Replace SpecificityCalculator with TestConfiguration-driven logic.
Created attachment 102049 [details] Patch
Comment on attachment 102049 [details] Patch Clearing flags on attachment: 102049 Committed r91792: <http://trac.webkit.org/changeset/91792>
All reviewed patches have been landed. Closing bug.
This patch broke expectations parsing in GTK and Qt. Investigating the issue, any hint Dimitri?
Created attachment 102116 [details] proposed GTK follow-up patch Something similar is needed for Qt.
Reopening as the patch was rolled out. See Bug 65242.
Comment on attachment 102049 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=102049&action=review > Tools/Scripts/webkitpy/layout_tests/models/test_expectations.py:777 > + if self._test_config in expectation_line.matching_configurations: The root issue of GTK and Qt failure is here. Those ports originally don't have the version field of their test config set and this test fails for all expectations.
Comment on attachment 102116 [details] proposed GTK follow-up patch Your sense of smell is correct, the problem was that GTK actual bot configs aren't matching anything in the provided all_test_configurations(). The right fix is to teach all ports about all configs that are valid for them. I'll work on this today.
Committed r91878: <http://trac.webkit.org/changeset/91878>
(In reply to comment #9) > Committed r91878: <http://trac.webkit.org/changeset/91878> Looks like it worked this time! Phillippe, Ossy, please don't hesitate to roll out if you find troubles.