Closed Bug 857380 Opened 11 years ago Closed 11 years ago

21b1 update tests failing with NS_ERROR_UNEXPECTED in nsIFileOutputStream.init

Categories

(Mozilla QA Graveyard :: Mozmill Tests, defect, P1)

defect

Tracking

(Not tracked)

VERIFIED FIXED

People

(Reporter: u279076, Assigned: mario.garbi)

References

Details

(Keywords: regression, Whiteboard: s=130401 u=failure c=update p=1)

I was testing 21.0b1 update tests on betatest this evening and all the Mac testruns are failing with the following exception in /testFallbackUpdate/test2.js::teardownModule:

Component returned failure code: 0x8000ffff (NS_ERROR_UNEXPECTED) [nsIFileOutputStream.init]

I am not seeing this error on Linux or Windows testruns. Unfortunately, I don't have access to a Mac OSX computer to check if this is a problem with the updates or if it's a failure of the automation infrastructure.
A report for such a failing case can be found here:

http://mozmill-ondemand.blargon7.com/#/update/report/25ad365ca7bcf4905e9b700b4fecaa1d

This has most likely regressed by the changes in bug 842571. Mario, please investigate the issue as soon as possible. We might want to reopen bug 842571 and backout the changes for a full fix.
Blocks: 842571
Status: NEW → ASSIGNED
Keywords: regression
Priority: -- → P1
Severity: normal → major
OS: Mac OS X → All
Assignee: nobody → mario.garbi
Whiteboard: s=130401 u=failure c=update p=1
It would seem that UpdRootD no longer works properly with the Beta21 and therefore the patch in bug 842571 causes the failure.

If we back out the patch we stop the failures, at least until we can come with a fix.If a back out isn't possible due to merge conflicts I can make a patch that adds the fallback on again.
Assignee: mario.garbi → nobody
Whiteboard: s=130401 u=failure c=update p=1
Please be careful with mid-air alerts and don't proceed in those cases. Restoring previous values now.
Assignee: nobody → mario.garbi
Whiteboard: s=130401 u=failure c=update p=1
Whatever needs to happen to get update tests working, please make it happen ASAP. I consider this to be a blocker to us signing off 21b1 in time for tomorrow's release.
Fixed with the backout of the patch on bug 842571.

Anthony, if you want you should retrigger all ondemand update tests for beta given that it failed across platforms.
Status: ASSIGNED → RESOLVED
Closed: 11 years ago
Resolution: --- → FIXED
(In reply to Henrik Skupin (:whimboo) from comment #5)
> Fixed with the backout of the patch on bug 842571.
> 
> Anthony, if you want you should retrigger all ondemand update tests for beta
> given that it failed across platforms.

Thanks Henrik. I will try this now and report back soon.
I'm still seeing this failure though far less frequent now:
http://mozmill-ondemand.blargon7.com/#/update/report/740ab23e062758ae8a240b263d0e41f0
(In reply to Anthony Hughes, Mozilla QA (:ashughes) from comment #7)
> I'm still seeing this failure though far less frequent now:
> http://mozmill-ondemand.blargon7.com/#/update/report/
> 740ab23e062758ae8a240b263d0e41f0

Looks like this is resolved now. I had to re-trigger a get_mozmill-tests job and I don't see this failure anymore.
Status: RESOLVED → VERIFIED
Strange. I did that after the landing of all the patches in both CI. Not sure why one of them hasn't fetched the changes.
Whoever triggered the ondemand update tests yesterday for the 20.0.1 release on releasetest was affected by this bug again. Please see bug 842571 comment 51 for further details.

My take from all this mess is that we will no longer trying to fix bug 842571 but will revise it when we no longer have to test those old releases. Most likely post Firefox 24.0.
Severity: major → critical
Product: Mozilla QA → Mozilla QA Graveyard
You need to log in before you can comment on or make changes to this bug.