The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/68f66618bed5d366bd6a68060d726ae3b6f27d87

commit 68f66618bed5d366bd6a68060d726ae3b6f27d87
Author: shans <shans@chromium.org>
Date: Wed Jun 29 03:30:56 2016

Mark virtual/gpu-rasterization/fast/images/color-profile-background-clip-text.html as failing on win10

BUG= 624233 
TBR=shans@chromium.org

Review-Url: https://codereview.chromium.org/2103953004
Cr-Commit-Position: refs/heads/master@{#402700}

[modify] https://crrev.com/68f66618bed5d366bd6a68060d726ae3b6f27d87/third_party/WebKit/LayoutTests/TestExpectations

Here's another instance of the same bug on Win7
https://build.chromium.org/p/chromium.win/builders/Win7%20Tests%20%28dbg%29%281%29/builds/50203

Looking at the expected image vs the actual image, they look identical to me, but the diff image looks odd, with big red bars - I've been seeing the same red bars in other tests, I wonder if something is wrong with the color setup on the machine or the diff framework.

Will take a look - some slight (not perceptable) changes are expected, but flakiness is not... now that my change has stuck, will take a look. We can hopefully just rebase/re-enable.

This is continuing to be flaky. It failed on https://build.chromium.org/p/chromium.webkit/builders/WebKit%20Win7%20%28dbg%29/builds/6242, and also on builds 6250, 6259, 6337, 6340, 6352, and 6354.

That's a rate of about once every twenty builds. I'm going to mark the test as failing on Win instead of just Win10.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/7e299f374dcf416f3b15e02f7836eef2d8ff140c

commit 7e299f374dcf416f3b15e02f7836eef2d8ff140c
Author: brucedawson <brucedawson@chromium.org>
Date: Thu Jul 07 03:15:05 2016

Mark virtual/gpu-rasterization/fast/images/color-profile-background-clip-text.html as failing on win

Previously this test was marked as failing on Windows 10 only, but it
is failing about 5% of the time on Windows 7 also, so marking as failing
on Windows in general.

BUG= 624233 
TBR=shans@chromium.org

Review-Url: https://codereview.chromium.org/2121273004
Cr-Commit-Position: refs/heads/master@{#404059}

[modify] https://crrev.com/7e299f374dcf416f3b15e02f7836eef2d8ff140c/third_party/WebKit/LayoutTests/TestExpectations

 Issue 649589  has been merged into this issue.

Looked into this a bit. There are a couple problems here. Marking this test flaky across all platforms while investigation continues.

The problems are:
- The test doesn't correctly pull in the "Ahem" font, leading to different fonts being used on each platform.
- Skia renders test context differently depending on whether the font-cache is populated or not at the start of the test.

To summarize the font rendering issue - we get different results when running this test if run in isolation vs if the run following another test which uses fonts.

Note that while the fonts aren't obvious in the test, we are using rectangular block characters to mask the images you can see.

While the errors themselves are small of-by-1 type erros, and would normally be rebaselined away, rebaseline isn't possible as the results change depending on the ordering of tests and whether cache cleanup has happened between tests.

To see the issue, run layout tests in the following patterns:

Scenario 1, isolation:
virtual/gpu-rasterization/fast/images/color-profile-background-clip-text.html

Scenario 2, in sequence:
virtual/gpu-rasterization/fast/images/alt-text-wrapping.html
virtual/gpu-rasterization/fast/images/color-profile-background-clip-text.html

You should see that one of these two fails.

It appears that Skia's rendering of these characters may differ slightly depending on whether the font cache is pre-populated at the time this test's glyph is added.

bsalomon@, are there any known issues where a glyph being at a different location in the atlas would cause a rendering difference (may be something else as well). Thanks!

Deleted: expected.png 23.0 KB

	expected.png 23.0 KB View Download

Deleted: actual.png 23.0 KB

	actual.png 23.0 KB View Download

Deleted: diff.png 24.9 KB

	diff.png 24.9 KB View Download

Am I correct that there is a very large glyph that is clipped? If so I'd guess that we are not using the text atlas at all but drawing the glyph as a path.

This has the same large glyph repeated 4 times. I had skia dump the atlas as a BMP earlier today and the very large glyph did appear to be in the atlas as a DF.

Attaching the two cache dumps. The glyph is a 25% height block, so although it looks like a rectangle here, I think that's the one glyph.

Deleted: existing_cache.png 14.9 KB

	existing_cache.png 14.9 KB View Download

Deleted: clean_cache.png 12.5 KB

	clean_cache.png 12.5 KB View Download

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/122df617d8e9d0856dd0baeb5db1fcfda31267de

commit 122df617d8e9d0856dd0baeb5db1fcfda31267de
Author: ericrk <ericrk@chromium.org>
Date: Mon Sep 26 22:14:14 2016

Mark color-profile-background-clip-text.html flaky

There are a couple issues with this test:
- Ahem font is not always used.
- Skia appears to rasterize content differently w fresh cache vs
  existing cache.

Marking this test as flaky while we fix these.

BUG= 624233 

Review-Url: https://codereview.chromium.org/2376523002
Cr-Commit-Position: refs/heads/master@{#421017}

[modify] https://crrev.com/122df617d8e9d0856dd0baeb5db1fcfda31267de/third_party/WebKit/LayoutTests/TestExpectations

Hm, wonder if it is just be texture sampling precision differences at different texture coords in osmesa.

Entirely possible - let me confirm whether we see differences with --use-gl=desktop

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/030f03cb171559c460ee53021bba37186035bb7e

commit 030f03cb171559c460ee53021bba37186035bb7e
Author: ericrk <ericrk@chromium.org>
Date: Thu Sep 29 00:55:05 2016

Use Ahem font in fast/images/color-profile-background-clip-text.html

We weren't pulling in the Ahem font correctly, leading to different
results on each platform (as a different fallback was chosen).

R=ccameron
BUG= 624233 

Review-Url: https://codereview.chromium.org/2368113003
Cr-Commit-Position: refs/heads/master@{#421699}

[modify] https://crrev.com/030f03cb171559c460ee53021bba37186035bb7e/third_party/WebKit/LayoutTests/TestExpectations
[modify] https://crrev.com/030f03cb171559c460ee53021bba37186035bb7e/third_party/WebKit/LayoutTests/fast/images/color-profile-background-clip-text.html

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/be6f269fec69d2ebe40382efa011bf7659a78b28

commit be6f269fec69d2ebe40382efa011bf7659a78b28
Author: Rebaseline Bot <blink-rebaseline-bot@chromium.org>
Date: Thu Sep 29 02:35:18 2016

Auto-rebaseline for r421699

https://chromium.googlesource.com/chromium/src/+/030f03cb1

BUG= 624233 
TBR=ericrk@chromium.org

Review URL: https://codereview.chromium.org/2375303002 .

Cr-Commit-Position: refs/heads/master@{#421723}

[modify] https://crrev.com/be6f269fec69d2ebe40382efa011bf7659a78b28/third_party/WebKit/LayoutTests/TestExpectations
[modify] https://crrev.com/be6f269fec69d2ebe40382efa011bf7659a78b28/third_party/WebKit/LayoutTests/fast/images/color-profile-background-clip-text-expected.txt
[add] https://crrev.com/be6f269fec69d2ebe40382efa011bf7659a78b28/third_party/WebKit/LayoutTests/platform/android/fast/images/color-profile-background-clip-text-expected.txt
[modify] https://crrev.com/be6f269fec69d2ebe40382efa011bf7659a78b28/third_party/WebKit/LayoutTests/platform/linux/fast/images/color-profile-background-clip-text-expected.png
[delete] https://crrev.com/5e5689bba0a51bad28c08bf3a3820da57a185c3b/third_party/WebKit/LayoutTests/platform/mac-mac10.10/fast/images/color-profile-background-clip-text-expected.png
[delete] https://crrev.com/5e5689bba0a51bad28c08bf3a3820da57a185c3b/third_party/WebKit/LayoutTests/platform/mac-mac10.9/fast/images/color-profile-background-clip-text-expected.png
[modify] https://crrev.com/be6f269fec69d2ebe40382efa011bf7659a78b28/third_party/WebKit/LayoutTests/platform/mac/fast/images/color-profile-background-clip-text-expected.png
[modify] https://crrev.com/be6f269fec69d2ebe40382efa011bf7659a78b28/third_party/WebKit/LayoutTests/platform/win/fast/images/color-profile-background-clip-text-expected.png

I've confirmed that we get a *very* minimal diff (1 pixel) when using desktop GL rather than osmesa - although a much smaller diff, I would still expect these to match, so likely not entirely an osmesa issue?

Build sheriff: This test crashed on Win7 dbg: http://build.chromium.org/p/chromium.webkit/builders/WebKit%20Win7%20(dbg)/builds/7868

Next build sheriff: It has been crashing every run since, with this output:

crash log for renderer (pid <unknown>):
STDOUT: ppp
STDOUT: 
STDOUT: ppp
STDERR: us bytes before marker 0xc4
STDERR: Corrupt JPEG data: 29 extraneous bytes before marker 0xc4
STDERR: Corrupt JPEG data: 29 extraneous bytes before marker 0xc4
STDERR: Corrupt JPEG data: 29 extraneous bytes before marker 0xc4
STDERR: Corrupt JPEG data: 29 extraneous bytes before marker 0xc4
STDERR: Corrupt JPEG data: 29 extraneous bytes before marker 0xc4
STDERR: Corrupt JPEG data: 29 extraneous bytes before marker 0xc4
STDERR: Corrupt JPEG data: 29 extraneous bytes before marker 0xc4
STDERR: Corrupt JPEG data: 29 extraneous bytes before marker 0xc4
STDERR: Corrupt JPEG data: 29 extraneous bytes before marker 0xc4
STDERR: Corrupt JPEG data: 29 extraneous bytes before marker 0xc4
STDERR: Corrupt JPEG data: 29 extraneous bytes before marker 0xc4
STDERR: Corrupt JPEG data: 29 extraneous bytes before marker 0xc4
STDERR: Corrupt JPEG data: 29 extraneous bytes before marker 0xc4
STDERR: Corrupt JPEG data: 29 extraneous bytes before marker 0xc4
STDERR: Corrupt JPEG data: 29 extraneous bytes before marker 0xc4
STDERR: Corrupt JPEG data: 29 extraneous bytes before marker 0xc4
STDERR: Corrupt JPEG data: 29 extraneous bytes before marker 0xc4
STDERR: Corrupt JPEG data: 29 extraneous bytes before marker 0xc4
STDERR: Corrupt JPEG data: 29 extraneous bytes before marker 0xc4
STDERR: Corrupt JPEG data: 29 extraneous bytes before marker 0xc4
STDERR: Corrupt JPEG data: 29 extraneous bytes before marker 0xc4
STDERR: Corrupt JPEG data: 29 extraneous bytes before marker 0xc4
STDERR: Corrupt JPEG data: 29 extraneous bytes before marker 0xc4
STDERR: Corrupt JPEG data: 29 extraneous bytes before marker 0xc4
STDERR: Corrupt JPEG data: 29 extraneous bytes before marker 0xc4
STDERR: Corrupt JPEG data: 29 extraneous bytes before marker 0xc4
STDERR: [4328:4332:1026/010324:480685:INFO:grdistancefieldgenfromvector.cpp(641)] e:\b\c\b\win_layout\src\third_party\skia\src\gpu\grdistancefieldgenfromvector.cpp:641: fatal error: "assert((endRow <= height) && "EndRow > height!")"
STDERR: 
STDERR: [4328:4332:1026/010324:480685:INFO:grmemorypool.cpp(35)] e:\b\c\b\win_layout\src\third_party\skia\src\gpu\grmemorypool.cpp:35: fatal error: "assert(0 == fAllocationCnt)"
STDERR: 

src/gpu/GrDistanceFieldGenFromVector.cpp was added just yesterday here:
https://codereview.chromium.org/1643143002

Which is already reverted in Skia, but a roll needs to happen.

No, I misread, it hasn't been reverted yet. I'll revert the Skia DEPS roll:
https://codereview.chromium.org/2449343002

CC current skia sheriff reed@

I can't read today, reverted the wrong roll and am now relanding that. I have to leave the office and will mark the test as crashing for now.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/573263ce1c07afc7d3bcd1a92ffc18e8c6fd7f54

commit 573263ce1c07afc7d3bcd1a92ffc18e8c6fd7f54
Author: schenney <schenney@chromium.org>
Date: Thu Apr 06 22:09:38 2017

Rebaseline some color profile tests

virtual/gpu-rasterization/images/webp-color-profile-lossy.html
is passing, it seems, and needs new baselines.

virtual/gpu-rasterization/images/color-profile-filter.html
seems to be passing everywhere already.

virtual/gpu-rasterization/images/color-profile-background-clip-text.html
is no longer crashing and produces consistent result, but those results
are wrong due to crbug.com/670292. So rebaseline until we fix the bug.

TBR=noel@chromium.org
BUG=624709,587737, 624233 

Review-Url: https://codereview.chromium.org/2794403002
Cr-Commit-Position: refs/heads/master@{#462642}

[modify] https://crrev.com/573263ce1c07afc7d3bcd1a92ffc18e8c6fd7f54/third_party/WebKit/LayoutTests/TestExpectations
[delete] https://crrev.com/205ee990e92962dbe5a0cceed476680c055c43b1/third_party/WebKit/LayoutTests/platform/linux/images/color-profile-filter-expected.png
[modify] https://crrev.com/573263ce1c07afc7d3bcd1a92ffc18e8c6fd7f54/third_party/WebKit/LayoutTests/platform/linux/virtual/gpu-rasterization/images/color-profile-background-clip-text-expected.png
[delete] https://crrev.com/205ee990e92962dbe5a0cceed476680c055c43b1/third_party/WebKit/LayoutTests/platform/linux/virtual/gpu-rasterization/images/webp-color-profile-lossy-expected.png
[modify] https://crrev.com/573263ce1c07afc7d3bcd1a92ffc18e8c6fd7f54/third_party/WebKit/LayoutTests/platform/mac-mac10.10/virtual/gpu-rasterization/images/color-profile-background-clip-text-expected.png
[delete] https://crrev.com/205ee990e92962dbe5a0cceed476680c055c43b1/third_party/WebKit/LayoutTests/platform/mac-mac10.10/virtual/gpu-rasterization/images/color-profile-filter-expected.png
[add] https://crrev.com/573263ce1c07afc7d3bcd1a92ffc18e8c6fd7f54/third_party/WebKit/LayoutTests/platform/mac-mac10.11/virtual/gpu-rasterization/images/color-profile-background-clip-text-expected.png
[modify] https://crrev.com/573263ce1c07afc7d3bcd1a92ffc18e8c6fd7f54/third_party/WebKit/LayoutTests/platform/mac-mac10.9/virtual/gpu-rasterization/images/color-profile-background-clip-text-expected.png
[add] https://crrev.com/573263ce1c07afc7d3bcd1a92ffc18e8c6fd7f54/third_party/WebKit/LayoutTests/platform/mac-retina/virtual/gpu-rasterization/images/color-profile-background-clip-text-expected.png
[delete] https://crrev.com/205ee990e92962dbe5a0cceed476680c055c43b1/third_party/WebKit/LayoutTests/platform/mac-retina/virtual/gpu-rasterization/images/color-profile-filter-expected.png
[modify] https://crrev.com/573263ce1c07afc7d3bcd1a92ffc18e8c6fd7f54/third_party/WebKit/LayoutTests/platform/mac/virtual/gpu-rasterization/images/color-profile-background-clip-text-expected.png
[modify] https://crrev.com/573263ce1c07afc7d3bcd1a92ffc18e8c6fd7f54/third_party/WebKit/LayoutTests/platform/mac/virtual/gpu-rasterization/images/color-profile-filter-expected.png
[modify] https://crrev.com/573263ce1c07afc7d3bcd1a92ffc18e8c6fd7f54/third_party/WebKit/LayoutTests/platform/win/virtual/gpu-rasterization/images/color-profile-background-clip-text-expected.png
[modify] https://crrev.com/573263ce1c07afc7d3bcd1a92ffc18e8c6fd7f54/third_party/WebKit/LayoutTests/platform/win/virtual/gpu-rasterization/images/webp-color-profile-lossy-expected.png
[add] https://crrev.com/573263ce1c07afc7d3bcd1a92ffc18e8c6fd7f54/third_party/WebKit/LayoutTests/platform/win7/virtual/gpu-rasterization/images/color-profile-background-clip-text-expected.png
[delete] https://crrev.com/205ee990e92962dbe5a0cceed476680c055c43b1/third_party/WebKit/LayoutTests/platform/win7/virtual/gpu-rasterization/images/webp-color-profile-lossy-expected.png

fixed?