566614 - Crash [@dlopen ] on Thunderbird startup

Status: RESOLVED WORKSFORME

(Core :: Widget: Cocoa, defect)

Description

[Ludovic Hirlimann [:Usul]]

0  	 	@0x8fe0cb71  	
1 		@0x8fe01a97 	
2 		@0x8fe04af7 	
3 		@0x8fe1058a 	
4 		@0x8fe051ad 	
5 		@0x8fe0bc32 	
6 	libSystem.B.dylib 	dlopen 	
7 	thunderbird-bin 	nsAppShell::Init 	widget/src/cocoa/nsAppShell.mm:433
8 	thunderbird-bin 	nsAppShellInit 	widget/src/xpwidgets/nsAppShellSingleton.h:74
9 	libxpcom_core.dylib 	nsGenericModule::Initialize 	nsGenericFactory.cpp:273
10 	libxpcom_core.dylib 	nsGenericModule::GetClassObject 	nsGenericFactory.cpp:361
11 	libxpcom_core.dylib 	nsFactoryEntry::GetFactory 	xpcom/components/nsComponentManager.cpp:3706
12 	libxpcom_core.dylib 	nsComponentManagerImpl::CreateInstance 	xpcom/components/nsComponentManager.cpp:1593
13 	libxpcom_core.dylib 	nsComponentManagerImpl::GetService 	xpcom/components/nsComponentManager.cpp:1901
14 	libxpcom_core.dylib 	nsGetServiceByCID::operator const 	nsComponentManagerUtils.cpp:83
15 	libxpcom_core.dylib 	nsCOMPtr_base::assign_from_gs_cid 	nsCOMPtr.cpp:114
16 	thunderbird-bin 	nsChromeRegistry::CheckForOSAccessibility 	
17 	thunderbird-bin 	ScopedXPCOMStartup::SetWindowCreator 	toolkit/xre/nsAppRunner.cpp:1231
18 	thunderbird-bin 	XRE_main 	toolkit/xre/nsAppRunner.cpp:3277
19 	thunderbird-bin 	main 	mail/app/nsMailApp.cpp:101
20 	thunderbird-bin 	thunderbird-bin@0x2351 	
21 	thunderbird-bin 	thunderbird-bin@0x2278 	
22 		@0x0

Comment 1

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

Steven, this is crashing in dlopen in the code you added in bug 396680.

Comment 2

[Steven Michaud [:smichaud] (Retired)]

> Steven, this is crashing in dlopen in the code you added in bug 396680.

Why do you say this?  It's not apparent from the stack in comment #0.

Please post a stack that shows what you say.

Comment 3

[Steven Michaud [:smichaud] (Retired)]

OK, now I see that the crash happens when code in nsAppShell::Init() tries to dlopen() an Apple system library:

http://hg.mozilla.org/releases/mozilla-1.9.2/diff/ae68009f697c/widget/src/cocoa/nsAppShell.mm

The crash isn't caused by my patch -- the call to dlopen() should just fail if this library is absent.  But the call to dlopen() does seem to trigger it.

Do we know how to reproduce these crashes?

Comment 4

[Steven Michaud [:smichaud] (Retired)]

Also note that this crash is unique in the last 4 weeks.

There are other crashes (in Thunderbird) at dlopen, but only one that happens at http://hg.mozilla.org/releases/mozilla-1.9.2/diff/ae68009f697c/widget/src/cocoa/nsAppShell.mm

http://crash-stats.mozilla.com/query/query?product=Thunderbird&version=ALL%3AALL&platform=mac&range_value=3&range_unit=weeks&date=05%2F18%2F2010+12%3A17%3A17&query_search=signature&query_type=exact&query=dlopen&build_id=&process_type=any&hang_type=any&do_query=1

Comment 5

[Steven Michaud [:smichaud] (Retired)]

Searching on all "products", there are *no* crashes like https://crash-stats.mozilla.com/report/index/c33d0fab-f7e8-4129-86ad-fd08d2100429 in the last week.  Though (once again) there are a fair number (69) at dlopen:

http://crash-stats.mozilla.com/query/query?product=ALL&version=ALL%3AALL&platform=mac&range_value=1&range_unit=weeks&date=05%2F18%2F2010+12%3A26%3A57&query_search=signature&query_type=exact&query=dlopen&build_id=&process_type=any&hang_type=any&do_query=1

Comment 6

[Steven Michaud [:smichaud] (Retired)]

If you're looking for the cause of the crashes at dlopen(), https://crash-stats.mozilla.com/report/index/c33d0fab-f7e8-4129-86ad-fd08d2100429 is clearly not the place to start -- it's not at all typical.

I these are all caused by an Apple bug, which I might be able to work around if we could reproduce even one of the crashes at dlopen.

Comment 7

[Steven Michaud [:smichaud] (Retired)]

> I these are all caused by an Apple bug

I suspect these are all caused by an Apple bug.

Comment 8

[Steven Michaud [:smichaud] (Retired)]

Another possibility, I suppose, is that these crashes happen using dlopen() on a corrupt binary.  Especially if the problem keeps happening on a given machine.

Comment 9

[Ludovic Hirlimann [:Usul]]

(In reply to comment #8)
> Another possibility, I suppose, is that these crashes happen using dlopen() on
> a corrupt binary.  Especially if the problem keeps happening on a given
> machine.

It only happened once on my machine - so I have no STRs nor ways to easily reproduce.

Comment 10

[Wayne Mery (:wsmwk)]

all thunderbird crashes are thunderbird 3.0.x. none newer. 
https://crash-stats.mozilla.com/report/list?product=Thunderbird&query_search=signature&query_type=exact&query=dlopen&reason_type=contains&date=02%2F25%2F2012%2016%3A04%3A06&range_value=16&range_unit=weeks&hang_type=any&process_type=any&do_query=1&admin=1&signature=dlopen

but there are firefox crashes
https://crash-stats.mozilla.com/report/list?product=Firefox&query_search=signature&query_type=exact&query=dlopen&reason_type=contains&date=02%2F25%2F2012%2016%3A11%3A11&range_value=16&range_unit=weeks&hang_type=any&process_type=any&do_query=1&admin=1&signature=dlopen - different stacks though