Update the testing infrastructure to support the WinCairo port. This consists of: 1. Teaching "old-run-webkit-tests" to recognize when it is being asked to test a WinCairo build. 2. Add a "platform/wincairo" directory and a "Skipped" list. 3. Populate the "Skipped" list with the current set of failing WinCairo tests so that the build bot does not turn red when we activate tests.
This first patch simply adds a 'wincairo' platform directory to the LayoutTests/platform tree, and a new Skipped file that instructs the test infrastructure to ignore all tests that currently fail when running the WinCairo build.
Created attachment 109699 [details] Patch
Created attachment 109733 [details] Create a Skiplist for WinCairo.
Created attachment 109735 [details] Update old-build-webkit-tests to know about wincairo
Comment on attachment 109735 [details] Update old-build-webkit-tests to know about wincairo View in context: https://bugs.webkit.org/attachment.cgi?id=109735&action=review > Tools/Scripts/old-run-webkit-tests:2479 > - > + Might as well undo this. > Tools/Scripts/webkitdirs.pm:238 > if ($configuration && isWinCairo()) { > - $configuration .= "_Cairo_CFLite"; > + unless ($configuration =~ /_Cairo_CFLite$/) { > + $configuration .= "_Cairo_CFLite"; > + } > } Was this always buggy? Why hadn't we run into it before?
Comment on attachment 109733 [details] Create a Skiplist for WinCairo. You can list directories in the Skipped file to skip all tests in that directory and its subdirectories. Maybe that would be a simpler way to start out? (And it would guard you against new tests being added in a directory that you haven't un-skipped yet.)
Comment on attachment 109735 [details] Update old-build-webkit-tests to know about wincairo View in context: https://bugs.webkit.org/attachment.cgi?id=109735&action=review >> Tools/Scripts/webkitdirs.pm:238 >> } > > Was this always buggy? Why hadn't we run into it before? I don't think the code had been executed previously. I ran across this when passing the command line "--wincairo" argument to the test tool. I think an earlier step generates the configuration name based on the build (which causes the _Cairo_CFLite to be affixed to the end). Then, the test harness processes the name again and added a second application of the suffix. There is probably a cleaner way to address this...
Committed r96744: <http://trac.webkit.org/changeset/96744>