Bug 89534 – [Chromium][Win] Missing trailing newlines in one text test expectation, debug only

• Status: RESOLVED
• Resolution: FIXED
• Component: Tools / Tests

[Kenneth Russell]

platform/chromium/virtual/gpu/fast/canvas/resize-while-save-active.html is failing on Windows Debug only (not Windows Release) because the text test expectations are missing the few trailing newlines. It's very strange. I updated the test expectations in http://trac.webkit.org/changeset/120758 , but this only shifted the failure from the Debug to the Release bot. Unfortunately I haven't been able to build on Windows today so haven't been able to investigate the failure at all.

Flakiness dashboard:

http://test-results.appspot.com/dashboards/flakiness_dashboard.html#tests=platform%2Fchromium%2Fvirtual%2Fgpu%2Ffast%2Fcanvas%2Fresize-while-save-active.html

Example failing build:

http://build.chromium.org/p/chromium.webkit/builders/Webkit%20Win/builds/15179

That's the first build containing the updated test expectations. The last failing Debug build before those expectations were checked in:

http://build.chromium.org/p/chromium.webkit/builders/Webkit%20Win%20%28dbg%29%282%29/builds/10642

[Dirk Pranke]

I have repro'ed the failure locally in debug build; I am now rebuilding my release build as well to make sure that passes ...

[Dirk Pranke]

okay, reproduced that this is passing in release. Also, I should not that I'm reproducing this on win 7, so it probably fails on all debug win. I think we just have debug xp bots, though (should probably change that).

[Dirk Pranke]

Okay, it looks like what is going on is that in release mode we're getting an extra empty text box inserted into the end of the render tree which isn't getting coalesced, and this is resulting in the extra newlines. Apparently debug is slow enough that you're getting a chance to yield in the parser and that is letting the block get coalesced -> no extra newlines.

Looks like this was caused by bug 86165, which changed the yielding logic.

Tony, Shrey, I'll let you decide whether you want to revert bug 86165 or do something else here, since this might be the only test that failed.

[Kenneth Russell]

Thanks Dirk for tracking down the cause of this crazy bug.

[Tony Gentilcore]

Created attachment 148696 [details]
Patch

[Tony Gentilcore]

Yeah, great find and sorry for the trouble. I believe this patch makes the render tree deterministic now although I haven't tested on win debug and wasn't able to repro the failure.

[Kenneth Russell]

Comment on attachment 148696 [details]
Patch

BTW, could you remove the associated line from LayoutTests/platform/chromium/TestExpectations along with this or any other patch? Thanks in advance.

[Tony Gentilcore]

Created attachment 148700 [details]
Patch

[Kenneth Russell]

Comment on attachment 148700 [details]
Patch

Thanks. LGTM FWIW.

[WebKit Review Bot]

Comment on attachment 148700 [details]
Patch

Rejecting attachment 148700 [details] from commit-queue.

Failed to run "['/mnt/git/webkit-commit-queue/Tools/Scripts/webkit-patch', '--status-host=queues.webkit.org', '-..." exit_code: 2

Last 500 characters of output:
z 3.
patching file LayoutTests/fast/canvas/resize-while-save-active-expected.txt
patching file LayoutTests/fast/canvas/resize-while-save-active.html
patching file LayoutTests/platform/chromium/TestExpectations
Hunk #1 FAILED at 3710.
1 out of 1 hunk FAILED -- saving rejects to file LayoutTests/platform/chromium/TestExpectations.rej

Failed to run "[u'/mnt/git/webkit-commit-queue/Tools/Scripts/svn-apply', u'--force', u'--reviewer', u'Kenneth Ru..." exit_code: 1 cwd: /mnt/git/webkit-commit-queue/

Full output: http://queues.webkit.org/results/13039250

[Tony Gentilcore]

Created attachment 148939 [details]
Patch

[Tony Gentilcore]

Rebased and carried forward kbr's r+

[WebKit Review Bot]

Comment on attachment 148939 [details]
Patch

Clearing flags on attachment: 148939

Committed r120999: <http://trac.webkit.org/changeset/120999>

[WebKit Review Bot]

All reviewed patches have been landed.  Closing bug.