380228 - Assertion failure at shutdown if console listener isn't unregistered: PR_ASSERT(lock->owner == 0);

Status: RESOLVED INCOMPLETE

(Core :: XPCOM, defect)

Description

[Alex Vincent [:WeirdAl]]

Attachment: stack trace

I was writing a xpcshell testcase for bug 380169, when I stumbled upon this crash.  Apparently, if you fail to unregister a console listener for any reason (such as an exception being thrown and not caught), you will crash on shutdown.

Testcase coming up as an attachment in a few minutes.

Comment 1

[Alex Vincent [:WeirdAl]]

Attachment: xpcshell testcase (will crash on shutdown!)

Comment 2

[timeless]

violates API:
04 xpcom_core!nsProxyObjectManager::~nsProxyObjectManager(void)+0x31
Hold Monitor:
0b xpcom_core!nsProxyEventObject::Release(void)+0xa0

This has nothing to do w/ the error console, it's either a bug in POM/PEO or a bug in XPCOM Startup/Shutdown, both of which are the responsibility of XPCOM (well, it could be the fault of the Error Service, but that too is in XPCOM, the error console is some ui).  

Comment 3

[timeless]

Attachment: this seems to work

Comment 4

[Alex Vincent [:WeirdAl]]

Comment on attachment 264453 [details] [diff] [review]
this seems to work

Why are you asking me to review?  I don't know this code and have never touched anything close to it.  I don't know what proxies mean in this context.  Never mind how simple the patch is, I don't think I'm an appropriate reviewer.

Comment 5

[Alex Vincent [:WeirdAl]]

I do not crash with timeless's patch, it does indeed fix things.

Comment 6

[Benjamin Smedberg]

Comment on attachment 264453 [details] [diff] [review]
this seems to work

The call to mProxyObject->LockedRemove(this) *must* be run within the POM monitor! In this case you're releasing the monitor before calling the destructor, which solves your bug but introduces unlocked access to the POM data structures.

Comment 7

[Alex Vincent [:WeirdAl]]

It's worth asking if anyone thinks this bug is still possible.  It's been over a decade.

Comment 8

[Negritas Sergiu, Desktop QA]

Hello Alex!

Is this issue still available? If so could you please provide some steps in order to reproduce it?
If the issue is not available anymore can we close it?

Comment 9

[Andrew McCreight [:mccr8]]

nsProxyEventObject doesn't seem to exist any more, and I'm not sure I've even heard of it, so let's just go ahead and close this.