1866884 - Leak of window created during docshell/test/browser/browser_alternate_fixup_middle_click_link.js

Status: RESOLVED FIXED

(Core :: DOM: Navigation, defect)

Description

[Andrew McCreight [:mccr8]]

This leak has shown up twice in bug 1866073, and I think I've seen leaks before while running this docshell directory so maybe it is the same thing. I'm not sure what the cause is.

Comment 1

[Andrew McCreight [:mccr8]]

It looks like this test intentionally loads an about:neterror page, so maybe this is another instance of the shutdown issue Emilio worked around for a few tests in bug 1865771.

Comment 2

[Andrew McCreight [:mccr8]]

Emilio, any ideas how we could stop this testing leak? Like maybe the test could explicitly wait until Stylo has finished dealing with about:neterror somehow? It seems to be one of the more frequent leaks. Unfortunately unlike bug 1865771 this test is intentionally loading a page that causes about:neterror so we can't fix it to a valid URL. Thanks.

Comment 3

[Emilio Cobos Álvarez (:emilio)]

Not sure if we can wait for the load event of about:neterror somehow. Nika, do you happen to know?

Comment 4

[:Gijs (out for now; he/him)]

(In reply to Emilio Cobos Álvarez (:emilio) from comment #3)

Not sure if we can wait for the load event of about:neterror somehow. Nika, do you happen to know?

There's a waitForErrorPage helper on BTU right? https://searchfox.org/mozilla-central/rev/f465027ef4d334dbc9ad270718c8a5e8045e7a2b/testing/mochitest/BrowserTestUtils/BrowserTestUtils.sys.mjs#1657

I think this is already being used in this test, cf. https://searchfox.org/mozilla-central/rev/f465027ef4d334dbc9ad270718c8a5e8045e7a2b/docshell/test/browser/browser_alternate_fixup_middle_click_link.js#26-39 - but not all the time. And I expect that when we hit the condition checked there (ie the currentURI has already updated), it may not have finished loading, and although that's sufficient for the test it's not sufficient to stop it leaking.

Comment 5

[Emilio Cobos Álvarez (:emilio)]

Attachment: Bug 1866884 - Convert aboutNetError to an html page. r=Gijs

While I'm looking at this / bug 1866884, this it probably makes sense,
wdyt?

Comment 6

[Phabricator Automation]

Comment on attachment 9372847 [details]
Bug 1866884 - Convert aboutNetError to an html page. r=Gijs

Revision D198571 was moved to bug 1874701. Setting attachment 9372847 [details] to obsolete.

Comment 7

[Emilio Cobos Álvarez (:emilio)]

Attachment: Bug 1866884 - Make browser_alternate_fixup_middle_click_link.js wait for error page properly. r=Gijs

Depends on D198571

Comment 8

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/30195878230c
Make browser_alternate_fixup_middle_click_link.js wait for error page properly. r=Gijs

Comment 9

[Sandor Molnar[:smolnar]]

https://hg.mozilla.org/mozilla-central/rev/30195878230c

Comment 10

[Andrew McCreight [:mccr8]]

Thanks Emilio and Gijs for your work here! Hopefully this will help reduce the failure rate of these window leaks.