I've seen this several times on our builders, and keep forgetting to work it out. Therefore filing bug so I remember. The details. Using the following versions: mozrunner: 2.5.2 jsbridge: 2.4.1 mozmill: 1.5.1 MozMill fails to run properly on the Thunderbird tinderboxes. Typical errors are: INFO | (runtestlist.py) | Running directory: account ['python', 'runtest.py', '-t', 'account', '--binary', '.././Shredder.app', '--symbols-path', '../symbols'] WARNING | endRunner was never called. There must have been a failure in the framework. NEXT ERROR TEST-UNEXPECTED-FAIL | (runtestlist.py) | Exited with code 1 during directory run or: ['python', 'runtest.py', '-t', 'account', '--binary', '.././Shredder.app', '--symbols-path', '../symbols'] error: uncaptured python exception, closing channel <jsbridge.network.Bridge 127.0.0.1:24242 at 0x1011cea28> (<class 'socket.error'>:[Errno 9] Bad file descriptor [/tools/python/lib/python2.6/asyncore.py|readwrite|107] [/tools/python/lib/python2.6/asyncore.py|handle_expt_event|441] [<string>|getsockopt|1] [/tools/python/lib/python2.6/socket.py|_dummy|165]) error: uncaptured python exception, closing channel <jsbridge.network.BackChannel 127.0.0.1:24242 at 0x1011cef80> (<class 'socket.error'>:[Errno 9] Bad file descriptor [/tools/python/lib/python2.6/asyncore.py|readwrite|107] [/tools/python/lib/python2.6/asyncore.py|handle_expt_event|441] [<string>|getsockopt|1] [/tools/python/lib/python2.6/socket.py|_dummy|165]) WARNING | endRunner was never called. There must have been a failure in the framework. NEXT ERROR TEST-UNEXPECTED-FAIL | (runtestlist.py) | Exited with code 1 during directory run INFO | (runtestlist.py) | account: 0 passed, 0 failed For reference, these versions do work: mozrunner: 2.5.2 jsbridge: 2.3.7b1 mozmill: 1.4.2b1 So I'm suspecting there's something in the jsbridge/mozmill code that's upsetting us, I need to work out what.

Could you also test with 1.4.2b2 and later beta releases for all packages? Would be good to nail down possible checkins.

This problem still exists with 1.5.2 RC1.

Does anyone know if this failure happens in mozmill trunk (mozmill 2.0)? There has been a lot of refactoring around the mozrunner and stopping code.

Could anyone point to a singular thunderbird test to try? they are no longer in the hg.m.o/qa/mozmill-tests

(In reply to comment #4) > Could anyone point to a singular thunderbird test to try? they are no longer > in the hg.m.o/qa/mozmill-tests I did some work last week to ensure that 1.5.2 ran fine on Thunderbird, so that's solved. However, I'm not clear if 2.0 does. Leaving this bug open until we hear back from that investigation. The tbird tests are in the comm central tree. https://developer.mozilla.org/en/Thunderbird/Thunderbird_MozMill_Testing

see also https://bugzilla.mozilla.org/show_bug.cgi?id=656736 Unsure how these are related, exactly

We're green with MozMill 1.5.4 now when the fix in bug 656736 is included. Duping.