem. > > > > Also: GetResourceAsData() complains when a resource is missing. > > This is usually an error. > > > > Change-Id: Iaf70b71b0ca5ed8cd1a5538a60ef185ae8736188 > > Reviewed-on: https://skia-review.googlesource.com/82642 > > Reviewed-by: Hal Canary <halcanary@google.com> > > Commit-Queue: Hal Canary <halcanary@google.com> > > TBR=halcanary@google.com,scroggo@google.com > > Change-Id: Ic5a7c0167c995a672e6b06dc92abe00564432214 > No-Presubmit: true > No-Tree-Checks: true > No-Try: true > Reviewed-on: https://skia-review.googlesource.com/83001 > Reviewed-by: Mike Reed <reed@google.com> > Commit-Queue: Mike Reed <reed@google.com> TBR=halcanary@google.com,scroggo@google.com,reed@google.com Change-Id: I5a46e4de61186a8a5eb9cacd3275e24e311d5a07 No-Presubmit: true No-Tree-Checks: true No-Try: true Reviewed-on: https://skia-review.googlesource.com/82942 Reviewed-by: Mike Reed <reed@google.com> Commit-Queue: Mike Reed <reed@google.com>
2017-12-09 01:27:41 +00:00
if (nullptr == GetResourceAsData("images/color_wheel.png")) {
info("Some resources are missing. Do you need to set --resourcePath?\n");
}
gather_gold();
gather_uninteresting_hashes();
if (!gather_srcs()) {
return 1;
}
// TODO(dogben): This is a bit ugly. Find a cleaner way to do this.
bool defaultConfigs = true;
for (int i = 0; i < argc; i++) {
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
static constexpr char kConfigArg[] = "--config";
if (strcmp(argv[i], kConfigArg) == 0) {
defaultConfigs = false;
break;
}
}
if (!gather_sinks(grCtxOptions, defaultConfigs)) {
return 1;
}
gather_tests();
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
gPending = gSrcs->count() * gSinks->count() + gParallelTests->count() + gSerialTests->count();
info("%d srcs * %d sinks + %d tests == %d tasks\n",
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
gSrcs->count(), gSinks->count(), gParallelTests->count() + gSerialTests->count(),
gPending);
// Kick off as much parallel work as we can, making note of any serial work we'll need to do.
SkTaskGroup parallel;
SkTArray<Task> serial;
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
for (TaggedSink& sink : *gSinks) {
for (TaggedSrc& src : *gSrcs) {
if (src->veto(sink->flags()) ||
should_skip(sink.tag.c_str(), src.tag.c_str(),
src.options.c_str(), src->name().c_str())) {
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
SkAutoSpinlock lock(*gMutex);
gPending--;
continue;
}
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
Task task(src, sink);
if (src->serial() || sink->serial()) {
serial.push_back(task);
} else {
parallel.add([task] { Task::Run(task); });
}
}
}
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
for (skiatest::Test& test : *gParallelTests) {
parallel.add([test, grCtxOptions] { run_test(test, grCtxOptions); });
}
// With the parallel work running, run serial tasks and tests here on main thread.
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
for (Task& task : serial) { Task::Run(task); }
for (skiatest::Test& test : *gSerialTests) { run_test(test, grCtxOptions); }
// Wait for any remaining parallel work to complete (including any spun off of serial tasks).
parallel.wait();
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
gDefinitelyThreadSafeWork->wait();
// At this point we're back in single-threaded land.
// We'd better have run everything.
SkASSERT(gPending == 0);
// Make sure we've flushed all our results to disk.
dump_json();
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
if (!gFailures->empty()) {
info("Failures:\n");
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
for (const SkString& fail : *gFailures) {
info("\t%s\n", fail.c_str());
}
Prevent non-trivial destructors from running after exit in DM. Previously, DM destroyed a large number of non-trivial objects at shutdown time. Because no shutdown order is promised across translation units by the standard, this can lead to bugs which only reproduce capriciously, at the whim of the linker. http://go/totw/110#the-fix-safe-initialization-no-destruction "Destruction issues are usually solved by defining your static data in such a way that the destructor never runs. The most common way to do this is to heap allocate the static object - pointers don't have destructors." http://go/cstyle#decision_on_destruction "Global and static variables that use dynamic initialization or have non-trivial destructors create complexity that can easily lead to hard- to-find bugs. Dynamic initialization is not ordered across translation units, and neither is destruction (except that destruction happens in reverse order of initialization). When one initialization refers to another variable with static storage duration, it is possible that this causes an object to be accessed before its lifetime has begun (or after its lifetime has ended). Moreover, when a program starts threads that are not joined at exit, those threads may attempt to access objects after their lifetime has ended if their destructor has already run." Change-Id: I54eedcd813295a23923deb925b0ca2adfff69f7d Reviewed-on: https://skia-review.googlesource.com/c/skia/+/297872 Auto-Submit: John Stiles <johnstiles@google.com> Commit-Queue: Mike Klein <mtklein@google.com> Reviewed-by: Mike Klein <mtklein@google.com>
2020-06-19 22:34:51 +00:00
info("%d failures\n", gFailures->count());
return 1;
}
SkGraphics::PurgeAllCaches();
info("Finished!\n");
return 0;
}