200424 - nsEventQueueServiceImpl::GetThreadEventQueue might return NS_OK w/ null out [@ nsMemoryImpl::FlushMemory]

Status: RESOLVED FIXED

(Core :: XPCOM, defect)

Description

[timeless]

I finally cvs updated raistlin. but my profiles are all in use, so mozilla
wasn't very happy.

WARNING: NS_ENSURE_TRUE(globalObject) failed, file
i:/build/mozilla/xpfe/appshell/src/nsWebShellWindow.cpp, line 1073
WARNING: NS_ENSURE_TRUE(globalObject) failed, file
i:/build/mozilla/xpfe/appshell/src/nsWebShellWindow.cpp, line 1073
WARNING: NS_ENSURE_TRUE(globalObject) failed, file
i:/build/mozilla/xpfe/appshell/src/nsWebShellWindow.cpp, line 1073
WARNING: nsTimeoutImpl::Release() proceeding without context., file
i:/build/mozilla/dom/src/base/nsGlobalWindow.cpp, line 4937
WEBSHELL- = 0
###!!! ASSERTION: Could not get profile directory lock.: 'Error', file
i:/build/mozilla/profile/src/nsProfile.cpp, line 1175
WEBSHELL+ = 1
### nsCacheProfilePrefObserver::Observe [topic=xpcom-shutdown data=]
nsPluginHostImpl::Observe "xpcom-shutdown"
###!!! ASSERTION: You can't dereference a NULL nsCOMPtr with operator->().:
'mRawPtr != 0', file i:/build/mozilla/xpcom/build/../glue\nsCOMPtr.h, line 691
Break: at file i:/build/mozilla/xpcom/build/../glue\nsCOMPtr.h, line 691

nsMemoryImpl::FlushMemory(const unsigned short * 0x002ec5a8, int 0) line 446 +
15 bytes
MemoryFlusher::Run(MemoryFlusher * const 0x00437390) line 191 + 43 bytes
nsThread::Main(void * 0x00437558) line 121 + 26 bytes
_PR_NativeRunThread(void * 0x00437658) line 433 + 13 bytes
MSVCRTD! beginthreadex + 307 bytes
KERNEL32! CreateFileA + 283 bytes

            rv = eqs->GetThreadEventQueue(NS_UI_THREAD, getter_AddRefs(eq));
            if (NS_SUCCEEDED(rv)) {
                PL_InitEvent(&mFlushEvent.mEvent, this, HandleFlushEvent,
DestroyFlushEvent);
                mFlushEvent.mReason = aReason;

                rv = eq->PostEvent(NS_REINTERPRET_CAST(PLEvent*, &mFlushEvent));

+	eq	{0x00000000}

So... I looked at nsEventQueueServiceImpl::GetThreadEventQueue and it looks like
it could theoretically return NS_OK+NULL:

Comment 1

[timeless]

Attachment: fix

Comment 2

[Darin Fisher]

Comment on attachment 119258 [details] [diff] [review]
fix

>Index: mozilla/xpcom/threads/nsEventQueueService.cpp

>+  // XXX: Need error code for requesting an event queue when none exists...
>+  if (!youngestQueue) {
>+    return NS_ERROR_FAILURE;

how about NS_ERROR_NOT_AVAILABLE?  that seems better than NS_ERROR_FAILURE.

r=darin with that change.

Comment 3

[timeless]

checked in