GTK: - before: https://build.webkit.org/builders/GTK%20Linux%2064-bit%20Release%20%28Tests%29/builds/6464 - after: https://build.webkit.org/builders/GTK%20Linux%2064-bit%20Release%20%28Tests%29/builds/6465 EFL: - before: https://build.webkit.org/builders/EFL%20Linux%2064-bit%20Release%20WK2/builds/20006 - after: https://build.webkit.org/builders/EFL%20Linux%2064-bit%20Release%20WK2/builds/20021 Unfortunately there were build break between r180585-r180603 due to the "awesome" WK2 owner model. But I checked it manually, and fast/workers tests pass on r180601 with the buildfix applied, so r180602 must be the culprit.
I skipped fast/workers tests on EFL bots, because it increased the testing time from 15 minutes to 35 minutes, and then the bot died with timeout. Skip landed in http://trac.webkit.org/changeset/180671 .
One of the side effects of r180602 is that it results in more threads being scanned by GC. Specifically, that means a VM on the main thread will also scan the stacks of all worker threads because they have had a VM run on them before. This means they are doing more work, though we had anticipated that it shouldn’t be that costly. I’ll prepare a test patch to see if eliminating / reducing this unnecessary work makes a difference.
FYI, the r180602 patch was rolled out in r180690: <http://trac.webkit.org/r180690>.
With the solution for https://bugs.webkit.org/show_bug.cgi?id=141990, the GTK test bot seems to be happy. The previously skipped tests have been un-skipped when I rolled out r180690 earlier. Nothing more to do. Closing this bug.