Bug 97465 – test-webkitpy fails on windows when running with more than one child process

• Component: Tools / Tests
• Status: RESOLVED FIXED

[Lucas Forschler]

On the mac windows debug test bot, the python tests are constantly failing.

I've noticed that passing in -j 1 (to run with only one child process) seems to fix the issue.  the stack trace is as follows:


$ python ./Tools/Scripts/test-webkitpy -v -j 2
Suppressing most webkitpy logging while running unit tests.
Skipping tests in the following modules or packages because they are really,
lly, slow:
    webkitpy.common.checkout.scm.scm_unittest
    (https://bugs.webkit.org/show_bug.cgi?id=31818; use --all to include)

Running the tests ...Traceback (most recent call last):
  File "/home/buildbot/slave/windows-debug-tests/build/OpenSource/Tools/Scri
webkitpy/test/main.py", line 227, in <module>
    sys.exit(main())
  File "/home/buildbot/slave/windows-debug-tests/build/OpenSource/Tools/Scri
webkitpy/test/main.py", line 69, in main
    return not tester.run()
  File "/home/buildbot/slave/windows-debug-tests/build/OpenSource/Tools/Scri
webkitpy/test/main.py", line 119, in run
    return self._run_tests(names)
  File "/home/buildbot/slave/windows-debug-tests/build/OpenSource/Tools/Scri
webkitpy/test/main.py", line 154, in _run_tests
    test_runner.run(parallel_tests, self._options.child_processes)
  File "/home/buildbot/slave/windows-debug-tests/build/OpenSource/Tools/Scri
webkitpy/test/runner.py", line 52, in run
    with message_pool.get(self, self.worker_factory, num_workers) as pool:
  File "/home/buildbot/slave/windows-debug-tests/build/OpenSource/Tools/Scri
webkitpy/common/message_pool.py", line 61, in get
    return _MessagePool(caller, worker_factory, num_workers, worker_startup_
y_secs, host)
  File "/home/buildbot/slave/windows-debug-tests/build/OpenSource/Tools/Scri
webkitpy/common/message_pool.py", line 79, in __init__
    self._messages_to_worker = multiprocessing.Queue()
  File "/usr/lib/python2.6/multiprocessing/__init__.py", line 213, in Queue
    return Queue(maxsize)
  File "/usr/lib/python2.6/multiprocessing/queues.py", line 37, in __init__
    self._rlock = Lock()
  File "/usr/lib/python2.6/multiprocessing/synchronize.py", line 117, in __i
_
    SemLock.__init__(self, SEMAPHORE, 1, 1)
  File "/usr/lib/python2.6/multiprocessing/synchronize.py", line 49, in __in

    sl = self._semlock = _multiprocessing.SemLock(kind, value, maxvalue)
OSError: [Errno 17] File exists

[Lucas Forschler]

On the one mac bot we have, running with j=1 takes about 70 seconds.  It is not significantly faster running parallel (around 40 seconds), so we could turn off parallel running to get the bot back online.

[Lucas Forschler]

Created attachment 165412 [details]
Tell Windows bot to use only one child process

[Tim Horton]

Comment on attachment 165412 [details]
Tell Windows bot to use only one child process

Is there a bug to actually fix the failure?

[WebKit Review Bot]

Comment on attachment 165412 [details]
Tell Windows bot to use only one child process

Clearing flags on attachment: 165412

Committed r129561: <http://trac.webkit.org/changeset/129561>

[WebKit Review Bot]

All reviewed patches have been landed.  Closing bug.