631633 - objc_msgSend no longer treated as a signature prefix

Status: VERIFIED FIXED

(Socorro :: General, task)

Description

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

I noticed this morning that objc_msgSend is no longer being treated as a frame to be prefixed; this means that tons of distinct Mac crashes are going to be lumped together.

The first Camino crash I see exhibiting this problem is at 18:47 2-2, so something around that time probably broke this (Camino's lower volume, so I'd say ±2 hours).

I suspect it was bug 630230 that broke this, based on the timeframe and FIXED bugs.

I'm unsure how many other signatures that are supposed to be prefixed were broken at this same time; objc_msgSend jumped right out at me, though.

Comment 1

[Benjamin Smedberg]

Hrm, it's ceratinly still listed: http://code.google.com/p/socorro/source/browse/trunk/scripts/config/processorconfig.py.dist#131

Comment 2

[Laura Thomson :laura]

the .dist file doesn't necessarily reflect prod.

Comment 3

[Laura Thomson :laura]

Prod config is:
prefixSignatureRegEx = '|'.join([
  '@0x0',
  'strchr',
  'strstr',
  'strlen',
  'PL_strlen',
  'strcmp',
  'wcslen',
  'memcpy',
  'memmove',
  'memcmp',
  'malloc',
  'realloc',
  '.*free',
  'arena_dalloc_small',
  'arena_alloc',
  'arena_dalloc',
  'nsObjCExceptionLogAbort(\(.*?\)){0,1}',
  'libobjc.A.dylib@0x1568.'
  'objc_msgSend',
  '_purecall',
  'PL_DHashTableOperate',
  'EtwEventEnabled',
  'RtlpFreeHandleForAtom',
  'RtlpDeCommitFreeBlock',
  'RtlpAllocateAffinityIndex',
  'RtlAddAccessAllowedAce',
  'RtlQueryPerformanceFrequency',
  'RtlpWaitOnCriticalSection',
  'RtlpWaitForCriticalSection',
  '_PR_MD_ATOMIC_(INC|DEC)REMENT',
  'nsCOMPtr.*',
  'nsRefPtr.*',
  'operator new\([^,\)]+\)',
  'CFRelease',
  'objc_exception_throw',
  '[-+]\[NSException raise(:format:(arguments:)?)?\]',
  'mozalloc_handle_oom',
  'nsTArray_base<.*',
  'nsTArray<.*',
  'KiFastSystemCallRet',
  '(Nt|Zw)?WaitForSingleObject(Ex)?',
  '(Nt|Zw)?WaitForMultipleObjects(Ex)?',
  'NtUserWaitMessage',
  'NtUserMessageCall',
  ])

which is actually the same as .dist.  As you can see, objc_msgSend is in there.

Comment 4

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

FWIW, the only other one I've found so far is the 'libobjc.A.dylib@0x1568.' entry, which is right before objc_msgSend in the .dist file.

So, we had this problem before in bug 557747 and apparently never figured out what caused it :(

Those frames (which were to be ignored, rather than prefixed as here) were not being ignored at all, even though they apparently *were* in the production config at the time (bug 557747 comment 6).  (The problem with those frames persists, although now they're not even in the SVN copy, so there's certainly no hope of them being in production.)

Comment 5

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

Exactly the same problem here as in bug 631036 comment 29 (yay rhelmer for finding it there) ;)

(In reply to comment #3)
>   'libobjc.A.dylib@0x1568.'
                             ^^^^ missing comma!

Any chance of getting this fixed, tested, and deployed before we accumulate lots of unprefixed objc_msgSend crashes over the weekend?

Comment 6

[Robert Helmer [:rhelmer]]

Attachment: fix typo

(In reply to comment #5)
> Exactly the same problem here as in bug 631036 comment 29 (yay rhelmer for
> finding it there) ;)
> 
> (In reply to comment #3)
> >   'libobjc.A.dylib@0x1568.'
>                              ^^^^ missing comma!
> 
> Any chance of getting this fixed, tested, and deployed before we accumulate
> lots of unprefixed objc_msgSend crashes over the weekend?

Checked in a fix for this:
Committed revision 2915.

I'll try to get this into bug 631690.

Comment 7

[Robert Helmer [:rhelmer]]

(In reply to comment #6)
> Checked in a fix for this:
> Committed revision 2915.
> 
> I'll try to get this into bug 631690.

This will roll with the other updates in bug 631690.

Comment 8

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

Thanks, Robert!

Comment 9

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

Verifying this; I no longer see new stand-alone objc_msgSend signatures being generated, and objc_msgSend | foo signatures started reappearing sometime after 21:00 last night :)