It's because WTF_Lock tests, they actually pass, but they are slower than our timeout. $ time bin/TestWebKitAPI/WTF/TestWTF --gtest_filter=WTF_Lock.* **PASS** WTF_Lock.UncontentedShortSection **PASS** WTF_Lock.UncontentedLongSection **PASS** WTF_Lock.ContentedShortSection **PASS** WTF_Lock.ContentedLongSection **PASS** WTF_Lock.ManyContentedShortSections **PASS** WTF_Lock.ManyContentedLongSections real 0m15.656s user 0m51.736s sys 0m36.848s We need a way to mark tests as slow and use a longer timeout like WTR does. I don't know if these tests are supposed to be so slow, though. Filip?
Created attachment 258479 [details] Patch Marking WTF_Lock.ContentedShortSection as slow seems to be enough
Comment on attachment 258479 [details] Patch Alternatively you could just make the tests run for less time. In each test, the number of iterations is the last argument to the runTest template.
Committed r188125: <http://trac.webkit.org/changeset/188125>
Unfortunately this was not enough for the bots, that seem to be slower than my laptop: $ time (../../Tools/jhbuild/jhbuild-wrapper --gtk run xvfb-run bin/TestWebKitAPI/WTF/TestWTF --gtest_filter=WTF_Lock.*) **PASS** WTF_Lock.UncontentedShortSection **PASS** WTF_Lock.UncontentedLongSection **PASS** WTF_Lock.ContentedShortSection **PASS** WTF_Lock.ContentedLongSection **PASS** WTF_Lock.ManyContentedShortSections **PASS** WTF_Lock.ManyContentedLongSections real 0m43.051s user 3m32.349s sys 1m44.799s
(In reply to comment #4) > Unfortunately this was not enough for the bots, that seem to be slower than > my laptop: > > $ time (../../Tools/jhbuild/jhbuild-wrapper --gtk run xvfb-run > bin/TestWebKitAPI/WTF/TestWTF --gtest_filter=WTF_Lock.*) > **PASS** WTF_Lock.UncontentedShortSection > **PASS** WTF_Lock.UncontentedLongSection > **PASS** WTF_Lock.ContentedShortSection > **PASS** WTF_Lock.ContentedLongSection > **PASS** WTF_Lock.ManyContentedShortSections > **PASS** WTF_Lock.ManyContentedLongSections > > real 0m43.051s > user 3m32.349s > sys 1m44.799s Increased the timeout for slow tests to timeout * 5 and it stills times out . . . Maybe we should just try to make the test faster.