On Sun, Sep 8, 2013 at 5:29 PM, Nicholas Nethercote <n.netherc...@gmail.com> wrote: > > I want to land https://bugzilla.mozilla.org/show_bug.cgi?id=910517, > which is just a clean-up of memory reporters (and is blocking a bunch > of follow-up work.) But I'm blocked by some baffling time-outs > occurring only on 32-bit opt Linux builds. (32-bit debug Linux builds > are fine.) > > About 90% of mochitest-1 runs hit this time-out. The problem is that > in a few tests this error from TestRunner.js is hit: > > Error: Unable to restore focus, expect failures and timeouts. > > and then the test times out.
Turns out the problem was a debug.js NS_ASSERT failure. Thanks to karlt for running the tests on 32-bit Linux (at least, I think that's what he did) and showing me the failure occurrence. Ed Morley filed https://bugzilla.mozilla.org/show_bug.cgi?id=914092 to make these debug.js NS_ASSERT failures cause tests to fail in a more obvious and reliable manner. > In order to try to debug it, I've been attempting to do 32-bit Linux > builds. > >> (The full list is below.) I apparently have a copy of a 32-bit >> libgtk-x11-2.0.so: > > In fact, you *don't* have a copy of a 32-bit libgtk-x11-2.0.so, which is > required for linking. You have libgtk-x11-2.0.so.0*. Those won't be used > by the linker. True! I tried adding symlinks for the missing filenames. It was enough to get past these errors, but then I got other link errors for missing symbols... but now that the test failure problem is solved I don't need these builds any more, so I won't pursue it further. Nick _______________________________________________ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform