Bug 45991 – [chromium] Unskip the canvas philip suite

• Component: New Bugs
• Status: RESOLVED WONTFIX

[James Robinson]

The philip canvas test suite was marked as SKIP in chromium's test_expectations.txt when it was initially checked in.  We should be running these tests, there's no reason to skip them.

[James Robinson]

canvas/philip contains 782 tests.  On my linux box 97 fail and 2 are flaky.  The rest pass with the existing baselines.

http/tests/canvas/philip contains 10 tests, all of which pass locally.

[James Robinson]

Created attachment 67942 [details]
Patch

[Dimitri Glazkov (Google)]

Comment on attachment 67942 [details]
Patch

Weeee!

[James Robinson]

Committed r67744: <http://trac.webkit.org/changeset/67744>

[James Robinson]

Created attachment 67989 [details]
specific expectations

[Andreas Kling]

Comment on attachment 67989 [details]
specific expectations

Awesome stuff!

[James Kozianski]

Created attachment 68054 [details]
Patch

[James Kozianski]

Comment on attachment 68054 [details]
Patch

This just takes the mac tests from the above patch (which seems to be out of date) as they are failing on the canaries right now.

[James Robinson]

This is not right - I intentionally left the entire suite as = TEXT until I had a chance to land the 'specific expectations' patch.  Looks like abarth messed up the expectations in http://trac.webkit.org/changeset/67820 and the followup.  For shame, Adam!

[James Robinson]

A bunch fail on windows still.  I recommend reverting the entire section specific expectations patch, which contains a list of the tests that fail on each platform as of last Friday.

I won't be able to look at this for ~10 hours more.

[Adam Barth]

> For shame, Adam!

Huh?  I'm not sure what I did wrong.  Maybe my attempt to help garden was misguided.

[Ojan Vafai]

(In reply to comment #9)
> This is not right - I intentionally left the entire suite as = TEXT until I had a chance to land the 'specific expectations' patch.  Looks like abarth messed up the expectations in http://trac.webkit.org/changeset/67820 and the followup.  For shame, Adam!

James, I'm not sure the history of this set of expectations, but someone else went ahead and took away the test suite listing. This patch just cleaned up some of the fallout on the mac bot. As best I can tell, there's only one failing test left on the webkit chromium canaries and thats canvas/philip/tests/2d.gradient.radial.touch2.html on the windows bot.

[James Robinson]

James' patch looks fine, I'm just frustrated that you guys are having to clean up expectations after I generated expectations for each platform already.  I can clean up the remaining fallout.

[James Robinson]

Committed r67872: <http://trac.webkit.org/changeset/67872>

[James Robinson]

Comment on attachment 67989 [details]
specific expectations

clearing flags, obsolete

[James Robinson]

Comment on attachment 68054 [details]
Patch

Clearing flags, obsolete

[James Robinson]

Not resolved.  The remaining failures need to be triaged into separate bugs.

[Justin Novosad]

Change set http://trac.webkit.org/changeset/81489 resolved the following layout test failures on windows and linux, which were associated with this bug: 

canvas/philip/tests/2d.shadow.pattern.alpha.html
canvas/philip/tests/2d.shadow.pattern.basic.html 
canvas/philip/tests/2d.shadow.pattern.transparent.2.html

[Marcus Bulach]

(In reply to comment #18)
> Change set http://trac.webkit.org/changeset/81489 resolved the following layout test failures on windows and linux, which were associated with this bug: 
> 
> canvas/philip/tests/2d.shadow.pattern.alpha.html
> canvas/philip/tests/2d.shadow.pattern.basic.html 
> canvas/philip/tests/2d.shadow.pattern.transparent.2.html

Hi,

I was gardening today, and 
canvas/philip/tests/2d.composite.canvas.source-over.html
started failing on Windows GPU.

Added a TEXT failure baseline at:
http://trac.webkit.org/changeset/88162

and filed a bug with more details at:
http://code.google.com/p/chromium/issues/detail?id=85051

Is anyone familiar with this test? Thanks!

[Adam Klein]

The Skia roll in http://crrev.com/111151 caused the following tests to start passing:

canvas/philip/tests/2d.shadow.canvas.alpha-expected.txt
canvas/philip/tests/2d.shadow.canvas.basic-expected.txt
canvas/philip/tests/2d.shadow.canvas.transparent.2-expected.txt
canvas/philip/tests/2d.shadow.image.alpha-expected.txt
canvas/philip/tests/2d.shadow.image.basic-expected.txt
canvas/philip/tests/2d.shadow.image.scale-expected.txt
canvas/philip/tests/2d.shadow.image.transparent.2-expected.txt

[Stephen Chenney]

LayoutTest failures for Chromium are being marked WontFix. The Bug is still accessible and referenced from TestExpectations.