webkitpy.layout_tests.layout_package.manager_worker_broker_unittest.FunctionTests.test_get__processes is sometimes failing on Windows. For example: Did not fail: http://build.webkit.org/builders/Windows%207%20Release%20%28Tests%29/builds/9673 Did fail: http://build.webkit.org/builders/Windows%207%20Release%20%28Tests%29/builds/9679 So far I've only seen it fail on apple-windows-5, which is running Python 2.6.5. I believe apple-windows-6 is running Python 2.5.x.
Hm. That test is showing python-level failures in the "multiprocessing" module, which is only available under Python 2.6. So that test won't run (or fail) under 2.5. We'll have to watch it to see how flaky it is ...
Another failure: http://build.webkit.org/builders/Windows%207%20Release%20%28Tests%29/builds/9691/steps/webkitpy-test/logs/stdio
More failures: http://build.webkit.org/builders/Windows%207%20Release%20%28Tests%29/builds/9662 http://build.webkit.org/builders/Windows%207%20Release%20%28Tests%29/builds/9671 So it failed 4 times from build 9662 to build 9691, inclusive. That's once every 7 builds or so.
<rdar://problem/9059908>
We've been seeing cases where lots of python processes are left around on the bots. I wonder if this bug is the/a cause of that?
(In reply to comment #5) > We've been seeing cases where lots of python processes are left around on the bots. I wonder if this bug is the/a cause of that? It's quite possible. When I briefly had the multiprocessing module enabled for new-run-webkit-tests, we also seemed to see a bunch of processes lying around. I think I'll disable this test for now until I can do some more testing on Windows and see if there are bugs I can reproduce/triage/fix.
Assigning to myself and skipping for now until I can triage it further.
Created attachment 84152 [details] Patch
I don't like skipping tests ... anyone have any ideas before I disable it?
Skipped the test on Windows to get the bots green Committed r79993: <http://trac.webkit.org/changeset/79993>
http://trac.webkit.org/changeset/79993 might have broken GTK Linux 32-bit Debug
clearing ownership now that I've disabled the bug, in the off chance that someone else wants to take this.
Attachment 84152 [details] was posted by a committer and has review+, assigning to Dirk Pranke for commit.
It looks like aroben landed this in r79993. Closing.
Has this been further investigated? I was looking at eliminating more uses of sys.platform and came across this one.
Nope, I haven't had a chance to look into it further, and it's not super-high on my priority list.