The following test is failing in latest builds of the "GTK Linux 32-bit Release" and "GTK Linux 64-bit Release" bots. See the current diff below: --- /var/lib/buildbot/build/gtk-linux-32-release/build/layout-test-results/plugins/refcount-leaks-expected.txt 2010-12-14 04:27:14.000000000 -0800 +++ /var/lib/buildbot/build/gtk-linux-32-release/build/layout-test-results/plugins/refcount-leaks-actual.txt 2010-12-14 04:27:14.000000000 -0800 @@ -7,6 +7,6 @@ --- refcount on plug.testObject: originally: 2 -after GC: 2 +after GC: 3 after passing: 2 SUCCESS Looks like this test comes from the Chromium port (see revisions 52462 and 72302) and they started failing recently in those two bots because of some (unknown yet) reason. The only piece of (hopefully) useful information I can give at this point is that it seems the test started failing at some point between revisions 73569 and 73580, as far as I can check it from the output logs in the buildbots. So the problem must be originated somewhere between those 11 commits. Reporting this patch at the same time I'm skipping that test in the gtk platform, till we found a proper solution for it.
This test is flaky. It often fails for one or two runs and then passes. I've CC'd the original author.