Closed Bug 1369413 Opened 4 years ago Closed 4 years ago

Intermittent test_switch_remote_frame.py TestSwitchRemoteFrame.test_we_can_switch_to_a_remote_frame_by_index | IOError: Process killed because the connection to Marionette server is lost. Check gecko.log for errors

Categories

(Testing :: Marionette, defect)

Product:
Testing
Component:
Marionette
Version:
Version 3
Type:
defect
Priority:
Not set
Severity:
normal

Tracking

(firefox-esr52 unaffected, firefox53 unaffected, firefox54 unaffected, firefox55 fixed)

Status:
RESOLVED FIXED
Milestone:
mozilla55
Tracking Flags:
Tracking Status
firefox-esr52 --- unaffected
firefox53 --- unaffected
firefox54 --- unaffected
firefox55 --- fixed

People

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

References

Details

(Keywords: intermittent-failure, regression, Whiteboard: [stockwell fixed:other])

Attachments

(1 file)

patch, v1.
1.63 KB, patch
ato
: review+
Details | Diff | Splinter Review 
The problem here is:

1496328219064	Marionette	TRACE	81 -> [0,49,"switchToFrame",{"focus":true,"id":0}]
JavaScript error: chrome://marionette/content/listener.js, line 1: SyntaxError: redeclaration of const INPUT_DATETIME_PREF

Seems to have been caused by https://hg.mozilla.org/integration/autoland/rev/f4448d4bcf98. So a regression from bug 1366188.

Jessica, can you please have a look?

PS: Your author name for the patch doesn't contain a valid email address.
Blocks: 1366188
Flags: needinfo?(jjong)
Keywords: regression
Thanks for reporting this, I could not reproduce it locally.
But to avoid polluting the global scope, I will put the const inside my function or just use literal string in Preferences.get().
Flags: needinfo?(jjong)
When you check Orange Factor the issue only hits some platform and then specific builds. For example on Linux only ASAN builds seem to be affected.
Attached patch patch, v1. — — Splinter Review 
Hi ato, I think the root cause might be loading the frame script multiple times in switchToFrame(), but that is just a guess and I don't really know the code, so I'll just avoid the issue by removing the const.
Attachment #8873778 - Flags: review?(ato)
Attachment #8873778 - Flags: review?(ato) → review+
(In reply to Jessica Jong [:jessica] from comment #5)
> Hi ato, I think the root cause might be loading the frame script multiple
> times in switchToFrame(), but that is just a guess and I don't really know
> the code, so I'll just avoid the issue by removing the const.

Yes, that is pretty worrying.  We should look into that as part of
https://bugzilla.mozilla.org/show_bug.cgi?id=1157192.
(In reply to Andreas Tolfsen ‹:ato› from comment #6)
> (In reply to Jessica Jong [:jessica] from comment #5)
> > Hi ato, I think the root cause might be loading the frame script multiple
> > times in switchToFrame(), but that is just a guess and I don't really know
> > the code, so I'll just avoid the issue by removing the const.
> 
> Yes, that is pretty worrying.  We should look into that as part of
> https://bugzilla.mozilla.org/show_bug.cgi?id=1157192.

I see, thanks for the info.
Pushed by ryanvm@gmail.com:
https://hg.mozilla.org/integration/mozilla-inbound/rev/dbad3f1bb947
Remove the use of const INPUT_DATETIME_PREF. r=ato
Keywords: checkin-needed
https://hg.mozilla.org/mozilla-central/rev/dbad3f1bb947
Status: NEW → RESOLVED
Closed: 4 years ago
status-firefox55: affected → fixed
Resolution: --- → FIXED
Target Milestone: --- → mozilla55
Whiteboard: [stockwell fixed:other]
You need to log in before you can comment on or make changes to this bug.