2002313 - Perma dom/media/tests/crashtests/1429507_1.html | load failed: timed out waiting for reftest-wait to be removed when Gecko 147 merges to beta on 2025-12-08

Status: VERIFIED FIXED

(Core :: DOM: Navigation, defect, P1)

Description

[agoloman]

Central-as-beta-simulation
 
How to run these simulations
 
Failure log

Comment 1

[Vincent Hilla [:vhilla]]

Can reproduce with android, central-as-beta, nofis. The window.open("/") call fails.

Here's the log

1-27 14:22:27.435  6747  6788 I GeckoDump: TEST START
11-27 14:22:27.441  6691  6713 I Gecko   : [Parent 6691, Main Thread] WARNING: NS_ENSURE_TRUE(domWin) failed: file /home/vincent/mozilla-unified/widget/android/GeckoProcessManager.h:41
11-27 14:22:27.446  6747  6788 W GeckoEditableChild: No editable parent
11-27 14:22:27.451  6691  6713 I Gecko   : [Parent 6691, Main Thread] WARNING: NS_ENSURE_TRUE(domWin) failed: file /home/vincent/mozilla-unified/widget/android/GeckoProcessManager.h:41
11-27 14:22:27.451  6691  6713 I Gecko   : [Parent 6691, Main Thread] WARNING: '!browserDOMWin', file /home/vincent/mozilla-unified/dom/ipc/ContentParent.cpp:5256
11-27 14:22:27.451  6691  6713 I Gecko   : [Parent 6691, Main Thread] WARNING: 'NS_FAILED(rv)', file /home/vincent/mozilla-unified/dom/ipc/ContentParent.cpp:5490
11-27 14:22:27.452  6747  6788 I Gecko   : [Child 6747, Main Thread] WARNING: 'aOwner->IsDiscarded()', file /home/vincent/mozilla-unified/obj-debug-android/dist/include/mozilla/dom/SyncedContextInlines.h:95
11-27 14:22:27.453  6747  6788 E Web Content: [JavaScript Error: "TypeError: can't access property "foo", o1 is null" {file: "http://192.168.184.28:8888/tests/dom/media/tests/crashtests/1429507_1.html" line: 9}]

Comment 2

[Vincent Hilla [:vhilla]]

On android, central-as-beta, with fission, the !browserDOMWin is passing.

On Linux, !browserDOMWin seems to fail always. I checked central-as-beta with and without fission, as well as central before bug 543435.

Comment 3

[Vincent Hilla [:vhilla]]

A webpage in GeckoView can do window.open('/'). Perhaps it's an issue with reftests.

Comment 4

[Vincent Hilla [:vhilla]]

Hm, it seems android nofis ends up using a different, earlier created BrowserParent.

In both cases, the logs show

• nsFrameLoader::TryRemoteBrowserInternal passing rootWin->GetBrowserDOMWindow() to topParent, which is null right now.
• Then GeckoViewNavigation onInitBrowser sets browserDOMWindow on the inner window, which propagates to some root window.
• Another TryRemoteBrowserInternal occurs and sets browserDOMWindow on a different topParent.
• The test runs and does window.open().

With fission, we use the second BrowserParent and with nofis, the first one.

Comment 5

[Vincent Hilla [:vhilla]]

Oh, I added the window reuse check in D265268. Everything else is wrapped in a try-catch, so perhaps the window.open call already failed before bug 543435.

Comment 6

[Vincent Hilla [:vhilla]]

I verified it with the parent rev of bug 543435 (https://hg-edge.mozilla.org/integration/autoland/rev/37f1c536fdb89cd69039774a87f3ad8a94b95328).

Already before the patch, with android nofis, central-as-beta, o1 was null. This just wasn't noticeable.

I'll remove the window reuse check and update the comment. With the window reuse change in bug 543435, the only necessary change to this test would've been the added / to have a non-sync load.

Comment 7

[Vincent Hilla [:vhilla]]

Attachment: Bug 2002313 - Continue to fail tests for bug 1429507 silently if window.open fails. r=#dom-core

Bug 543435 changed when window reuse occured and made the initial about:blank
load sync.
This required changing the loaded URL of the tests for bug 1429507 to non-blank
URLs. I added code to verify the window is reused, but aparently window.open
can fail in reftests in certain configurations, already before bug 543435.

Comment 8

[Vincent Hilla [:vhilla]]

If someone thinks it's needed, I'm happy to file a follow-up bug for this. But I don't think it's needed. My best guess is there is some race in GeckoView reftest infrastructure that leads to there being no nsIBrowserDOMWindow associated with the test window. But if it only occurs with android nofis, and we'll have fission soon, it's probably not worth spending effort on.

Comment 9

[Cristina Horotan [:chorotan]]

This is failing on early beta too, as seen here

Comment 10

[Intermittent Failures Robot]

7 failures were associated with this bug in the last 7 days.

Repository breakdown:

• try: 7

Table

	nofis
android-x86/opt	7

For more details, see:

https://treeherder.mozilla.org/intermittent-failures/bugdetails?bug=2002313&startday=2025-11-24&endday=2025-11-30&tree=all

Comment 11

[Pulsebot]

Pushed by [e-mail]:
https://github.com/mozilla-firefox/firefox/commit/5980df272ec0
https://hg.mozilla.org/integration/autoland/rev/25fa90f4cc11
Continue to fail tests for bug 1429507 silently if window.open fails. r=webrtc-reviewers,dom-core,smaug,pehrsons

Comment 12

[agoloman]

https://hg.mozilla.org/mozilla-central/rev/25fa90f4cc11

Comment 13

[amarc]

Verified fixed in today's simulation