982072 - RegisterStrongMemoryReporter leaks on failure

Status: RESOLVED FIXED

(Core :: XPCOM, defect)

Description

[Andrew McCreight [:mccr8]]

I just noticed this while reading code, and it might be cause of bug 981966 (another possibility is the doomsday scenario, bug 980401).  How could we be leaking a memory reporter (OggReporter in this case) without any indication, if it is an ISupports class?  Well, one way for that to happen is to do |RegisterStrongMemoryReporter(new WhateverReporter())| and have the call to |do_GetService| fail.  The caller expects the register function to deal with it, so we leak, and we never called AddRef or Release on the reporter, so I think it is invisible to the XPCOM leak checker.  I haven't checked if this is actually happening for the OggReporter, but we want to fix this problem in any case.

Comment 1

[Andrew McCreight [:mccr8]]

Okay, Nathan pointed out that I was reading the stack incorrectly, so we're not actually leaking a reporter in bug 981966, so I think this isn't actually happening right now.  But it would still be good to fix, with at least an NS_ASSERTION or something on failure.

Comment 2

[(no longer active)]

Attachment: Don't leak the argument passed to RegisterStrongMemoryReporter if we fail to get the service; r=froydnj

Comment 3

[(no longer active)]

https://hg.mozilla.org/integration/mozilla-inbound/rev/ae7ee1ee97a3

Comment 4

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

https://hg.mozilla.org/mozilla-central/rev/ae7ee1ee97a3