1532054 - Crash in [@ mozilla::dom::PWindowGlobalChild::SendBecomeCurrentWindowGlobal]

Status: RESOLVED FIXED

(Core :: General, defect)

Description

[Calixte Denizet (:calixte)]

This bug is for crash report bp-21843d00-fbba-46e0-a485-d63b30190302.

Top 10 frames of crashing thread:

0 xul.dll mozilla::dom::PWindowGlobalChild::SendBecomeCurrentWindowGlobal ipc/ipdl/PWindowGlobalChild.cpp:98
1 xul.dll nsresult nsGlobalWindowOuter::SetNewDocument dom/base/nsGlobalWindowOuter.cpp:2257
2 xul.dll nsresult nsDocumentViewer::InitInternal layout/base/nsDocumentViewer.cpp:964
3 xul.dll nsresult nsDocumentViewer::Init layout/base/nsDocumentViewer.cpp:712
4 xul.dll nsresult nsDocShell::Embed docshell/base/nsDocShell.cpp:6260
5 xul.dll nsresult nsDocShell::CreateAboutBlankContentViewer docshell/base/nsDocShell.cpp:7053
6 xul.dll nsresult nsWebShellWindow::Initialize xpfe/appshell/nsWebShellWindow.cpp:233
7 xul.dll nsresult nsAppShellService::JustCreateTopWindow xpfe/appshell/nsAppShellService.cpp:673
8 xul.dll nsresult nsAppShellService::CreateHiddenWindowHelper xpfe/appshell/nsAppShellService.cpp:135
9 xul.dll nsresult nsAppShellService::GetHiddenDOMWindow xpfe/appshell/nsAppShellService.cpp:739

There are 280 crashes (from 81 installations) in nightly 67 with buildid 20190302094156. In analyzing the backtrace, the regression may have been introduced by patch [1] to fix bug 827976.

[1] https://hg.mozilla.org/mozilla-central/rev?node=e254496ff95b

Comment 1

[Benjamin Bouvier [:bbouvier] (inactive)]

Instacrash on todoist.com (may require to be logged in) with Nightly 67.0a1 (2019-03-02) (64-bit).

edit: on Windows 10 at least. Examples of local crashes on todoist.com:

bp-05f6559a-dbe9-4496-8a90-b82830190302
bp-837d0e64-8b81-4585-b0c8-c01c40190302
bp-de770ee7-8378-4027-8db5-ecdd10190302

Comment 2

[Thomas Andre]

Crashing on https://displaycal.net/ also

Comment 3

[Marco Zehe (:MarcoZ)]

Crashing on Gmail with report bp-a22afc68-cf40-4a17-be0c-b54190190302 and bp-6ef4b2ce-2369-4b52-9274-2b5e50190302, both on Windows 10.

Comment 4

[Geobert Quach]

https://stackoverflow.com/ as well

Comment 5

[Ted Campbell [:tcampbell]]

We are up to 1000+ crashing installs. https://crash-stats.mozilla.org/topcrashers/?product=Firefox&version=67.0a1&process_type=content&days=1&platform=Windows

Can we pull this nightly release?

Comment 6

[(No longer employed by Mozilla) Aaron Klotz]

This was backed out and new Nightlies have been spun up. See https://bugzilla.mozilla.org/show_bug.cgi?id=827976#c44.

Comment 7

[Julien Cristau [:jcristau]]

Fixed by backout.

Comment 10

[Firefox Bug Husbandry Bot]

Please specify a root cause for this bug. See :tmaity for more information.

Comment 11

[Andrew Overholt [:overholt]]

(In reply to Firefox Bug Husbandry Bot from comment #10)

Please specify a root cause for this bug. See :tmaity for more information.

I'm going to put "Testing error" because we didn't have a test for this but Felipe wrote https://bugzilla.mozilla.org/show_bug.cgi?id=827976#c45 and maybe has time to elaborate on the root cause of the issue here? (Hi, Felipe!)

Comment 12

[:Felipe Gomes (needinfo for replies!)]

oh, Hello Andrew!

I'm changing this to "Internal API Issue" because the issue was that there was code in the child process calling an API only meant for the parent process (in this case, GetHiddenDOMWindow).

The test issue that was mentioned earlier is that it was surprising that this did not get caught by a test, because the bad caller was from a CSS media query that for sure should have been exercised by a test (hint: it wasn't :P)