1005035 - Increase 'security.dialog_enable_delay' to work around JSBridge disconnect issue

Status: RESOLVED FIXED

(Mozilla QA Graveyard :: Mozmill Tests, defect)

Description

[Andrei Eftimie]

As identified in bug 994658 increasing the delay in 'security.dialog_enable_delay' to 1000 (which is the default value) seems to alleviate the jsbridge disconnect issues we're having.

Comment 1

[Andrei Eftimie]

Attachment: fix1_increase_delay.patch

Comment 2

[Andrei Eftimie]

We should land this on default and mozilla-aurora.

Not sure about other branches.
Applies to all branches except mozilla-esr24.

Comment 3

[Henrik Skupin [:whimboo][⌚️UTC+1]]

Comment on attachment 8416479 [details] [diff] [review]
fix1_increase_delay.patch

Review of attachment 8416479 [details] [diff] [review]:
-----------------------------------------------------------------

When we finally fix this lets make sure we have a single constant only for this. Lets get it in. No need to wait for a review from Andreea.

Comment 4

[Andrei Eftimie]

Comment on attachment 8416479 [details] [diff] [review]
fix1_increase_delay.patch

Review of attachment 8416479 [details] [diff] [review]:
-----------------------------------------------------------------

Landed:
http://hg.mozilla.org/qa/mozmill-tests/rev/0a338a551b5c (default)
http://hg.mozilla.org/qa/mozmill-tests/rev/4e2937c6dcb4 (mozilla-aurora)

Comment 5

[Andrei Eftimie]

Lets see how this fares on Nightly and Aurora over the weekend.

Comment 6

[Andrei Eftimie]

Good news. No JSBridge disconnects over the weekend!

Now to make sure this change fixed the problem, I've backed it out:
http://hg.mozilla.org/qa/mozmill-tests/rev/5178a30216fe (default)

We've also had a fix for the crash land in mc on Friday.
With this backed out of Nightly, we should see jsbridge dc's

Comment 7

[Andrei Eftimie]

Now that we've run Nightly without this patch we've had some failures (2 out of 4 testruns didn't properly complete on OSX).

We've also seen this fail a lot on the beta and on the release branch.

Reenabled on default:
http://hg.mozilla.org/qa/mozmill-tests/rev/0c72cea9aa59 (default)

Transplanted:
http://hg.mozilla.org/qa/mozmill-tests/rev/5370687ecc7d (mozilla-beta)
http://hg.mozilla.org/qa/mozmill-tests/rev/12df70b9abf1 (mozilla-release)

This issue should now be fixed.

When we find exactly who the culprit is and if that will need changes to this pref, it will be done in another bug.

Comment 8

[Henrik Skupin [:whimboo][⌚️UTC+1]]

As you say we only landed a temporary fix here, which ensures that our tests hopefully do not fail anymore during restart of Firefox. So please file a new bug immediately, which handles the investigation and can be used for a final fix. Thanks.

Comment 9

[Henrik Skupin [:whimboo][⌚️UTC+1]]

Oh, we have bug 994658. :)

Comment 10

[Andrei Eftimie]

We had a failure which looks like the one we're trying to fix here on ESR24:
http://mm-ci-master.qa.scl3.mozilla.com:8080/job/mozilla-esr24_functional/2078/console

We might want to backport this to mozilla-esr24 as well.

Comment 11

[Henrik Skupin [:whimboo][⌚️UTC+1]]

Maybe. So lets get it backported. That way we have the same code across branches.

Comment 12

[Andrei Eftimie]

Transplanted to ESR24:
http://hg.mozilla.org/qa/mozmill-tests/rev/e11404d4bbca (mozilla-esr24)