701352 - Firefox 8 crash on quit after opening in 32 bit mode

Status: RESOLVED DUPLICATE of bug 700835

(Firefox :: General, defect)

Description

[Vishvesh]

User Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_6_8) AppleWebKit/535.2 (KHTML, like Gecko) Chrome/15.0.874.106 Safari/535.2

Steps to reproduce:

1. Set Firefox to launch in "Open in 32-bit mode"
2. Open a page which loads my plug-in
3. Quit the page.



Actual results:

Firefox 8 crashed. 




Expected results:

Firefox 8 should have terminated

Comment 1

[Vishvesh]

The crash reports
https://crash-stats.mozilla.com/report/index/bp-d3f30b91-4914-477e-835d-2bf0b2111110
https://crash-stats.mozilla.com/report/index/bp-6354e0ff-a6a9-46cc-9d6d-6ec2c2111110
https://crash-stats.mozilla.com/report/index/bp-d3f30b91-4914-477e-835d-2bf0b2111110

Comment 2

[Vishvesh]

This issues occurs in both 32-bit mode and 64-bit mode. But it is very consistent in 32-bit mode. In 64-bit mode, If I quit the browser few minutes after closing the tab that had loaded the plug-in I see a crash.

Crash is seen only on 10.7.2 machines with Firefox 8. Tried it on a Mac book pro and a Mac mini machine. No crashes are seen on 10.7.

Comment 3

[Vishvesh]

Attachment: Apple crash report for the crash 32-bit mode

Apple crash report for the crash 32-bit mode

Comment 4

[Vishvesh]

Attachment: Apple crash report for the crash 64-bit mode

Comment 5

[Thomas Ahlblom]

0 	libsystem_c.dylib 	__cxa_finalize 	
1 		@0x8fe65c6b 	
2 		@0x8fe6ead2 	
3 	libdyld.dylib 	NSUnLinkModule 	
4 	libnspr4.dylib 	PR_UnloadLibrary 	nsprpub/pr/src/linking/prlink.c:1023
5 	libnss3.dylib 	SECMOD_UnloadModule 	pk11load.c:622
6 	libnss3.dylib 	SECMOD_SlotDestroyModule 	pk11util.c:884
7 	libnss3.dylib 	PK11_DestroySlot 	pk11slot.c:451
8 	libnss3.dylib 	SECMOD_DestroyModule 	pk11util.c:855
9 	XUL 	nsNSSComponent::UnloadLoadableRoots 	security/manager/ssl/src/nsNSSComponent.cpp:936
10 	XUL 	nsNSSComponent::ShutdownNSS 	security/manager/ssl/src/nsNSSComponent.cpp:1925
11 	XUL 	nsNSSComponent::DoProfileBeforeChange 	security/manager/ssl/src/nsNSSComponent.cpp:2609
12 	XUL 	nsNSSComponent::Observe 	security/manager/ssl/src/nsNSSComponent.cpp:2243
13 	XUL 	nsObserverList::NotifyObservers 	xpcom/ds/nsObserverList.cpp:130
14 	XUL 	nsObserverService::NotifyObservers 	xpcom/ds/nsObserverService.cpp:182
15 	XUL 	nsXREDirProvider::DoShutdown 	toolkit/xre/nsXREDirProvider.cpp:802
16 	XUL 	ScopedXPCOMStartup::~ScopedXPCOMStartup 	toolkit/xre/nsAppRunner.cpp:1079
17 	XUL 	XRE_main 	toolkit/xre/nsAppRunner.cpp:1086
18 	firefox 	main 	browser/app/nsBrowserApp.cpp:198
19 	firefox 	firefox@0xbe5 	
20 		@0x1

Comment 6

[Steven Michaud [:smichaud] (Retired)]

Do you have a public URL (one you can post here) with which you can reproduce this bug?

Or failing that can you attach a zipped-up testcase to this bug, containing your plugin and an *.html file that loads it, so that we can put it on our servers to test with?

Without either of these, there's not much we can do.

However there's a pretty good chance that a fix for bug 700835 will also fix your bug.  So you can wait until we post a patch for that bug (with a tryserver build) and then try the tryserver build.

By the way, if your bug *is* a dup of bug 700835, it will no longer happen in today's mozilla-central nightly.

Comment 7

[Steven Michaud [:smichaud] (Retired)]

> By the way, if your bug *is* a dup of bug 700835, it will no longer happen in
> today's mozilla-central nightly.

This may not be correct on OS X 10.6.8.

Comment 8

[Steven Michaud [:smichaud] (Retired)]

https://crash-stats.mozilla.com/ has only three of these crashes in the last week:

https://crash-stats.mozilla.com/query/query?product=Firefox&version=ALL%3AALL&range_value=1&range_unit=weeks&date=11%2F10%2F2011+16%3A26%3A24&query_search=signature&query_type=contains&query=__cxa_finalize&reason=&build_id=&process_type=any&hang_type=any&do_query=1

Comment 9

[Vishvesh]

Please visit :
http://qawatch.slingbox.com/
User Name : [e-mail]
Password :  vishvesh

You would be prompted to download a plug-in. Download it.

There is one box in this account (FF8 Test), connect to the box and stream for few minutes and quit the browser. This crash occurs in 32 bit mode..

Comment 10

[Vishvesh]

(In reply to Steven Michaud from comment #8)
> https://crash-stats.mozilla.com/ has only three of these crashes in the last
> week:
> 
> https://crash-stats.mozilla.com/query/
> query?product=Firefox&version=ALL%3AALL&range_value=1&range_unit=weeks&date=1
> 1%2F10%2F2011+16%3A26%3A24&query_search=signature&query_type=contains&query=_
> _cxa_finalize&reason=&build_id=&process_type=any&hang_type=any&do_query=1

Initial update on the issue. I started testing the issue with the following build

http://ftp.mozilla.org/pub/mozilla.org/firefox/try-builds/[e-mail]-f3b5cce53f90/try-macosx64/firefox-11.0a1.en-US.mac.dmg

I haven't seen this issue yet. I will provide an final update by the end of the day.

Comment 11

[Vishvesh]

(In reply to Steven Michaud from comment #8)
> https://crash-stats.mozilla.com/ has only three of these crashes in the last
> week:
> 
> https://crash-stats.mozilla.com/query/
> query?product=Firefox&version=ALL%3AALL&range_value=1&range_unit=weeks&date=1
> 1%2F10%2F2011+16%3A26%3A24&query_search=signature&query_type=contains&query=_
> _cxa_finalize&reason=&build_id=&process_type=any&hang_type=any&do_query=1

The fix for https://bugzilla.mozilla.org/show_bug.cgi?id=700835 fixes issue with our plug-in. I still don't understand why this would effect a non-java plug-in. 

I am marking the bug as resolved. Thanks.

Comment 12

[Steven Michaud [:smichaud] (Retired)]

> I am marking the bug as resolved. Thanks.

You're quite welcome.

> The fix for https://bugzilla.mozilla.org/show_bug.cgi?id=700835
> fixes issue with our plug-in. I still don't understand why this
> would effect a non-java plug-in.

As best I can tell, the only way my patch could fix your plugin is if
it calls NPN_SetValue to turn on NPPVpluginKeepLibraryInMemory, and if
it has the same kind of bug as Apple's newest Java updates (if it
tries to call a method on (or otherwise access) a plugin instance
after that plugin instance has been destroyed).

Are you the plugin's author?  If not you can ask the developers who
wrote or currently work on the plugin.

Comment 13

[Vishvesh]

I am co-author of the plug-in. I spent the entire day investigating the issue. It looks like the issue is on the plug-in side and not on the browser. But for some reason it is seen only on FF8 running on 10.7.2.

The plug-in doesn't use "NPPVpluginKeepLibraryInMemory".