Maciej committed a debug print accidentally (http://trac.webkit.org/changeset/75682) and it made all layout tests fail and buildbots sick. The size of results for ~22000 failing layout tests is more than 100Mb. I propose some changes to avoid overloading buildbot slaves and master: 1.) add --exit-after-n-failures 500 option to the old-run-webkit-tests It is enough for online rebaselineing and avoiding 100Mb sizes results 2.) old-run-webkit-tests should remove results directory before testing (Now test-result-archive does it after zipping. But if it is terminated or killed, the results directory will contain some mess.) Patches are coming soon.
Created attachment 78798 [details] old-run-webkit-tests fix
Created attachment 78799 [details] exit-after-n-failures 500 patch
Comment on attachment 78798 [details] old-run-webkit-tests fix Seems OK. As long as rmtree is available in all the versions of perl we support, and error handling works. I’ll be optimistic.
Comment on attachment 78799 [details] exit-after-n-failures 500 patch View in context: https://bugs.webkit.org/attachment.cgi?id=78799&action=review Not great to attach two patches to the same bug. > Tools/BuildSlaveSupport/build.webkit.org-config/master.cfg:196 > + WithProperties("--%(configuration)s"), "--exit-after-n-crashes-or-timeouts", "20", "--exit-after-n-failures", "500"] Would be slightly better to have a comment explaining the reason for this. There’s also an extra space after the comma.
Comment on attachment 78798 [details] old-run-webkit-tests fix Clearing flags on attachment: 78798 Committed r75721: <http://trac.webkit.org/changeset/75721>
Comment on attachment 78799 [details] exit-after-n-failures 500 patch Clearing flags on attachment: 78799 Committed r75726: <http://trac.webkit.org/changeset/75726>
All reviewed patches have been landed. Closing bug.
http://trac.webkit.org/changeset/75726 might have broken Leopard Intel Release (Tests)