A little JIT proof of concept for SkRasterPipeline, using xbyak, which is a header-only assembler. It's x86-only, but supports x86 very thoroughly, and it's very user friendly (at least as far as assembler libraries go...).
CQ_INCLUDE_TRYBOTS=skia.primary:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD
Change-Id: Ie17e562b0f3fff5914041badfb2c1fe4f86efab8
Reviewed-on: https://skia-review.googlesource.com/5730
Reviewed-by: Herb Derby <herb@google.com>
Reviewed-by: Heather Miller <hcm@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>
CQ_INCLUDE_TRYBOTS=skia.primary:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD
Change-Id: I75e232faee6ad48f65bac5b119a461280b27bbc8
Reviewed-on: https://skia-review.googlesource.com/6661
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Mike Klein <mtklein@chromium.org>
This reverts commit efcc41805b.
Reason for revert: Possible culprit for Chromium failures.
Original change's description:
> Explicitly fail read/writePixels in invalid color space scenarios
>
> It's not well defined what to do when moving from a nullptr color space to
> a tagged destination (drawing, reading, writing, etc...). In these
> scenarios, at least, we can choose to disallow the operation (rather than
> produce an unexpected or inconsistent result).
>
> BUG=skia:
>
> Change-Id: I033b23c6f2bb00664efc8fdab1b3f52053d77695
> Reviewed-on: https://skia-review.googlesource.com/6600
> Commit-Queue: Brian Osman <brianosman@google.com>
> Reviewed-by: Brian Salomon <bsalomon@google.com>
> Reviewed-by: Matt Sarett <msarett@google.com>
>
TBR=mtklein@google.com,bsalomon@google.com,msarett@google.com,brianosman@google.com,reviews@skia.org
BUG=skia:
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
Change-Id: I17791f9285089ede42b7921324e0dc264865be1d
Reviewed-on: https://skia-review.googlesource.com/6628
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
It's not well defined what to do when moving from a nullptr color space to
a tagged destination (drawing, reading, writing, etc...). In these
scenarios, at least, we can choose to disallow the operation (rather than
produce an unexpected or inconsistent result).
BUG=skia:
Change-Id: I033b23c6f2bb00664efc8fdab1b3f52053d77695
Reviewed-on: https://skia-review.googlesource.com/6600
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Reviewed-by: Matt Sarett <msarett@google.com>
Trying to get better information about what's failing (and make this
easier in the future).
BUG=skia:6086
Change-Id: Iedb1269abb4527170b919bd90bce625a7f78f05a
Reviewed-on: https://skia-review.googlesource.com/6584
Commit-Queue: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Removes the feedback to GrDrawOp via GrPipelineOptimizations.
Change-Id: I3cb17cad41779af292a92385fcd5ac23ae5a1ffd
Reviewed-on: https://skia-review.googlesource.com/6561
Reviewed-by: Greg Daniel <egdaniel@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
BUG=skia:
Change-Id: I8d4594fcf0eeebf598871bfe9203ed52460c98ce
Reviewed-on: https://skia-review.googlesource.com/6558
Reviewed-by: Mike Klein <mtklein@chromium.org>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Matt Sarett <msarett@google.com>
BUG=skia:
Change-Id: I744af0efd4d48a8932b834092ed2dbad13008c1d
Reviewed-on: https://skia-review.googlesource.com/6556
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Matt Sarett <msarett@google.com>
When the path was "large" (as defined by ScaleFactor(...)), the computed
bounds would not be adjusted to the correct space. Make sure to scale
the result in those cases.
BUG=chromium:678162
Change-Id: Ia2eb94050c4620286e9abb69976dbc0202ecc307
Reviewed-on: https://skia-review.googlesource.com/6501
Reviewed-by: Cary Clark <caryclark@google.com>
Commit-Queue: Cary Clark <caryclark@google.com>
If a frame does not have a valid transparent index and it covers the
prior frame, it does not really depend on that frame. Instead, it
depends on the frame that the prior frame depends on.
Determine this once we have parsed the local color map (if any), so a
transparent index out of range of the color map is not considered
valid.
Share code that determines whether a frame has a transparent pixel.
Add a test that we compute the dependencies correctly. randPixelsAnim.gif
has 13 frames. After the first, the frames cover all combinations of
- Whether the prior frame was keep, restoreBG or restoreToPrevious
- Whether the new frame covers the prior frame
- Whether the new frame has a transparent pixel
(It only does so when using a global color table. It may make sense to
expand the test to also cover using local color tables.)
The test caught a bug where we incorrectly reused an existing
SkColorTable for a different frame. Fix that bug by keeping track of
the transparent index associated with the current SkColorTable.
Change-Id: I3cf6be7f612990fa7a00d9e74d116d31bd227526
Reviewed-on: https://skia-review.googlesource.com/6402
Reviewed-by: Matt Sarett <msarett@google.com>
Commit-Queue: Leon Scroggins <scroggo@google.com>
Reenable test on GTX10700 bot.
BUG=skia:6080
Change-Id: Ieb4292e88fc337c226dad7ac82c6da84879e9522
Reviewed-on: https://skia-review.googlesource.com/6523
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Reviewed-by: Ben Wagner <benjaminwagner@google.com>
abort if incoming data is out of range
TBR=reed@google.com
BUG=676866
Change-Id: I7d4850611654a399e32ea2012b23ca369dc53e70
Reviewed-on: https://skia-review.googlesource.com/6525
Reviewed-by: Cary Clark <caryclark@google.com>
Commit-Queue: Cary Clark <caryclark@google.com>
Instead use std::unique_ptr to manage GrOp lifetime.
Change-Id: Ic1dc1e0ffd7254c3994221f498677af5bbf66a71
Reviewed-on: https://skia-review.googlesource.com/6479
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
BUG=skia:
Change-Id: Icf616bec73e81aad97815b519566ff5b9db611e3
Reviewed-on: https://skia-review.googlesource.com/6495
Commit-Queue: Hal Canary <halcanary@google.com>
Reviewed-by: Mike Klein <mtklein@chromium.org>
Change-Id: I80f951976558a284e55386e0a368f08bd835d8ca
Reviewed-on: https://skia-review.googlesource.com/6359
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Make the test factories use sk_sp.
Change-Id: Idba630b84deb2848f2203a80fd72e1efa5fc6acf
Reviewed-on: https://skia-review.googlesource.com/6342
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>
This matches the version used by Chromium (crbug.com/675306) and
Android (https://android-review.googlesource.com/#/c/316352/).
Update our test. In the new version, we successfully decode a subset
from incomplete input that we previously did not.
Change-Id: I3442bf59ffdf223a723d8aa75f5b9b816b9e9c3c
Reviewed-on: https://skia-review.googlesource.com/6320
Reviewed-by: Matt Sarett <msarett@google.com>
Commit-Queue: Leon Scroggins <scroggo@google.com>
BUG=skia:
Change-Id: Idd950c3da2c517780b24e312a4e7cc16ee413c99
Reviewed-on: https://skia-review.googlesource.com/6270
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Matt Sarett <msarett@google.com>
The ICC errata supports the opposite of what we do.
http://www.color.org/icc_specs2.xalterTBR=reed@google.com
BUG=skia:
CQ_INCLUDE_TRYBOTS=skia.primary:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD
Change-Id: I18ace7f312926b264e624c30d8cb983eff5c434b
Reviewed-on: https://skia-review.googlesource.com/6277
Commit-Queue: Matt Sarett <msarett@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
SkClipOp.h & SkPictureFlat.h
Invalid SkClipOps were getting through - the question here is where (for a class enum) is a good place to put the k*Mask definition.
SkPath1DPathEffect
NaNs were getting past.
SkBlurMaskFilter
Assert wasn't necessary since we whacked the flag on the next line.
Change-Id: I87f95ad39f4760284f881d7c4500eb82fcdba282
Reviewed-on: https://skia-review.googlesource.com/6194
Commit-Queue: Robert Phillips <robertphillips@google.com>
Reviewed-by: Herb Derby <herb@google.com>
If the input SkStream has a length and position, do not copy and store
LZW blocks or ColorMaps. Instead, mark the position and size, and read
from the stream when necessary.
This will save memory in Chromium's use case, which has already
buffered all of its data.
In the case where we *do* need to copy, store it on the SkStreamBuffer.
This allows SkGifImageReader to have simpler code.
Add tests.
Change-Id: Ic65fa766328ae2e5974b2084bc2099e19aced731
Reviewed-on: https://skia-review.googlesource.com/6157
Reviewed-by: Matt Sarett <msarett@google.com>
Commit-Queue: Leon Scroggins <scroggo@google.com>
This reverts commit eb733fbf56.
Reason for revert: Revert patch was automatically merged incorrectly?
Original change's description:
> Revert "WIP: Skia support library for ICC tasks"
>
> This reverts commit fc8dc3194a.
>
> Reason for revert: Breaks Build-Mac-Clang-Arm7-{Debug,Release}-iOS builds.
> Example tasks:
> * https://chromium-swarm.appspot.com/task?id=3322f668620b9e10&refresh=10
> * https://chromium-swarm.appspot.com/task?id=332296146331e810&refresh=10
>
> Original change's description:
> > WIP: Skia support library for ICC tasks
> >
> > As a starting point, this would be mostly trivial to implement using
> > SkColorSpace.
> >
> > This also would give us the flexibility to begin to move all of
> > the ICC related code from SkColorSpace to SkICC.
> >
> > What are the advantages of moving this away from SkColorSpace?
> > (1) A long term goal (once Chrome uses SkCodec), might be to
> > move SkColorSpace::MakeICC() out of the public API. That way,
> > we can guarantee that we can draw to/from *any* SkColorSpace.
> > (2) Keeps SkColorSpace separate from ICC-specific representations
> > like SkColorSpaceTransferFn etc.
> >
> > BUG=skia:
> >
> > Change-Id: Iddeb9903221fb57fbfc01218d8641c928b4a5165
> > Reviewed-on: https://skia-review.googlesource.com/5676
> > Commit-Queue: Matt Sarett <msarett@google.com>
> > Reviewed-by: Brian Osman <brianosman@google.com>
> > Reviewed-by: Mike Reed <reed@google.com>
> >
>
> TBR=mtklein@google.com,msarett@google.com,brianosman@google.com,reed@google.com,reviews@skia.org
> BUG=skia:
> NOPRESUBMIT=true
> NOTREECHECKS=true
> NOTRY=true
>
> Change-Id: Ibdf272fce25892402bd3e85595fb8814cdf59856
> Reviewed-on: https://skia-review.googlesource.com/6232
> Commit-Queue: Ravi Mistry <rmistry@google.com>
> Reviewed-by: Ravi Mistry <rmistry@google.com>
>
TBR=mtklein@google.com,rmistry@google.com,msarett@google.com,reviews@skia.org,brianosman@google.com,reed@google.com
BUG=skia:
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
Change-Id: I68b1624cfab8adfe31b17e1193a7766507dec8b0
Reviewed-on: https://skia-review.googlesource.com/6233
Commit-Queue: Ravi Mistry <rmistry@google.com>
Reviewed-by: Ravi Mistry <rmistry@google.com>
This reverts commit fc8dc3194a.
Reason for revert: Breaks Build-Mac-Clang-Arm7-{Debug,Release}-iOS builds.
Example tasks:
* https://chromium-swarm.appspot.com/task?id=3322f668620b9e10&refresh=10
* https://chromium-swarm.appspot.com/task?id=332296146331e810&refresh=10
Original change's description:
> WIP: Skia support library for ICC tasks
>
> As a starting point, this would be mostly trivial to implement using
> SkColorSpace.
>
> This also would give us the flexibility to begin to move all of
> the ICC related code from SkColorSpace to SkICC.
>
> What are the advantages of moving this away from SkColorSpace?
> (1) A long term goal (once Chrome uses SkCodec), might be to
> move SkColorSpace::MakeICC() out of the public API. That way,
> we can guarantee that we can draw to/from *any* SkColorSpace.
> (2) Keeps SkColorSpace separate from ICC-specific representations
> like SkColorSpaceTransferFn etc.
>
> BUG=skia:
>
> Change-Id: Iddeb9903221fb57fbfc01218d8641c928b4a5165
> Reviewed-on: https://skia-review.googlesource.com/5676
> Commit-Queue: Matt Sarett <msarett@google.com>
> Reviewed-by: Brian Osman <brianosman@google.com>
> Reviewed-by: Mike Reed <reed@google.com>
>
TBR=mtklein@google.com,msarett@google.com,brianosman@google.com,reed@google.com,reviews@skia.org
BUG=skia:
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
Change-Id: Ibdf272fce25892402bd3e85595fb8814cdf59856
Reviewed-on: https://skia-review.googlesource.com/6232
Commit-Queue: Ravi Mistry <rmistry@google.com>
Reviewed-by: Ravi Mistry <rmistry@google.com>
Pathops is very well exercised with winding paths,
but less so with xor (even odd) paths.
Rewrite the xor main loop to look like the winding
one to take advantage of the latter's bug fixes.
TBR=reed@google.com
BUG=skia:6041
Change-Id: Ied8d522254a327b1817b54f0abbf4414f5fab7da
Reviewed-on: https://skia-review.googlesource.com/6228
Reviewed-by: Cary Clark <caryclark@google.com>
Commit-Queue: Cary Clark <caryclark@google.com>
As a starting point, this would be mostly trivial to implement using
SkColorSpace.
This also would give us the flexibility to begin to move all of
the ICC related code from SkColorSpace to SkICC.
What are the advantages of moving this away from SkColorSpace?
(1) A long term goal (once Chrome uses SkCodec), might be to
move SkColorSpace::MakeICC() out of the public API. That way,
we can guarantee that we can draw to/from *any* SkColorSpace.
(2) Keeps SkColorSpace separate from ICC-specific representations
like SkColorSpaceTransferFn etc.
BUG=skia:
Change-Id: Iddeb9903221fb57fbfc01218d8641c928b4a5165
Reviewed-on: https://skia-review.googlesource.com/5676
Commit-Queue: Matt Sarett <msarett@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Reviewed-by: Mike Reed <reed@google.com>
Recent changes (crrev.com/2045293002) made it so that a GIF may not
support index 8. In that case, make SkAndroidCodec not suggest index 8.
Add a test and a new test file. randPixelsOffset.gif is the same as
randPixels.gif, except its frame is offset. Since it does not have a
transparent index, we have to decode to kN32.
Change-Id: I1c09ab9094083de3dfc436632b3c26dbde1dccbd
Reviewed-on: https://skia-review.googlesource.com/6196
Commit-Queue: Leon Scroggins <scroggo@google.com>
Reviewed-by: Matt Sarett <msarett@google.com>
This method does the same thing as ::bytesWritten but has a confusing
name. It appears there are no external users, so remove it.
Change-Id: I06aed269200c34c2dda36605092f8ea37fcec693
Reviewed-on: https://skia-review.googlesource.com/6188
Reviewed-by: Mike Reed <reed@google.com>
Commit-Queue: Ben Wagner <bungeman@google.com>
SkPath::dump() and SkPath::dumpHex() dump the fill type
in addition to the data so that the original path can
be faithfully reconstructed.
This may be a small part of why some error cases
aren't reproduced.
R=reed@google.com
BUG=skia:6041
Change-Id: Ice86bf08ea907a6b87ceef182a9316a3c979af0b
Reviewed-on: https://skia-review.googlesource.com/6185
Reviewed-by: Mike Reed <reed@google.com>
Commit-Queue: Cary Clark <caryclark@google.com>
This gives a picture image a preferred "native" color space, which
facilitates caching and other things.
BUG=skia:
Change-Id: I95988c14d17f96d7d870b3d1c3b723c36e2c170d
Reviewed-on: https://skia-review.googlesource.com/6158
Reviewed-by: Mike Reed <reed@google.com>
Reviewed-by: Florin Malita <fmalita@chromium.org>
Commit-Queue: Brian Osman <brianosman@google.com>
Change-Id: I6410eae41f051ce38bef6f38d670924c3483c325
Reviewed-on: https://skia-review.googlesource.com/6163
Commit-Queue: Brian Salomon <bsalomon@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
In SkBmpCodec, if the header size does not match a known header,
stop trying to create an SkCodec. We do not know of any BMPs with
arbitrarily sized headers, so this should not cause any real
regressions.
In addition, this fixes a bug where we attempt to read too much data
from a file. Since we attempt to read the header size in one read,
and a size reported by the "BMP" may be larger than SSIZE_MAX, this
will crash when reading from a file.
Add a test.
BUG:b/33651913
Change-Id: I0f3292db3124dc5ac5cbdbc07196bda130a49ba7
Reviewed-on: https://skia-review.googlesource.com/6150
Commit-Queue: Leon Scroggins <scroggo@google.com>
Reviewed-by: Matt Sarett <msarett@google.com>
Checking to invalidate this on every write() call has a measurable cost, so removing it both simplifies the class and speeds it up.
BUG=skia:
Change-Id: Idf0baa265c9a0b5d26d82fce948c61ed9b0810b1
Reviewed-on: https://skia-review.googlesource.com/6096
Reviewed-by: Ben Wagner <bungeman@google.com>
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Mike Reed <reed@google.com>
This CL forces all GrSurface copies to go through a GrSurfaceContext (rather than GrContext).
There is a bit of goofiness going on here until read/writePixels is also consolidated in GrSurfaceContext and a proxy-backed SkImage/SkSurface is added.
This is a reland of https://skia-review.googlesource.com/c/5773/ (Add a deferred copy surface)
Change-Id: Ib8fd96d0569274ef781366eb900ed8ee839ae9bd
Reviewed-on: https://skia-review.googlesource.com/6109
Reviewed-by: Brian Salomon <bsalomon@google.com>
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
This indicates whether the frame has been fully received, i.e. the
stream contains enough data to decode to the end of the frame.
A client may want to use this to know whether they should attempt to
decode this frame, if they do not want to decode partial frames.
Change-Id: I336c7031b0c0b8c1401ce040f5372aedc87fdc14
Reviewed-on: https://skia-review.googlesource.com/5703
Reviewed-by: Chris Blume <cblume@google.com>
Reviewed-by: Derek Sollenberger <djsollen@google.com>
Commit-Queue: Leon Scroggins <scroggo@google.com>
SkOpAngle::alignmentSameSide()
Shifting an edge to align it for angle sorting may move a compared edge to the opposite side.
For lines that are shifted, check to see if this is so.
class SkOpContourBuilder
If the path contains a pair of lines that cancel, skip them as early as possible.
While not strictly necessary, this optimization is cheap and makes debugging much easier.
SkOpEdgeBuilder::walk()
case SkPath::kCubic_Verb:
If max curvature or inflections break a cubic into pieces, make sure that the pieces are
large enough to process. If not, add the broken piece back to a neighbor.
Correct debugging that had gone stale.
Add active span debugging cache so only changes are shown.
TBR=reed@google.com
BUG=skia:6401
Change-Id: I766f77e4fb9b76537cf5464961addb103114f5db
Reviewed-on: https://skia-review.googlesource.com/5764
Reviewed-by: Cary Clark <caryclark@google.com>
Commit-Queue: Cary Clark <caryclark@google.com>
This reverts commit 398487a850.
Reason for revert: See if this is causing the roll failure
Original change's description:
> Add a deferred copy surface (take 2)
>
> This CL forces all GrSurface copies to go through a GrSurfaceContext (rather than GrContext).
>
> There is a bit of goofiness going on here until read/writePixels is also consolidated in GrSurfaceContext and a proxy-backed SkImage/SkSurface is added.
>
> This is a reland of https://skia-review.googlesource.com/c/5773/ (Add a deferred copy surface)
>
> Change-Id: Ide560f569aede5e622420dc2f30eef76357d69f4
> Reviewed-on: https://skia-review.googlesource.com/5939
> Reviewed-by: Brian Osman <brianosman@google.com>
> Reviewed-by: Brian Salomon <bsalomon@google.com>
> Commit-Queue: Robert Phillips <robertphillips@google.com>
>
TBR=bsalomon@google.com,robertphillips@google.com,brianosman@google.com,reviews@skia.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
Change-Id: I1ef40f0d5fb0bca62031f94f10eb18acd753e913
Reviewed-on: https://skia-review.googlesource.com/6024
Commit-Queue: Robert Phillips <robertphillips@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
This CL forces all GrSurface copies to go through a GrSurfaceContext (rather than GrContext).
There is a bit of goofiness going on here until read/writePixels is also consolidated in GrSurfaceContext and a proxy-backed SkImage/SkSurface is added.
This is a reland of https://skia-review.googlesource.com/c/5773/ (Add a deferred copy surface)
Change-Id: Ide560f569aede5e622420dc2f30eef76357d69f4
Reviewed-on: https://skia-review.googlesource.com/5939
Reviewed-by: Brian Osman <brianosman@google.com>
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>
This reverts commit 4431de6af9.
Reason for revert: ANGLE errors (at the very least)
Original change's description:
> Add a deferred copy surface
>
> This CL forces all GrSurface copies to go through a GrSurfaceContext (rather than GrContext).
>
> There is a bit of goofiness going on here until read/writePixels is also consolidated in GrSurfaceContext and a proxy-backed SkImage/SkSurface is added.
>
> Change-Id: Iab1867668d8146a766201158a251b9174438ee2b
> Reviewed-on: https://skia-review.googlesource.com/5773
> Reviewed-by: Brian Osman <brianosman@google.com>
> Reviewed-by: Robert Phillips <robertphillips@google.com>
> Commit-Queue: Robert Phillips <robertphillips@google.com>
>
TBR=bsalomon@google.com,robertphillips@google.com,brianosman@google.com
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
Change-Id: I61408d9e306b9b1ab32f93ab086e95184e12857f
Reviewed-on: https://skia-review.googlesource.com/5938
Commit-Queue: Robert Phillips <robertphillips@google.com>
Reviewed-by: Robert Phillips <robertphillips@google.com>
The more I look at std::unordered_map and co., the less I like them.
I think we might want to bet on SkTHash*.
As a simple first improvement, add move support.
Next comes shrinking, and then I'll start moving over SkTDynamicHash users.
BUG=skia:6053
Change-Id: Ifdb5d713aab66434ca271c7f18a0cbbb0720099c
Reviewed-on: https://skia-review.googlesource.com/5943
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Herb Derby <herb@google.com>
Reviewed-by: Hal Canary <halcanary@google.com>
