Comment 2 for bug 237671

On Wed, 2008-06-11 at 15:03 +0000, Barry Warsaw wrote:
> Really, PQM should send a notification regardless of the exit status.
> Two other examples have been seen recently.
>
> 1. When the test runner runs for more than 15 minutes, it will get
> killed, but the exit status indicates signal death, and these exits
> don't trigger a notification.

I thought the killing was done by test.py or some wrapper *nothing*
should be killing PQM itself to stop a test run - thats nuts because it
will prevent cleanups happening. IF this is the case its a bug in the
deployment.

> 2. If the bzr merge command fails (e.g. because the target was a loom),
> this also doesn't trigger a notification.
>
> pqm should /always/ send a notification!

If the merge fails a conflict list is sent to the user. I presume you
mean the source being a loom causes no notification - I would guess that
this *actually* causes a traceback - and for some reason I am not
getting a copy of the traceback?

(PQM /always/ sends a notification except when PQM suffers an uncaught
exception, uncaught exceptions go to the output which cron captures and
sends to the operators. This lets them report bugs and fix the issue).

I haven't seen seeing any tracebacks for the launchpad instance in a
long time, so either the configuration is borked, the admins are not
reporting them anymore or ....

Anyhow, this is the code for handling test suite exit values:

   ecode = child.wait()
   if not ((ecode is None) or (ecode == 0)):
       self.output.append('\npre-commit hook failed with error code %d
at %s\n' % (ecode - 255, time.strftime('%c')))
       raise PQMCmdFailure(self.sender, self.successful, self.line,
self.output)

-Rob

--
GPG key available at: <http://www.robertcollins.net/keys.txt>.