1880131 - Remove "network.cookie.cookieBehavior" from CDP recommended preferences

Status: RESOLVED FIXED

(Remote Protocol :: WebDriver BiDi, defect, P2)

Description

[Alexandra Borovova [:Sasha]]

While working on the test, I've noticed that network.cookie.cookieBehavior which is set for CDP as 0 gets overridden (or reset) on the CI and get the default value 5 (see failing try push). At the same time, it works fine if the pref is added in COMMON_PREFERENCES (see passing try push).

Comment 1

[Henrik Skupin [:whimboo][⌚️UTC+2]]

My suggestion was to log more often the value of this preference so that it will be easier to see when the pref might get reset. A good place might be when a new BiDi command is send so that we can log it before and right after is has finished. Maybe this could give an idea why it gets reset.

Comment 2

[Alexandra Borovova [:Sasha]]

I've pushed twice to try once with the network test on top of central, and it failed, and once on top of other cookies patches and it passed.
What I've noticed, that since the second push had much more tests, it changed the list of tests which were running in the scope of the chunk. In the failing chunk before running the bidi tests, we had this test running, and it was not included in the passing chunk. I've tried to run locally this test together with get_cookies/partition.py, and it failed the same way as on CI.

Comment 3

[Alexandra Borovova [:Sasha]]

I've just also tried to add the test case:

def test_classic(session):
    print('test')

in the beginning of the get_cookies/partition.py file with network test, run only this file and that was enough to flip the pref

Comment 4

[Henrik Skupin [:whimboo][⌚️UTC+2]]

Yeah, the BiDi session is not used here and as such we restart the browser to have a classic only session. This means that CDP is not enabled as well and as such the preference is not getting set.

But any other test following using the BiDi session again should actually set the preference again given that CDP should be enabled as well.

Comment 5

[Julian Descottes [:jdescottes]]

As mentioned on Matrix, the test should actually pass regardless of the value of the preference and it's probably a sign that our implementation is not flexible enough here. However it could still be nice to have mozilla specific tests which make sure a similar test passes under different values of this preference.

Comment 6

[Alexandra Borovova [:Sasha]]

(In reply to Julian Descottes [:jdescottes] from comment #5)

As mentioned on Matrix, the test should actually pass regardless of the value of the preference and it's probably a sign that our implementation is not flexible enough here. However it could still be nice to have mozilla specific tests which make sure a similar test passes under different values of this preference.

This part is moved to bug 1879503.

To avoid the problem of different values for "network.cookie.cookieBehavior" preference between CDP and WebDriverBiDi, we're going to try removing this preference from CDP.

Comment 7

[Alexandra Borovova [:Sasha]]

Attachment: Bug 1880131 - Remove "network.cookie.cookieBehavior" preference from CDP recommended preferences.

Comment 8

[Alexandra Borovova [:Sasha]]

Attachment: Bug 1880131 - [wdspec] Remove test cases with cross-origin iframes which use document.cookie.

Comment 9

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/7e0fc55064a7
Remove "network.cookie.cookieBehavior" preference from CDP recommended preferences. r=webdriver-reviewers,jdescottes
https://hg.mozilla.org/integration/autoland/rev/fbb087d5e5b8
[wdspec] Remove test cases with cross-origin iframes which use document.cookie. r=webdriver-reviewers,whimboo,jdescottes

Comment 10

[Cristian Tuns]

https://hg.mozilla.org/mozilla-central/rev/7e0fc55064a7
https://hg.mozilla.org/mozilla-central/rev/fbb087d5e5b8