REGRESSION (r211845): [ios-simulator] LayoutTest compositing/masks/solid-color-masked.html is a flaky failure

[Ryan Haddad]

[ios-simulator] LayoutTest compositing/masks/solid-color-masked.html is a flaky failure

https://build.webkit.org/results/Apple%20iOS%2010%20Simulator%20Release%20WK2%20(Tests)/r211961%20(3594)/results.html

https://webkit-test-results.webkit.org/dashboards/flakiness_dashboard.html#showAllRuns=true&tests=compositing%2Fmasks%2Fsolid-color-masked.html

Direct link to images and diff: https://build.webkit.org/results/Apple%20iOS%2010%20Simulator%20Release%20WK2%20(Tests)/r211961%20(3594)/compositing/masks/solid-color-masked-diffs.html

[Ryan Haddad]

Flakiness dashboard suggests that this started with https://trac.webkit.org/changeset/211845

[Simon Fraser (smfr)]

I have a hunch about what the problem is here.

[Ryan Haddad]

Marked test as failing in http://trac.webkit.org/projects/webkit/changeset/212123

[Simon Fraser (smfr)]

I get:

 463$ $ wktests --debug --ios-simulator LayoutTests/compositing/masks/ --iterations=10
Using port 'ios-simulator'
Test configuration: <, x86_64, debug>
Placing test results in /Volumes/Data/Development/OSX/webkit/OpenSource/WebKitBuild/Debug-iphonesimulator/layout-test-results
Baseline search path: platform/ios-simulator-wk2 -> platform/ios-simulator -> platform/wk2 -> generic
Using Debug build
Pixel tests disabled
Regular timeout: 30000, slow test timeout: 150000
Command line: /Volumes/Data/Development/OSX/webkit/OpenSource/WebKitBuild/Debug-iphonesimulator/WebKitTestRunnerApp.app -

Found 20 tests; running 18 (10 times each: --repeat-each=1 --iterations=10), skipping 2.

Running 18 tests

This machine could support 12 simulators, but is only configured for 2.
Please see <https://trac.webkit.org/wiki/IncreasingKernelLimits>.
Running 1 WebKitTestRunnerApp.app.

[16/180] compositing/masks/solid-color-masked.html passed unexpectedly                               
[34/180] compositing/masks/solid-color-masked.html passed unexpectedly                                
[52/180] compositing/masks/solid-color-masked.html passed unexpectedly                                
[70/180] compositing/masks/solid-color-masked.html passed unexpectedly                                
[88/180] compositing/masks/solid-color-masked.html passed unexpectedly                                
[106/180] compositing/masks/solid-color-masked.html passed unexpectedly                               
[124/180] compositing/masks/solid-color-masked.html passed unexpectedly                                
[142/180] compositing/masks/solid-color-masked.html passed unexpectedly                                
[160/180] compositing/masks/solid-color-masked.html passed unexpectedly                                
[178/180] compositing/masks/solid-color-masked.html passed unexpectedly                                
stopping WebKitTestRunnerApp.app(pid 15605) timed out, killing it

170 tests ran as expected, 10 didn't:


Expected to fail, but passed: (1)
  compositing/masks/solid-color-masked.html

[Simon Fraser (smfr)]

Ah, i can reproduce this on first MobileSafari launch.

[Simon Fraser (smfr)]

Ah, we never run the code at the top of GraphicsLayerCA::recursiveCommitChanges() for mask layers.

[Simon Fraser (smfr)]

Actually GraphicsLayerCA::setVisibleAndCoverageRects() copies the m_intersectsCoverageRect flag to the mask layer.

The bug happens when GraphicsLayerCA::setVisibleAndCoverageRects() is called for the masked layer before it has a mask. After than, assigning the mask doesn't ever update the mask's m_intersectsCoverageRect.

[Radar WebKit Bug Importer]

<rdar://problem/30476807>

[Simon Fraser (smfr)]

Created attachment 301228 [details]
Patch

[Simon Fraser (smfr)]

https://trac.webkit.org/r212172

[Jonathan Bedard]

Someone forgot to remove this test expectation.

Removed in https://trac.webkit.org/r215324.