Bug 78370

Summary:

[chromium] compositing/rubberbanding tests failing

Product:

WebKit

Reporter:

Tony Chang <tony>

Component:

Tools / Tests

Assignee:

asvitkine

Status:

RESOLVED FIXED

Severity:

Normal

CC:

asvitkine, jamesr, webkit.review.bot

Priority:

Version:

528+ (Nightly build)

Hardware:

Unspecified

OS:

Unspecified

Attachments:

Description	Flags
Patch	none
Patch	none

Tony Chang

Reported 2012-02-10 12:09:10 PST

http://test-results.appspot.com/dashboards/flakiness_dashboard.html#tests=platform%2Fchromium%2Fcompositing%2Frubberbanding%2Ftransform-overhang-nw.html%2Cplatform%2Fchromium%2Fcompositing%2Frubberbanding%2Ftransform-overhang-sw.html%2Cplatform%2Fchromium%2Fcompositing%2Frubberbanding%2Ftransform-overhang-n.html%2Cplatform%2Fchromium%2Fcompositing%2Frubberbanding%2Ftransform-overhang-ne.html%2Cplatform%2Fchromium%2Fcompositing%2Frubberbanding%2Ftransform-overhang-size-change.html%2Cplatform%2Fchromium%2Fcompositing%2Frubberbanding%2Ftransform-overhang-w.html There has been a lot of test flipping that makes it hard for me to tell what caused the most recent failure. According to the flakiness dashboard, the regression is in this range: http://trac.webkit.org/log/?verbose=on&rev=107345&stop_rev=107313 The most likely change is http://trac.webkit.org/changeset/107335/ or http://trac.webkit.org/changeset/107329/ .

Attachments
Patch (4.77 KB, patch) 2012-02-16 08:47 PST, asvitkine	no flags	Details Formatted Diff Diff
Patch (15.73 KB, patch) 2012-02-16 11:18 PST, asvitkine	no flags	Details Formatted Diff Diff
Show Obsolete (1) View All Add attachment proposed patch, testcase, etc.

Tony Chang

Comment 1 2012-02-10 12:09:29 PST

James can you take a look? I'm going to mark the tests as failing for now.

Tony Chang

Comment 2 2012-02-10 12:12:27 PST

See also bug 78008.

asvitkine

Comment 3 2012-02-13 15:37:44 PST

Looking at the flaky runs of green/non-green results, the non-green results have tiny pixel differences in the rubber band gradients only. Is that a known cause of flakyness with the compositing path? It seems the non-compositing rubber-banding tests don't have this flakyness in the gradients. And then there's the more recent regression that just causes them to fail all the time due to not pattern/gradient area being drawn at all...

asvitkine

Comment 4 2012-02-16 07:06:56 PST

Pretty sure it's the mock scrollbars change. Rubber-band overhang painting is done by ScrollbarThemeChromiumMac.mm, while the mock scrollbar changes replaces this by ScrollbarThemeMock.

asvitkine

Comment 5 2012-02-16 08:47:02 PST

Created attachment 127389 [details] Patch

James Robinson

Comment 6 2012-02-16 10:52:33 PST

Comment on attachment 127389 [details] Patch If you want these specific tests to not use mock scrollbars, then just call window.internals.settings.setMockScrollbarsEnabled(false) in the tests

asvitkine

Comment 7 2012-02-16 11:18:17 PST

Created attachment 127415 [details] Patch

asvitkine

Comment 8 2012-02-16 11:19:03 PST

Comment on attachment 127415 [details] Patch New patch that just changes the tests to disable mock scrollbars.

James Robinson

Comment 9 2012-02-16 11:19:27 PST

Comment on attachment 127415 [details] Patch r=me

WebKit Review Bot

Comment 10 2012-02-16 16:49:58 PST

Comment on attachment 127415 [details] Patch Clearing flags on attachment: 127415 Committed r108000: <http://trac.webkit.org/changeset/108000>

WebKit Review Bot

Comment 11 2012-02-16 16:50:07 PST

All reviewed patches have been landed. Closing bug.

Note You need to log in before you can comment on or make changes to this bug.