With our Firefox UI tests we see a lot of the following type of failure especially in our update tests. [Errno 10054] An existing connection was forcibly closed by the remote host Traceback

Traceback (most recent call last): File "c:\jenkins\workspace\mozilla-central_update\venv\lib\site-packages\marionette\runner\base.py", line 561, in gather_debug with marionette.using_context(marionette.CONTEXT_CHROME): File "C:\Python27\Lib\contextlib.py", line 17, in __enter__ return self.gen.next() File "c:\jenkins\workspace\mozilla-central_update\venv\lib\site-packages\marionette_driver\marionette.py", line 1159, in using_context scope = self._send_message('getContext', 'value') File "c:\jenkins\workspace\mozilla-central_update\venv\lib\site-packages\marionette_driver\decorators.py", line 36, in _ return func(*args, **kwargs) File "c:\jenkins\workspace\mozilla-central_update\venv\lib\site-packages\marionette_driver\marionette.py", line 687, in _send_message response = self.client.send(message) File "c:\jenkins\workspace\mozilla-central_update\venv\lib\site-packages\marionette_transport\transport.py", line 101, in send raise e Affected line of code: https://dxr.mozilla.org/mozilla-central/source/testing/marionette/client/marionette/runner/base.py#562 I don't see a pattern yet when it happens. It seems to be random. Maybe I'm able to get a reduced testcase for it and have it reproducible all the time.

I ran this test a dozen of times but the given problem did not re-appear. So not exactly sure what triggers this disconnect. My feeling is that it happens due to one of the requested restarts of Firefox, and that in such a case the Marionette client is still trying to send a message while the server in Firefox already closed the connection. I will keep the machine mm-win-8-64-4 offline and do a long-time test on it. Maybe it will reproduce in at least one of thousand runs.

More testing has shown that this seems to only be an issue with the mm-win-8-64-4 VM in SCL3. All other VMs seem to work regarding this reported problem. But given that other machines of this Windows version seem to have focus issues, I requested a re-deployment of all VMs from the template. It should fix this particular issue, which I was not able to reproduce at all. I will mark the bug as incomplete for now and will reopen if I'm able to reproduce once it happens again.