KeyError in "virgin" runner.

[Stephen A. Goss]

Not sure what caused this error that showed up in mailman.log:

Aug 16 10:54:50 2011 (31400) Uncaught runner exception: u'foo.example.com'
Aug 16 10:54:50 2011 (31400) Traceback (most recent call last):
  File "/home/sgoss/mailman_clone/src/mailman/core/runner.py", line 138, in _one_iteration
    self._process_one_file(msg, msgdata)
  File "/home/sgoss/mailman_clone/src/mailman/core/runner.py", line 220, in _process_one_file
    keepqueued = self._dispose(mlist, msg, msgdata)
  File "/home/sgoss/mailman_clone/src/mailman/runners/virgin.py", line 37, in _dispose
    process(mlist, msg, msgdata, 'virgin')
  File "/home/sgoss/mailman_clone/src/mailman/core/pipelines.py", line 50, in process
    handler.process(mlist, msg, msgdata)
  File "/home/sgoss/mailman_clone/src/mailman/pipeline/cook_headers.py", line 360, in process
    process(mlist, msg, msgdata)
  File "/home/sgoss/mailman_clone/src/mailman/pipeline/cook_headers.py", line 202, in process
    listinfo = mlist.script_url('listinfo')
  File "/home/sgoss/mailman_clone/src/mailman/model/mailinglist.py", line 243, in script_url
    return urljoin(self.domain.base_url, target + '/' + self.fqdn_listname)
  File "/home/sgoss/mailman_clone/src/mailman/model/mailinglist.py", line 227, in domain
    return getUtility(IDomainManager)[self.mail_host]
  File "/home/sgoss/mailman_clone/src/mailman/model/domain.py", line 142, in __getitem__
    raise KeyError(email_host)
KeyError: u'foo.example.com'

[Stephen A. Goss (postfuturist)]

I am able to trigger this error by sending an email to a list from a nonmember, so the message gets put into held status. Then moderating the message with Action.accept. After that, the message does not get delivered, instead the above error appears in the log.

[Stephen A. Goss (postfuturist)]

Regarding last comment, this error does not always appear, but moderator approved message never gets sent, so I think they might be unrelated.

[Barry Warsaw (barry)]

In one sense script_url() is kind of bogus because now that the web ui is separate from the engine, the engine needs a better model for how to calculate its urls. This method is mostly a hold over from Mailman 2.

OTOH, its rather disturbing that you don't have a key in the Domain column of database for foo.example.com. Wait, did you possibly create foo.example.com, and then delete it before this message got put into the virgin queue?

[Stephen A. Goss (postfuturist)]

Ah, yep, it happens right at the end of the test suite (our internal tests, not MM3's) which cleans up everything, including the domain entry.

[Stephen A. Goss (postfuturist)]

I think what happened, was that before the fix for https://bugs.launchpad.net/flufl.lock/+bug/827052 was made, I was looping through and rejecting all pending messages before removing the list and then the domain. Maybe since I was rejecting instead of discarding, it was generating a bounce message for the virgin queue right before deleting the domain entry. This isn't really a problem any more. Sorry if I got some details wrong, I don't fully understand the Mailman pipeline yet.

[Barry Warsaw (barry)]

Yeah, I think this had to be caused by some unsupported external manipulation of the database. The base Runner class will refuse to process any message destined for a mailing list that does not exist, and every mailing list should be associated with an existing domain.

I'm going to mark this one incomplete, but if you find a reproducible test case to show the error, please re-open it.

[Abhilash Raj (raj-abhilash1)]

This bug has been moved to the new gitlab repo here: https://gitlab.com/mailman/mailman/issues/53