1053170 - Test failure "Cookie value is displayed in private mode - '' should equal 'cookieValue'" in testPrivateBrowsing/testFlashCookie.js

Status: RESOLVED WORKSFORME

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

Description

[Cosmin Malutan, [:cosmin-malutan]]

Test:      testCheckFlashCookie    
Failure:   Cookie value is displayed in private mode - '' should equal 'cookieValue'
Branches:  Default
Platforms: All

This reproduces in a complete testrun, we should skip it, I think is different from bug 958068, because here we have no value at all and this is reproducible. 

http://mozmill-daily.blargon7.com/#/remote/failure?app=Firefox&branch=All&platform=All&from=2014-08-06&to=&test=%2FtestPrivateBrowsing%2FtestFlashCookie.js&func=testCheckFlashCookie

Comment 1

[Andrei Eftimie]

Are you sure this is a new issue?
We have this already logged in bug 958068.

If this indeed is reproducible, it might indeed have a different cause.

Comment 2

[Cosmin Malutan, [:cosmin-malutan]]

Attachment: skip_patch

Yes I think this is a different issue, the value here is an empty string and not undefined and as I said it's reproducible the other one was intermittent.

Here is the skip patch.

Comment 3

[Andrei Eftimie]

Comment on attachment 8472312 [details] [diff] [review]
skip_patch

Review of attachment 8472312 [details] [diff] [review]:
-----------------------------------------------------------------

Disabled:
https://hg.mozilla.org/qa/mozmill-tests/rev/883d259a0ef1 (default)

Comment 4

[Cosmin Malutan, [:cosmin-malutan]]

Here is the changesets between the last unaffected and the first affected tinderbox build:
http://hg.mozilla.org/mozilla-central/pushloghtml?fromchangeset=b53c2753ce9a&tochangeset=ee1ad12a3939

To reproduce this you'll have to ran testGeolocation tests prior to this one.

Comment 5

[Andrei Eftimie]

You'll need to check either fx-team or inbound to get the actual regressor.
This pushlog is way too big to find something useful.

Also a testcase would be good.

Comment 6

[Cosmin Malutan, [:cosmin-malutan]]

Again, to reproduce this run geolocation test prior to private browsing.
[include:testGeolocation/manifest.ini]
[include:testPrivateBrowsing/manifest.ini]

It looks like after running geolocation tests, firefox can't display the input elements in private browsing, the exist, you can find them with inspect node but they are not displayed and the don't get any value.
http://mozqa.com/data/firefox/plugins/flash/cookies/flash_cookie.html

Comment 7

[Henrik Skupin [:whimboo][⌚️UTC+2]]

This is a very strange regression! We should really find the causing changeset ASAP. A testcase would really have been good.

Comment 8

[Henrik Skupin [:whimboo][⌚️UTC+2]]

So I tested this and its not what I'm seeing. There is indeed a small delay until the values are shown in the textboxes, but it happens also in older builds. Please try to trim the tests down, so we exactly know which ones have to be executed.

Comment 9

[Cosmin Malutan, [:cosmin-malutan]]

Attachment: test_case.js

This is the test-case, the bug has been fixed, it passes with the latest nightly and we can re-enable it. 
A failing build can be downloaded from here:
https://ftp.mozilla.org/pub/mozilla.org/firefox/nightly/2014-08-13-03-02-01-mozilla-central/

Comment 10

[Cosmin Malutan, [:cosmin-malutan]]

Andrei can you please backout the skip patch?

The changeset that fixed this is between:
http://hg.mozilla.org/mozilla-central/pushloghtml?fromchangeset=d7e78f0c1465&tochangeset=de1d3c229e5a

The first passing build is:
https://ftp.mozilla.org/pub/mozilla.org/firefox/tinderbox-builds/mozilla-central-linux64/1408002307/
The last failing is:
https://ftp.mozilla.org/pub/mozilla.org/firefox/tinderbox-builds/mozilla-central-linux64/1407959453/

Comment 11

[Andrei Eftimie]

This is weird, it looks like the pushlog functionality from http://hg.mozilla.org is broken :|

Comment 12

[Andrei Eftimie]

So the issue indeed appears to have been fixed.

Backed out skip:
https://hg.mozilla.org/qa/mozmill-tests/rev/1613b8c1959f (default)

This still was a regression, and we still want to find the culprit.
Since the web based pushlog appears to be broken, please use hg locally to try tracking down either the commit that broke this test or the one that fixed it, preferably both (but finding one should also find the other)

Comment 13

[Andrei Eftimie]

Actually the pushlog functionality is sound. The query arguments were wrong.


The actual pushlog that BROKE the test is:
http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=b53c2753ce9a&tochange=ee1ad12a3939

And this one FIXED the issue:
http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=d7e78f0c1465&tochange=8f922378fab0

*Sigh* there is no common bug no between them :(

I'm not sure this is worth investing more time into.
I will mark both bugs as WFM