Right now we fail in our mozrunner thread tests for stopping the application because the returncode is 0 instead of -9. This happens because in kill() we declare the returncode as 0, and if process is already dead because another thread stopped it, it overwrites the former -sig returncode. https://github.com/mozilla/mozbase/blob/master/mozprocess/mozprocess/processhandler.py#L114 If the returncode has been set, we should really take that one.

Actually this happens when you call wait() twice after the process has been stopped via kill(). I will have a fix for this today.

Simple patch with an additional test for this problem. It needs to be tested on OS X first before I want to ask for review.

Comment on attachment 8371460 [details] [diff] [review] Patch v1 Review of attachment 8371460 [details] [diff] [review]: ----------------------------------------------------------------- Tests are all green on OS X and Linux. Sadly tests are still all skipped on Windows.

Comment on attachment 8371460 [details] [diff] [review] Patch v1 Review of attachment 8371460 [details] [diff] [review]: ----------------------------------------------------------------- Looks like William and Andrew are out the whole week. Jonathan, would you mind to review? Also adding Joel in case he could do it.

Comment on attachment 8371460 [details] [diff] [review] Patch v1 Review of attachment 8371460 [details] [diff] [review]: ----------------------------------------------------------------- ::: mozprocess/tests/test_mozprocess_wait.py @@ +104,5 @@ > + "process_waittimeout_python.ini"], > + cwd=here) > + p.run() > + p.kill() > + p.wait() should you verify the returncode from the first wait?

Comment on attachment 8371460 [details] [diff] [review] Patch v1 Review of attachment 8371460 [details] [diff] [review]: ----------------------------------------------------------------- Ditto jmaher's comments.

Ups, you both are right. I missed to add this. Fixed it now. I will land it tomorrow with fresh eyes. Thanks!