623436 - "ASSERTION: no SHEntry for a non-transient viewer?"

Status: RESOLVED FIXED

(Core :: DOM: Navigation, defect)

Description

[Jesse Ruderman]

Attachment: testcase (requires popups to be enabled)

###!!! ASSERTION: no SHEntry for a non-transient viewer?: 'IsAboutBlank(mCurrentURI)', file /builds/slave/cen-osx64-dbg/build/docshell/base/nsDocShell.cpp, line 9278

Comment 1

[Jesse Ruderman]

Attachment: stack trace

Comment 2

[Jesse Ruderman]

Still happens on trunk.

###!!! ASSERTION: no SHEntry for a non-transient viewer?: 'NS_IsAboutBlank(mCurrentURI)', file docshell/base/nsDocShell.cpp, line 9411

Comment 3

[Mats Palmgren (inactive)]

Still happens with the test in bug 922976 comment 268.

Comment 4

[Henrik Skupin [:whimboo][⌚️UTC+2]]

So there was an attempt in 2010 to get this fixed by bug 583200. Maybe it didn't fix it completely?

I can see the same assertion when running the attachment 8693720 [details] via marionette in a debug build on Linux64: "mach marionette-test test_file".

Current line of failure:
https://dxr.mozilla.org/mozilla-central/source/docshell/base/nsDocShell.cpp#11521

I tried to get some nsDocShell log output ("MOZLOG=nsDocShell:5") but no output is shown. Maybe its a problem with the debug artifact build?

Comment 5

[Henrik Skupin [:whimboo][⌚️UTC+2]]

Attachment: newer stacktrace

Comment 6

[Henrik Skupin [:whimboo][⌚️UTC+2]]

Just to add this only happens in e10s mode. Let me know if we need another bug for this specific case. I didn't want to hi-jack a years old bug.

Comment 7

[Blake Kaplan (:mrbkap) (inactive)]

(In reply to Henrik Skupin (:whimboo) from comment #6)
> Just to add this only happens in e10s mode. Let me know if we need another
> bug for this specific case. I didn't want to hi-jack a years old bug.

It looks like this bug is unrelated to your assertion, please file a new bug for it (and CC me).

Comment 8

[Blake Kaplan (:mrbkap) (inactive)]

Attachment: Possible fix

In order to figure out if Henrik's assertion was related to this one, I
debugged this one. The problem appears to be that we call
w.location.replace() on a docshell before it's finished loading (and
therefore before there's a session history entry for it). Then, when we
try to navigate it, we find that we failed to ever create a history
entry and assert.

This patch tries to fix this by appending a new history entry if we end
up in this situation.

https://treeherder.mozilla.org/#/jobs?repo=try&revision=616bbe2fc73d

Comment 9

[Henrik Skupin [:whimboo][⌚️UTC+2]]

I filed bug 1301399.

Comment 10

[Olli Pettay [:smaug][[e-mail]]]

Comment on attachment 8789100 [details] [diff] [review]
Possible fix

ok

Comment 11

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/6e5c2daac4d3
Deal a little better with replacing inexistant session history entries. r=smaug

Comment 12

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/integration/mozilla-inbound/rev/6e5c2daac4d3

Comment 13

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/6e5c2daac4d3