Created attachment 139522 [details] Patch

Attachment 139522 [details] did not pass style-queue: Failed to run "['Tools/Scripts/check-webkit-style', '--diff-files', u'LayoutTests/ChangeLog', u'LayoutTests/iete..." exit_code: 1 LayoutTests/platform/gtk/test_expectations.txt:517: More specific entry on line 330 overrides line 517 fast/workers/storage/use-same-database-in-page-and-workers.html [test/expectations] [5] Total errors found: 1 in 54 files If any of these errors are false positives, please file a bug against check-webkit-style.

Style failure due to existing issue in gtk/test_expectations.txt : 517: More specific entry on line 330 overrides line 517 fast/workers/storage/use-same-database-in-page-and-workers.html. I'm inclined to let the gtk gardener deal with this rather than attempting a fix myself, but am open to suggestion.

Notes on this patch: - 24 total tests, 3 fail. - bugs have been written for the failures - All tests have correct reference tests - EWS failure due to existing problem in gtk test_expectations.txt, and also due to churn in the chromium test_expectations.txt. I will deal with these issues when I land the patch.

Created attachment 139893 [details] Patch

(In reply to comment #5) > Created an attachment (id=139893) [details] > Patch Rebased to include fix (by koz) to gtk test_expectations.txt so we don't faile the style check.

Committed r116018: <http://trac.webkit.org/changeset/116018>

None of these reference tests pass on the Mac. The test looks visually identical to the reference, but ImageDiff detects a difference between the color drawn by background-image and the color in the reference image. Might this be because the images have an sRGB color profile?

I filed <https://bugs.webkit.org/show_bug.cgi?id=85666> about this.

I had assumed the green in the pngs used in the MS tests were the same color and background "green". In fact, all my testing on my system (Linux: QT,GTK, and Chromium) went just fine. It's weird that QT on the bots also failed on minor pixel differences (can't repro this locally). Anyway, I've reworked the reference tests to use the pngs where necessary, so the pixel differences should disappear. I have repro'ed the issue on another developer's Mac, and will upload a patch as soon as I can get him to test the new ref tests.