Closed Bug 1314627 Opened 8 years ago Closed 7 years ago

Intermittent IOError: Process has been closed (Exit code: 0) (Reason: [Errno 10054] An existing connection was forcibly closed by the remote host)

Categories

(Testing :: Firefox UI Tests, defect)

Product:
Testing
Component:
Firefox UI Tests
Version:
Version 3
Type:
defect
Priority:
Not set
Severity:
normal

Tracking

(firefox50 fixed, firefox51 fixed, firefox52 wontfix)

Status:
RESOLVED FIXED
Milestone:
mozilla51
Tracking Flags:
Tracking Status
firefox50 --- fixed
firefox51 --- fixed
firefox52 --- wontfix

People

(Reporter: intermittent-bug-filer, Assigned: whimboo)

References

Details

(Keywords: intermittent-failure)

Attachments

(1 file)

last-update.log
26.77 KB, text/plain
Details 
Filed by: andrei.vaida [at] softvision.ro

https://treeherder.mozilla.org/logviewer.html#?job_id=328587&repo=mozilla-release

https://firefox-ui-tests.s3.amazonaws.com/94ebe73f-76c3-4a93-b135-1d44f0b90814/log_info.log

Seen while running ondemand update tests for 50.0-build1.
When observing a test run on one of our machines I can see that Firefox downloads the update and applies it. But after the restart nothing happens. In some cases I see bug 1256425 happening before but that doesn't cause a shutdown of Firefox either.

So I have the feeling that due to some reason the connection is broken between the marionette client and server. Not sure yet how to best investigate that, given that update tests are tricky. Especially because our usual functional tests for 50.0rc seem to work. So there is something happening with the restart or some brokenness in the updated Firefox.
status-firefox50: --- → affected
status-firefox51: --- → ?
status-firefox52: --- → ?
Attached file last-update.log — 
The last-update.log file from such a failing update test. As it can be seen the update has been applied correctly. Just not sure why we add and remove the same files a couple of times. Robert, could you shed some light on this? Thanks.

Otherwise I still don't know what could cause this problem. Running the functional tests with an updated build afterwards just works fine.
Flags: needinfo?(robert.strong.bugs)
(In reply to Henrik Skupin (:whimboo) from comment #2)
> Created attachment 8806850 [details]
> last-update.log
> 
> The last-update.log file from such a failing update test. As it can be seen
> the update has been applied correctly. Just not sure why we add and remove
> the same files a couple of times. Robert, could you shed some light on this?
> Thanks.
That is how complete updates are logged.
Flags: needinfo?(robert.strong.bugs)
(In reply to Robert Strong [:rstrong] (use needinfo to contact me) from comment #3)
> That is how complete updates are logged.

Ok, so that doesn't seem to be related then.

The last thing I tried yesterday was to not force an in-app restart of Firefox to get the update applied, but doing an external force-shutdown (kill) of the process. Starting Firefox then works. So I assume there is something busted with the respawned process. The strange thing still is why it starts to happen now. We never have seen such an outage all the time on mozilla-central and mozilla-aurora. Even not for the beta branch until now when we run based off the release branch.
Ok, so as it looks like this bug is indeed related to bug 1256425. The problem seems to be that we open a graphics test page after the update of the browser, and close it some milliseconds later. This causes a race and we somehow fail to continue with Marionette.

Means we need a fix for bug 1256425. I will leave this bug open until then, so we can use it for starring on Treeherder.
Depends on: 1256425
Bug 1256425 has been fixed via the patch on bug 1314891. Sadly it has not been backported to other branches yet, so we could still be affected on release.
Status: NEW → RESOLVED
Closed: 8 years ago
status-firefox51: ? → affected
status-firefox52: ? → fixed
Resolution: --- → FIXED
Target Milestone: --- → mozilla52
Assignee: nobody → hskupin
status-firefox50: affected → fixed
status-firefox51: affected → fixed
It looks like that there is still another issue here given the update test results for 50.0rc on the beta channel yesterday:

https://treeherder.mozilla.org/#/jobs?repo=mozilla-release&revision=dc617d65c9f0&group_state=expanded&filter-tier=3&filter-searchStr=Fxup-beta(&selectedJob=353827
Status: RESOLVED → REOPENED
status-firefox50: fixed → affected
status-firefox51: fixed → affected
status-firefox52: fixed → affected
Resolution: FIXED → ---
Target Milestone: mozilla52 → ---
Too late for firefox 52, mass-wontfix.
status-firefox52: affected → wontfix
Status: REOPENED → RESOLVED
Closed: 8 years ago → 7 years ago
Keywords: bulk-close-intermittents
Resolution: --- → INCOMPLETE
Looks like I accidentally removed some flags in comment 10. This bug is actually fixed.
status-firefox50: affected → fixed
status-firefox51: affected → fixed
Keywords: bulk-close-intermittents
Resolution: INCOMPLETE → FIXED
Target Milestone: --- → mozilla51
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Creator:
Created:
Updated:
Size: