More steps toward GN/Windows.

I think I'm now at the point of needing to just resolve missing symbols.

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=3201

Change-Id: Ib908bd72c23f2d4bafd17182eedcb2fc85c422e5
Reviewed-on: https://skia-review.googlesource.com/3201
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Mike Klein <mtklein@chromium.org>

implement A8 destination fast-path for SkPixelInfo::CopyPixels

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2407093002

Review-Url: https://codereview.chromium.org/2407093002

GN: windows compiles locally.

Not yet linking.  Ignoring bots for now.

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=3200

Change-Id: Idd75033313df60844c2ba602f7845b3c52987bc2
Reviewed-on: https://skia-review.googlesource.com/3200
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Mike Klein <mtklein@chromium.org>

fix vertex alpha calculation in GrTesselator

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2405813003

Review-Url: https://codereview.chromium.org/2405813003

Add AA Analytic Compile Flag (and small fixes)

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2410303002

Review-Url: https://codereview.chromium.org/2410303002

skpbench: add debug prints for thermal trip points

Prints thermal trip points that have been exceeded when a
HardwareException is raised, and verbosity is set to debug. This can
help us correlate thermal trip points with throttling and detect
future throttling before it occurs.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2408893002

Review-Url: https://codereview.chromium.org/2408893002

Fix 32-bit Vulkan build

In 32-bit land, VkFence is uint64_t, so reinterpret_cast (between two
identical integral types) is illegal.

BUG=skia:

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=3143

Change-Id: Iba9507f5678f647710f4abd35023c192bf6eed66
Reviewed-on: https://skia-review.googlesource.com/3143
Reviewed-by: Brian Osman <brianosman@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>

Add SkColorSpaceXform to the public API

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2390263002

Review-Url: https://codereview.chromium.org/2390263002

Fix GN formatting presubmit on Windows.

BUG=skia:

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=3182

Change-Id: I6e19b18d6851516357f71930e406b9fb324ea81f
Reviewed-on: https://skia-review.googlesource.com/3182
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Mike Klein <mtklein@chromium.org>
Reviewed-by: Herb Derby <herb@google.com>

Make `gn gen` work on Windows again.

Clang detection is specific to GCC-like toolchains.

BUG=skia:

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=3171

Change-Id: I7144bc8e5cd3e774625b51a6dda981284ed1fdc1
Reviewed-on: https://skia-review.googlesource.com/3171
Reviewed-by: Mike Klein <mtklein@chromium.org>
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>

Update to Android NDK r13.

The libc++ include paths have changed very slightly.  I've left GN compatible with both older r12 NDKs and the new r13 to smooth the transition.

The libc++ in r13 depends on long-double math.h functions (cosl, atanl, etc.) only available in Android API v21 (Lollipop) and up.  That's what the 64-bit bots were already on, so we just pull the 32-bit bots up to the same target.  Conveniently, the oldest bots we have (N7 and N10) are on Lollipop.

The r13 MIPS64 sysroots are a little weird... /usr/include and /usr/lib64 but no /usr/lib.  That'd be fine---we only want 64-bit builds---but Clang searches for /usr/lib64 via its path to /usr/lib, and without at least an empty /usr/lib, it can't find /usr/lib64.  So you'll see a special mips64el section in the GN config where we do this all manually (other platforms pick this all up correctly from --sysroot).  I've chosen to do this rather than fix it up in the asset create.py scripts so that we stay compatible with vanilla NDKs, which is convenient for developers.

CQ_INCLUDE_TRYBOTS=master.client.skia.compile:Build-Mac-Clang-arm64-Debug-GN_Android-Trybot,Build-Ubuntu-Clang-arm-Debug-GN_Android-Trybot,Build-Ubuntu-Clang-arm-Release-GN_Android-Trybot,Build-Ubuntu-Clang-arm64-Debug-GN_Android-Trybot,Build-Ubuntu-Clang-arm64-Debug-GN_Android_FrameworkDefs-Trybot,Build-Ubuntu-Clang-arm64-Debug-GN_Android_Vulkan-Trybot,Build-Ubuntu-Clang-arm64-Release-GN_Android-Trybot,Build-Ubuntu-Clang-mips64el-Debug-GN_Android-Trybot,Build-Ubuntu-Clang-mips64el-Release-GN_Android-Trybot,Build-Ubuntu-Clang-mips64el-Release-GN_Android-Trybot,Build-Ubuntu-Clang-mipsel-Debug-GN_Android-Trybot,Build-Ubuntu-Clang-mipsel-Release-GN_Android-Trybot,Build-Ubuntu-Clang-x64-Debug-GN_Android-Trybot,Build-Ubuntu-Clang-x64-Release-GN_Android-Trybot,Build-Ubuntu-Clang-x86-Debug-GN_Android-Trybot,Build-Ubuntu-Clang-x86-Release-GN_Android-Trybot

BUG=skia:

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=3180

Change-Id: I6f3b5d9411ded0ee49c1099490f41fa86a8736f8
Reviewed-on: https://skia-review.googlesource.com/3180
Reviewed-by: Eric Boren <borenet@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>

Revert of Turned on SkSL->GLSL compiler (patchset #49 id:1240001 of https://codereview.chromium.org/2288033003/ )

Reason for revert:
Pre-emptive revert. Sorry, but we're getting back up on our roll and Chrome reverted our last one for apparently no reason.

Original issue's description:
> Turned on SkSL->GLSL compiler
> GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2288033003
>
> Committed: https://skia.googlesource.com/skia/+/9b0fe3d125f237d9884732a48414fa85fc71b4e3
> Committed: https://skia.googlesource.com/skia/+/b12b3c6908c62c908b3680be01e3b5bfd30de310
> Committed: https://skia.googlesource.com/skia/+/f008b0a59f45c0d4bea3e66faf3b01805009ec89
> Committed: https://skia.googlesource.com/skia/+/08b2ccf398e2b81bc05d2c105837e5419899469b
> Committed: https://skia.googlesource.com/skia/+/dcfe6dba4a335e50e86ff68e3252065d4197432c
> Committed: https://skia.googlesource.com/skia/+/ccb1dd8f267f9d7fe7c9d0ce222ebc81b41853b3

TBR=benjaminwagner@google.com,bsalomon@google.com,egdaniel@google.com,ethannicholas@google.com
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true

Review-Url: https://codereview.chromium.org/2408193002

Add SkColorSpacePrimaries to help with making D50 matrices

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2304753002

Review-Url: https://codereview.chromium.org/2304753002

Remove restrictive format pattern for Fontconfig matching

This pattern pushes CFF fonts too far low in the list. Instead
it is enough to rely on the isValidPattern function to skip
Type 1 fonts.

BUG=skia:5846
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2410063002

Review-Url: https://codereview.chromium.org/2410063002

Turned on SkSL->GLSL compiler
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2288033003

Committed: https://skia.googlesource.com/skia/+/9b0fe3d125f237d9884732a48414fa85fc71b4e3
Committed: https://skia.googlesource.com/skia/+/b12b3c6908c62c908b3680be01e3b5bfd30de310
Committed: https://skia.googlesource.com/skia/+/f008b0a59f45c0d4bea3e66faf3b01805009ec89
Committed: https://skia.googlesource.com/skia/+/08b2ccf398e2b81bc05d2c105837e5419899469b
Committed: https://skia.googlesource.com/skia/+/dcfe6dba4a335e50e86ff68e3252065d4197432c
Review-Url: https://codereview.chromium.org/2288033003

Add new GM to demonstrates SkGaussianEdgeShader

This would've caught the incorrect distances in the center bug

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=3111

Change-Id: I9461a8865b561cc139a18a5e779e933d7979ee0d
Reviewed-on: https://skia-review.googlesource.com/3111
Commit-Queue: Robert Phillips <robertphillips@google.com>
Reviewed-by: Jim Van Verth <jvanverth@google.com>

Revert of Turned on SkSL->GLSL compiler (patchset #48 id:1220001 of https://codereview.chromium.org/2288033003/ )

Reason for revert:
https://build.chromium.org/p/chromium/builders/Linux%20x64/builds/26798

Original issue's description:
> Turned on SkSL->GLSL compiler
> GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2288033003
>
> Committed: https://skia.googlesource.com/skia/+/9b0fe3d125f237d9884732a48414fa85fc71b4e3
> Committed: https://skia.googlesource.com/skia/+/b12b3c6908c62c908b3680be01e3b5bfd30de310
> Committed: https://skia.googlesource.com/skia/+/f008b0a59f45c0d4bea3e66faf3b01805009ec89
> Committed: https://skia.googlesource.com/skia/+/08b2ccf398e2b81bc05d2c105837e5419899469b
> Committed: https://skia.googlesource.com/skia/+/dcfe6dba4a335e50e86ff68e3252065d4197432c

TBR=benjaminwagner@google.com,bsalomon@google.com,egdaniel@google.com,ethannicholas@google.com
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true

Review-Url: https://codereview.chromium.org/2408053002

fix 'GrClip::quickContains was hidden' warning in GrNoClip

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2394113002

Review-Url: https://codereview.chromium.org/2394113002

GN: build dump_record

BUG=skia:

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=3021

Change-Id: I447ef9a880e5b13103bac76d41e176c7940c8512
Reviewed-on: https://skia-review.googlesource.com/3021
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Mike Klein <mtklein@chromium.org>

More fixes for webpages_playback

BUG=skia:5843
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2409753002

Review-Url: https://codereview.chromium.org/2409753002

Simplify rowsDecoded returned by incrementalDecode

When sampling, we previously had to figure out the number of rows that
were written to the output based on the value reported by
incrementalDecode. (We also messed up the first time - see the fix in
crrev.com/2343153003.)

Instead, make incrementalDecode report the actual number of rows
written to. This can be provided directly to fill.

Make SkPngCodec report the correct number of rows, by incrementing its
count when it actually writes to the destination.

This also will simplify my in progress GIF change.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2407543002

Review-Url: https://codereview.chromium.org/2407543002

RecreateSKPs: Don't download and use the .boto file

It's not needed for the new bucket.

BUG=skia:5843
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2408793002

Review-Url: https://codereview.chromium.org/2408793002

Specify the Android build number as a swarming dimension

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2403103002

Review-Url: https://codereview.chromium.org/2403103002

Turned on SkSL->GLSL compiler
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2288033003

Committed: https://skia.googlesource.com/skia/+/9b0fe3d125f237d9884732a48414fa85fc71b4e3
Committed: https://skia.googlesource.com/skia/+/b12b3c6908c62c908b3680be01e3b5bfd30de310
Committed: https://skia.googlesource.com/skia/+/f008b0a59f45c0d4bea3e66faf3b01805009ec89
Committed: https://skia.googlesource.com/skia/+/08b2ccf398e2b81bc05d2c105837e5419899469b
Review-Url: https://codereview.chromium.org/2288033003

Update AndroidShadows slide with new stroking offset

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=3080

Change-Id: I0a60e99b357f930a5215201d8bb8b5f3c91819b1
Reviewed-on: https://skia-review.googlesource.com/3080
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>

Revert of Turned on SkSL->GLSL compiler (patchset #47 id:1200001 of https://codereview.chromium.org/2288033003/ )

Reason for revert:
Looks like it introduces new static initializers, and it's failing the Chrome roll.

Original issue's description:
> Turned on SkSL->GLSL compiler
> GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2288033003
>
> Committed: https://skia.googlesource.com/skia/+/9b0fe3d125f237d9884732a48414fa85fc71b4e3
> Committed: https://skia.googlesource.com/skia/+/b12b3c6908c62c908b3680be01e3b5bfd30de310
> Committed: https://skia.googlesource.com/skia/+/f008b0a59f45c0d4bea3e66faf3b01805009ec89
> Committed: https://skia.googlesource.com/skia/+/08b2ccf398e2b81bc05d2c105837e5419899469b

TBR=benjaminwagner@google.com,bsalomon@google.com,egdaniel@google.com,ethannicholas@google.com
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true

Review-Url: https://codereview.chromium.org/2403083002

Update comment, move constant inside helper struct

BUG=skia:

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=3109

Change-Id: Ife8a2434ff591bd77be1cd0fbcce50430e18c86c
Reviewed-on: https://skia-review.googlesource.com/3109
Reviewed-by: Florin Malita <fmalita@chromium.org>
Commit-Queue: Brian Osman <brianosman@google.com>

webpages_playback: Fix GS path with extra '/'

BUG=skia:5843
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2409593002

Review-Url: https://codereview.chromium.org/2409593002

Add description to Gold

BUG=skia:
R=hcm,jcgregorio
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2406753002
NOTRY=true
DOCS_PREVIEW= https://skia.org/?cl=2406753002

Review-Url: https://codereview.chromium.org/2406753002

Fix iOS surface creation for SampleApp

BUG=skia:5810
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2394843003
TBR=bsalomon@google.com

Review-Url: https://codereview.chromium.org/2394843003

Revert "SkRasterPipeline: 8x pipelines, attempt 2"

This reverts commit Id0ba250037e271a9475fe2f0989d64f0aa909bae.

crbug.com/654213
Looks like Chrome Canary's picking up Haswell code on non-Haswell machines.

Change-Id: I16f976da24db86d5c99636c472ffad56db213a2a
Reviewed-on: https://skia-review.googlesource.com/3108
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Mike Klein <mtklein@chromium.org>

Report 0 rowsDecoded for no rows (SkPngCodec interlaced)

incrementalDecode is supposed to report the number of rows decoded. It
failed to if none were decoded (for an interlaced png). Fix the bug and
add a test.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2402063002

Review-Url: https://codereview.chromium.org/2402063002

Fix an assert statement in SkPngCodec

allRowsCallback was asserting that
    rowNum - fFirstRow == fLinesDecoded

But it was expecting fFirstRow to be 0 (logically, it is, since we are
decoding all rows), and it never set it (because it doesn't really need
to, except for this assert). This is only a problem if we reuse an
SkPngCodec after previously scaling with it. Add a test that triggers
the assert, and fix it.
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2401133002

Review-Url: https://codereview.chromium.org/2401133002

Update Perf docs and add a couple screenshots.

BUG=skia:

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=3121
NOTRY=true
DOCS_PREVIEW= https://skia.org/?cl=3121

Change-Id: Ic2d1743b2dc42701202ff316277db4c944d969d3
Reviewed-on: https://skia-review.googlesource.com/3121
Reviewed-by: Joe Gregorio <jcgregorio@google.com>
Commit-Queue: Joe Gregorio <jcgregorio@google.com>

Turned on SkSL->GLSL compiler
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2288033003

Committed: https://skia.googlesource.com/skia/+/9b0fe3d125f237d9884732a48414fa85fc71b4e3
Committed: https://skia.googlesource.com/skia/+/b12b3c6908c62c908b3680be01e3b5bfd30de310
Committed: https://skia.googlesource.com/skia/+/f008b0a59f45c0d4bea3e66faf3b01805009ec89
Review-Url: https://codereview.chromium.org/2288033003

Convert gradient params test helper to a struct

Much less copy-pasted code, fewer implementation details leaking out,
and going to be easier to extend for 4f and color space testing.

BUG=skia:

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2900

Change-Id: Icc468c606aa35fbe82c64bcc398e7e348e0faa20
Reviewed-on: https://skia-review.googlesource.com/2900
Reviewed-by: Florin Malita <fmalita@chromium.org>
Commit-Queue: Brian Osman <brianosman@google.com>

webpages_playback: use gsutil instead of API, no shell_utils

BUG=skia:5843
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2406863002

Review-Url: https://codereview.chromium.org/2406863002

Roll recipe DEPS

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2402313002

Review-Url: https://codereview.chromium.org/2402313002

fix another fuzz

Add one more early exit.

TBR=reed@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2401673004

Review-Url: https://codereview.chromium.org/2401673004

fix fuzz busters

When fuzz is hit
just quit

TBR=reed@google.com
BUG=skia:5837, skia:5838, skia:5839
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2404483002

Review-Url: https://codereview.chromium.org/2404483002

SkRasterPipeline: 8x pipelines, attempt 2

Original review here: https://skia-review.googlesource.com/c/2990/

Changes since:
  - simpler implementations of load_tail() / store_tail(): slower, but more obviously correct to all compilers
  - fleshed out math ops on Sk8i and Sk8u to make unit tests happy on -Fast bot (where we always have AVX2)
  - now storing stage functions as void(*)() to avoid undefined behavior and/or linker problems.  This restores 32-bit Windows.
  - all AVX2 Sk8x methods are marked always-inline, to avoid linking the "wrong" version on Debug builds.

CQ_INCLUDE_TRYBOTS=master.client.skia:Perf-Ubuntu-Clang-GCE-CPU-AVX2-x86_64-Debug-ASAN-Trybot,Perf-Ubuntu-Clang-GCE-CPU-AVX2-x86_64-Debug-GN,Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot,Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-Fast-Trybot;master.client.skia.compile:Build-Win-MSVC-x86_64-Debug-Trybot

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=3064

Change-Id: Id0ba250037e271a9475fe2f0989d64f0aa909bae
Reviewed-on: https://skia-review.googlesource.com/3064
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Mike Klein <mtklein@chromium.org>

GrTessellator: refactor Line out of Edge.

There are cases where we need only a line equation, and not a full
Edge. Create a Line struct for this, and refactor Edge to use it.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2397963003

Review-Url: https://codereview.chromium.org/2397963003

Whitespace change to kick build.

Reduce geometry size for circles to help fill rate.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2391133004

Review-Url: https://codereview.chromium.org/2391133004

Disable svgs on nanobench when running on Valgrind

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2398173002

Review-Url: https://codereview.chromium.org/2398173002

Revert "SkRasterPipeline: 8x pipelines"

This reverts commit I1c82e5755d8e44cc0b9c6673d04b117f85d71a3a.

Reason for revert: lots of failing bots.

TBR=mtklein@chromium.org,msarett@google.com,reviews@skia.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true

Change-Id: I653bed3905187f43196504f19424985fa2a765b5
Reviewed-on: https://skia-review.googlesource.com/3063
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Mike Klein <mtklein@chromium.org>

Disable antialiasing on interior of filled RRects that need distance vectors

The interior antialiasing was causing a small transparent rectangle in the middle of filled rects

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=3062

Change-Id: If0bf26b96c8969af6d81f3a3d6fee59e8c1d747e
Reviewed-on: https://skia-review.googlesource.com/3062
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>

SkRasterPipeline: 8x pipelines

Bench runtime changes:
sRGB: 7194 -> 3735  = 1.93x faster
F16:  6531 -> 2559  = 2.55x faster

Instead of building 4x and 1-3x pipelines and then maybe 8x and 1-7x, instead build either the short ones or the long ones, but not both.  If we just take care to use a compatible run_pipeline(), there's some cross-module type disagreement but everything works out in the end.

Oddly, a few places that looked like they'd be faster using SkNx_fma() or Sk4f_round()/Sk8f_round() are actually faster the long way, e.g. multiply, add 0.5, truncate.  Curious!  In all the other places you see here that I've used SkNx_fma(), it's been a significant speedup.

This folds in a couple refactors and cleanups that I've been meaning to do.  Hope you don't mind... if find the new code considerably easier to read than the old code.

BUG=skia:

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2990
CQ_INCLUDE_TRYBOTS=master.client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot

Change-Id: I1c82e5755d8e44cc0b9c6673d04b117f85d71a3a
Reviewed-on: https://skia-review.googlesource.com/2990
Reviewed-by: Matt Sarett <msarett@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>

Assorted Android fixes

Ran into these while trying to follow https://skia.org/user/quick/android.

R=reed@google.com,djsollen@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2395273002

Review-Url: https://codereview.chromium.org/2395273002

Auto re-gen for gn cmake generator.

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=3028

Change-Id: I73c6ce9e852c280bbb2017c30717809b4cb44046
Reviewed-on: https://skia-review.googlesource.com/3028
Reviewed-by: Ben Wagner <bungeman@google.com>
Commit-Queue: Ben Wagner <bungeman@google.com>

Add distance values to interior of filled RRects

This doesn't compute the correct distance vector but it suffices for gaussian edges.

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2986

Change-Id: I8bd9e37b0f3788c30d85c95a3c0845d093f22554
Reviewed-on: https://skia-review.googlesource.com/2986
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Robert Phillips <robertphillips@google.com>

Harden SkPicturePlayback::handleOp() skips

SkValidatingReadBuffer::skip() may return null - tread more carefully
around it.

BUG=skia:5828
R=reed@google.com,mtklein@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2399043002

Review-Url: https://codereview.chromium.org/2399043002

Fix Vulkan orientation during screne rotations

This will fix the Vulkan orientation when the screen is rotated by having
the compositor do the corrections for us. However, it is suggested in the
Android Vulkan docs that for power reasons it may be better for the app
to get the needed rotation off of the VkAndroidSurface and then manually
adjust for this rotation in the rendering without the compositor doing any
corrections. This is something we can look into a later date as it would
require sending some info back from the VulkanWindowContext back to Skia
about what Matrix should be applied.

BUG=skia:

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=3052

Change-Id: I4d3d620ffcc77e607fce8d06b869c2912a61400c
Reviewed-on: https://skia-review.googlesource.com/3052
Reviewed-by: Jim Van Verth <jvanverth@google.com>
Commit-Queue: Greg Daniel <egdaniel@google.com>

Fix SkDeferredCanvas for use on android.

BUG=skia:

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=3030

Change-Id: Ie55023257736a12360a233d01096462ba2eb3e74
Reviewed-on: https://skia-review.googlesource.com/3030
Commit-Queue: Herb Derby <herb@google.com>
Reviewed-by: Mike Reed <reed@google.com>

remove sprintf

Most uses of sprintf are
in pathops testing.

Replace them with
SkString::appendf

and replace the remaining
with snprintf

R=scroggo@google.com
BUG=skia:2716
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2391133005

Review-Url: https://codereview.chromium.org/2391133005

Make load4 and store4 part of SkNx properly.

Every type now nominally has Load4() and Store4() methods.
The ones that we use are implemented.

BUG=skia:

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=3046
CQ_INCLUDE_TRYBOTS=master.client.skia:Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-SKNX_NO_SIMD-Trybot

Change-Id: I7984f0c2063ef8acbc322bd2e968f8f7eaa0d8fd
Reviewed-on: https://skia-review.googlesource.com/3046
Reviewed-by: Matt Sarett <msarett@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>

fuzzer fix

Error out when fuzzing
conditions are met.

Also, make degenerate
input line ends consistent.

TBR=reed@google.com
BUG=653452, 626164, skia:5829

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=3048

Change-Id: I23a01adde9dec07b54d66ab2418b3ea0b96e4456
Reviewed-on: https://skia-review.googlesource.com/3048
Commit-Queue: Cary Clark <caryclark@google.com>
Reviewed-by: Cary Clark <caryclark@google.com>

fix other printf warning for SkBlendMode

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2393253003

Review-Url: https://codereview.chromium.org/2393253003

fix mac all build

R=reed@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2394943002

Review-Url: https://codereview.chromium.org/2394943002

Fix SkPath::arcTo when sweepAngle is tiny and radius is big

In this function, it first check whether this arc is a lone point
or not. If not, it converts angles to unit vectors. The problem
here is that when the radius is huge and the sweepAngle is small,
the function angles_to_unit_vectors() could return a startV ==stopV.
When that happens, it will draw a dot at the point that corresponding
to the startAngle. This CL adds a special branch for this case, and
draw a connecting line between the points at startAngle and endAngle.

BUG=640031, skia:5807
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2388833002

Review-Url: https://codereview.chromium.org/2388833002

fix printf warning in dumpcanvas

BUG=skia:

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=3040

Change-Id: I850609e499a670cda0fd09941345fe221fbeeffb
Reviewed-on: https://skia-review.googlesource.com/3040
Reviewed-by: Mike Reed <reed@google.com>
Commit-Queue: Mike Reed <reed@google.com>

Revert[8] "replace SkXfermode obj with SkBlendMode enum in paints"

This reverts commit c245574ba3d0e2ade6c94b2812de3baa383bf4c4.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2396953002

TBR=

Review-Url: https://codereview.chromium.org/2396953002

tight bounds optimization

Add support for tight bounds to detect and return moveTo
followed by close or zero-length lineTo.

Also short circuit so that hard work is avoided when
the path bounds is also the tight bounds.

Avoid doing work if the bounds can be trivially computed.
Include naked moveTo coordinates in the tight bounds.

R=fmalita@chromium.org
BUG=skia:5555, skia:5553
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2394443004

Review-Url: https://codereview.chromium.org/2394443004

Revert[7] "replace SkXfermode obj with SkBlendMode enum in paints"

This reverts commit Ib4a154cdd5f5d1dcac921ef50d53b79a2d6a1be8.

Reason for revert: new assert from 100K bot

Original change's description:
> Revert[6] "replace SkXfermode obj with SkBlendMode enum in paints"
>
> - perform version check in CreateProc for XfermodeImageFilter and ArithmeticImageFilter
> This reverts commit 3ed485f4249e17abb4b11f5018d03175fd1afb44.
>
> BUG=skia:
>
> GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2992
>
> Change-Id: Ib4a154cdd5f5d1dcac921ef50d53b79a2d6a1be8
> Reviewed-on: https://skia-review.googlesource.com/2992
> Reviewed-by: Mike Reed <reed@google.com>
> Commit-Queue: Mike Reed <reed@google.com>
>

TBR=reed@google.com,reviews@skia.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true

Change-Id: I848e5a69c5cd67f2c14889f4f0a346652578c4ff
Reviewed-on: https://skia-review.googlesource.com/3023
Commit-Queue: Mike Reed <reed@google.com>
Reviewed-by: Mike Reed <reed@google.com>

Revert "add cast for printf"

This reverts commit Ib9277facab87747b3cf9b71585e63067b4be9a58.

Reason for revert: reverted related CL

Original change's description:
> add cast for printf
>
> BUG=skia:
>
> GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=3020
>
> Change-Id: Ib9277facab87747b3cf9b71585e63067b4be9a58
> Reviewed-on: https://skia-review.googlesource.com/3020
> Reviewed-by: Mike Reed <reed@google.com>
> Commit-Queue: Mike Reed <reed@google.com>
>

TBR=reed@google.com
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true

Change-Id: I1621d4ea42e9a9403ce78db147c62e8a458b4e00
Reviewed-on: https://skia-review.googlesource.com/3024
Commit-Queue: Mike Reed <reed@google.com>
Reviewed-by: Mike Reed <reed@google.com>

add cast for printf

BUG=skia:

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=3020

Change-Id: Ib9277facab87747b3cf9b71585e63067b4be9a58
Reviewed-on: https://skia-review.googlesource.com/3020
Reviewed-by: Mike Reed <reed@google.com>
Commit-Queue: Mike Reed <reed@google.com>

update deferred

BUG=skia:

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2994

Change-Id: I49e029d5825e84ba36331586bfcfbf0fe34f7c2c
Reviewed-on: https://skia-review.googlesource.com/2994
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Reed <reed@google.com>

validate using nx to match bounds

Check the path against the bounds
using Nx instead of straight scalars,

R=mtklein@google.com
BUG=skia:5541
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2388903006

Review-Url: https://codereview.chromium.org/2388903006

Use calendar.timegm instead of time.mktime in recipes

Timezones are hard.

BUG=skia:5823
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2392143004

Review-Url: https://codereview.chromium.org/2392143004

Revert[6] "replace SkXfermode obj with SkBlendMode enum in paints"

- perform version check in CreateProc for XfermodeImageFilter and ArithmeticImageFilter
This reverts commit 3ed485f4249e17abb4b11f5018d03175fd1afb44.

BUG=skia:

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2992

Change-Id: Ib4a154cdd5f5d1dcac921ef50d53b79a2d6a1be8
Reviewed-on: https://skia-review.googlesource.com/2992
Reviewed-by: Mike Reed <reed@google.com>
Commit-Queue: Mike Reed <reed@google.com>

Fix public API msg in PRESUBMIT.py

BUG=skia:5825

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2989

Change-Id: Ia074ab00ccd1500bee10a5b313636581e8265461
Reviewed-on: https://skia-review.googlesource.com/2989
Reviewed-by: Joe Gregorio <jcgregorio@google.com>
Commit-Queue: Ravi Mistry <rmistry@google.com>

Make test lower-level, make const_cast more visible.

I can only think there's something funky going on with the hidden
const_cast inside SkRasterPipeline.cpp, or with the Sk4h -> Sk4f
conversion.  So make the const_cast visible and write the test
directly in halfs instead of floats.

CQ_INCLUDE_TRYBOTS=master.client.skia:Test-Mac-Clang-MacMini6.2-CPU-AVX-x86_64-Release-GN-Trybot

BUG=skia:

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=3003

Change-Id: I3a7e19ae165fce44f177b4968a63ec04e25c4b93
Reviewed-on: https://skia-review.googlesource.com/3003
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Mike Klein <mtklein@chromium.org>

Presubmit recipe: call presubmit_support.py directly

For some reason, presubmit_support.py can't find PRESUBMIT.py
unless it has a diff. Calling this script directly allows us
to specify a fake upstream to force it to find a diff and
therefore be able to find PRESUBMIT.py.

BUG=skia:5823
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2398653002

Review-Url: https://codereview.chromium.org/2398653002

Make local coords be highp in default geometry processor

This fixes some regressions on the Nexus 10 that were introduced when the device coordinate feature was removed from GrCoordTransform.

BUG=skia:

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2987

Change-Id: Ib0286b1e056d6d8f875a45beee8b0eda85dc1382
Reviewed-on: https://skia-review.googlesource.com/2987
Reviewed-by: Robert Phillips <robertphillips@google.com>
Commit-Queue: Brian Salomon <bsalomon@google.com>

update defer-canvas for lattice and regions

BUG=skia:

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2985

Change-Id: I9106841d1cc246064d0a963b0a1e9feae109589d
Reviewed-on: https://skia-review.googlesource.com/2985
Reviewed-by: Herb Derby <herb@google.com>
Commit-Queue: Mike Reed <reed@google.com>

PRESUBMIT.py improvements

* Updated code to check for owners in reviewers for Gerrit issues instead for in the TBR= line.
* The previous TBR parsing code was not accounting for adding comments. Eg: "TBR=rmistry (Testing only)".

BUG=skia:5825

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2981

Change-Id: I910a3ae71a9f57c14f80c0b0404041cbe451a77c
Reviewed-on: https://skia-review.googlesource.com/2981
Reviewed-by: Joe Gregorio <jcgregorio@google.com>
Commit-Queue: Ravi Mistry <rmistry@google.com>

Harden degenerate gradient context handling

Certain inputs produce degenerate values at context creation time only.

Detect such cases after context creation, and abort drawing by returning
a null shader context instead.

BUG=skia:5821
R=reed@google.com,brianosman@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2397473003

Review-Url: https://codereview.chromium.org/2397473003

skpbench: add option for gpu timing

Adds a gpu timing option with a GL implementation.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2388433003

Committed: https://skia.googlesource.com/skia/+/c06720d06faab3b01eba1b8693e0ac791f06dc96
Review-Url: https://codereview.chromium.org/2388433003

webpages_playback: use new skia-skps bucket

BUG=skia:2969
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2393043003

Review-Url: https://codereview.chromium.org/2393043003

fix fuzz

This fix is slightly interesting; if the final
close of the contour degenerates into a zero-length
line, remove the previous line from the generated
contour.

TBR=reed@google.com
BUG=skia:5822

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=3004

Change-Id: Id668d13ccf6aad9bc81d78588fc77437527a0b7b
Reviewed-on: https://skia-review.googlesource.com/3004
Reviewed-by: Cary Clark <caryclark@google.com>

Fixes for presubmit recipe

- Add depot_tools to PATH
- Run on a branch
- Use max verbosity

BUG=skia:5823
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2394853002

Review-Url: https://codereview.chromium.org/2394853002

Fix bmp bug exposed by fuzzer

BUG=skia:5820
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2391933004

Review-Url: https://codereview.chromium.org/2391933004

Add action_foreach, copy, and proper target naming.

Also properly handles dependencies across OBJECT libraries.

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2939

Change-Id: I4aa48c896caf262772fe9769e742b54f6e265ab0
Reviewed-on: https://skia-review.googlesource.com/2939
Reviewed-by: Brian Salomon <bsalomon@google.com>

looks like red and blue start wrong (more unit test debugging)

BUG=skia:

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=3001

CQ_INCLUDE_TRYBOTS=master.client.skia:Test-Mac-Clang-MacMini6.2-CPU-AVX-x86_64-Release-GN-Trybot

Change-Id: I8d26b5484a2bf67d5d5891475640970046e470d8
Reviewed-on: https://skia-review.googlesource.com/3001
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Mike Klein <mtklein@chromium.org>

initialize result in SkRasterPipelineTest... more debugging

BUG=skia:

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2982

Change-Id: I7c40766c3df61fb66f63e17a2da61219d1d95d1b
Reviewed-on: https://skia-review.googlesource.com/2982
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Mike Klein <mtklein@chromium.org>

Add swarm_presubmit recipe

This allows us to add a Housekeeper-PerCommit-Presubmit bot/trybot

BUG=skia:5823
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2392663004

Review-Url: https://codereview.chromium.org/2392663004

remove dead code

Use code coverage and a new '-c' flag to find
and remove unneeded code.

TBR=reed@google.com

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2397593002

patch from issue 2397593002 at patchset 1 (http://crrev.com/2397593002#ps1)

Change-Id: I741a82de9191146c10457d4c0b9cb9605dff2069
Reviewed-on: https://skia-review.googlesource.com/2983
Reviewed-by: Cary Clark <caryclark@google.com>

more debugging for this test.

BUG=skia:

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2965

Change-Id: I583bcdeb199212840ff227942ab6ebbf57a481d7
Reviewed-on: https://skia-review.googlesource.com/2965
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Mike Klein <mtklein@chromium.org>

Mac: Remove use of deprecated font APIs

Remove use of CGContextSelectFont CGContextShowTextAtPoint APIs.
Verified that the new calls create the same pixel value for the
smoothing test as before.

Also remove workarounds for 10.6, because it is not supported anymore.

BUG=skia:5803
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2395613002

Review-Url: https://codereview.chromium.org/2395613002

Fold extra_*_flags into GN as a config.

This should make them visible to things like gn_to_cmake.py.
I'm not exactly sure what this implies about ordering and overriding.

BUG=skia:

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2938

Change-Id: I0740613993fb5bbfb8363cfa126d1f59768abf60
Reviewed-on: https://skia-review.googlesource.com/2938
Reviewed-by: Brian Salomon <bsalomon@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>

Revert "Debug Mac test failure."

This reverts commit I0ed569b585f4962a90a0b6993acc484a74055177.

Reason for revert: Mac bots look okay now.  I am puzzled.

Original change's description:
> Debug Mac test failure.
>
> CQ_EXTRA_TRYBOTS=master.client.skia:Test-Mac-Clang-MacMini6.2-CPU-AVX-x86_64-Release-GN-Trybot
> TBR=
>
> BUG=skia:
>
> GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2933
>
> Change-Id: I0ed569b585f4962a90a0b6993acc484a74055177
> Reviewed-on: https://skia-review.googlesource.com/2933
> Reviewed-by: Mike Klein <mtklein@chromium.org>
> Commit-Queue: Mike Klein <mtklein@chromium.org>
>

TBR=mtklein@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true

Change-Id: I4feef1f58172d1cc649bc311c9e6d3c86f915617
Reviewed-on: https://skia-review.googlesource.com/2963
Reviewed-by: Mike Klein <mtklein@chromium.org>
Commit-Queue: Mike Klein <mtklein@chromium.org>

Revert[5] "replace SkXfermode obj with SkBlendMode enum in paints"

This reverts commit I0fa5c58af428f3da8565465d1219a34ef8417d9a.

Reason for revert: failing to deserialize some of the 100K

Original change's description:
> Revert[4] "replace SkXfermode obj with SkBlendMode enum in paints"
>
> This reverts commit 2cbcd12281ee807214df094964c584c78932e10b.
>
> BUG=skia:
>
> GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2924
>
> Change-Id: I0fa5c58af428f3da8565465d1219a34ef8417d9a
> Reviewed-on: https://skia-review.googlesource.com/2924
> Reviewed-by: Mike Reed <reed@google.com>
> Commit-Queue: Mike Reed <reed@google.com>
>

TBR=reed@google.com,reviews@skia.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true

Change-Id: I1805a57eef5ebcac203da5989c8539345ecf806f
Reviewed-on: https://skia-review.googlesource.com/2962
Reviewed-by: Mike Reed <reed@google.com>
Commit-Queue: Mike Reed <reed@google.com>

Remove SK_SUPPORT_LEGACY_TEXTBLOB_BUILDER from public.bzl

BUG=skia:

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2960

Change-Id: I945782a5aaf416b630181bf65f517f89bae95d9f
Reviewed-on: https://skia-review.googlesource.com/2960
Commit-Queue: Mike Klein <mtklein@chromium.org>
Reviewed-by: Mike Klein <mtklein@chromium.org>

Use uint64_t for PlatformFence

VkFence is 64 bit even on 32-bit platforms.

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2391113002

Review-Url: https://codereview.chromium.org/2391113002

Make the housekeeper upload doxygen to a newer bucket

BUG=skia:2969
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2370153003
DOCS_PREVIEW= https://skia.org/?cl=2370153003

Review-Url: https://codereview.chromium.org/2370153003

Make global use-analytic-AA bit threadsafe.

I also had to cut it down to just a global atomic bool... as a field in a global singleton accessed through instance(), it's very hard to make threadsafe.

CQ_INCLUDE_TRYBOTS=master.client.skia:Test-Ubuntu-Clang-GCE-CPU-AVX2-x86_64-Release-TSAN-Trybot

BUG=skia:

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2937

Change-Id: If80be987906dd521fbe644d1d0d577009f06d0e3
Reviewed-on: https://skia-review.googlesource.com/2937
Reviewed-by: Yuqian Li <liyuqian@google.com>
Commit-Queue: Yuqian Li <liyuqian@google.com>
Commit-Queue: Mike Klein <mtklein@chromium.org>

Revert[4] "replace SkXfermode obj with SkBlendMode enum in paints"

This reverts commit 2cbcd12281ee807214df094964c584c78932e10b.

BUG=skia:

GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2924

Change-Id: I0fa5c58af428f3da8565465d1219a34ef8417d9a
Reviewed-on: https://skia-review.googlesource.com/2924
Reviewed-by: Mike Reed <reed@google.com>
Commit-Queue: Mike Reed <reed@google.com>

disallow -4 pointer

A user's homegrown unsigned integer overflow tool
complains if a nullptr is decremented.
The conicWeight pointer likes to predecrement
before walking, but this is unnecessary if
its value is nullptr.

R=reed@google.com
BUG=skia:5415
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2061833005

Review-Url: https://codereview.chromium.org/2061833005

Revert of skpbench: add option for gpu timing (patchset #7 id:120001 of https://codereview.chromium.org/2388433003/ )

Reason for revert:
many bots failing

Original issue's description:
> skpbench: add option for gpu timing
>
> Adds a gpu timing option with a GL implementation.
>
> BUG=skia:
> GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2388433003
>
> Committed: https://skia.googlesource.com/skia/+/c06720d06faab3b01eba1b8693e0ac791f06dc96

TBR=egdaniel@google.com,bsalomon@google.com,csmartdalton@google.com
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=skia:

Review-Url: https://codereview.chromium.org/2390383002

Add --test to gen_tasks.go, add presubmit check

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2392083002

Review-Url: https://codereview.chromium.org/2392083002