535548 - Add `Abort` to the Socorro signature skiplist

Status: RESOLVED FIXED

(Socorro :: General, task)

Description

[Benjamin Smedberg]

The electrolysis code makes use of NS_DebugBreak(NS_DEBUG_ABORT) fairly readily, and all the "crashes" that result have a signature of `Abort`. Please skip Abort so that the crash is charged to the caller.

e.g. http://crash-stats.mozilla.com/report/index/d540bbb4-93d4-4a95-a152-1f4c92091216

Comment 1

[K Lars Lohn [:lars] [:klohn]]

For clarification, in the quoted crash you want the resultant signature to be "mozilla::ipc::AsyncChannel::Close" ?  If not, what would it be?

We actually have three types of skip lists for signatures.  Your answer to the question above will make it clear which list gets "abort".

Comment 2

[Benjamin Smedberg]

Yes, the signature should be mozilla::ipc::AsyncChannel::Close

Comment 3

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

I think it's actually preferable if it shows up as:
  Abort | mozilla::ipc::AsyncChannel::Close

(I'd prefer to completely hide frames only in the cases where the hidden frames vary but the ones under them don't, i.e., random numeric addresses.)

Comment 4

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

But, either way, it would be really good to get this done soon; there are a lot of trunk crashes from out-of-process-plugins (OOPP) that are hard to distinguish (and therefore hard to prioritize, and hard to notice new regressions) because of this.

Comment 5

[K Lars Lohn [:lars] [:klohn]]

the first method can be done with a configuration change.  The second as suggested by :dbaron would require recoding.  I'm opting for the config change instead of reworking the signature generation code.

see bug 537135 for progress

Comment 6

[Jesse Ruderman]

Why does dbaron's request require recoding?  There are already many functions that are appended rather than skipped.  (Can you CC me to bug 537135?)

Comment 7

[K Lars Lohn [:lars] [:klohn]]

just putting 'Abort' into the 'prefixSignatureRegEx' would result in a signature of 'Abort | @0xbdd0707 | mozilla::ipc::AsyncChannel::Close' instead of the 'Abort | mozilla::ipc::AsyncChannel::Close' that he requested.

When the code was first written, it worked like dbaron's request.  However, a competing mandate said that frame signatures should never be skipped in the middle of a compound signature.

See http://code.google.com/p/socorro/wiki/SignatureGeneration for the explanation of the signature generation procedures.

Comment 8

[Benjamin Smedberg]

Is this bug staying open for the compound signature change, or can it be marked FIXED?

Comment 9

[Benjamin Smedberg]

Calling this FIXED to get it off the Lorentz radar.

Comment 10

[Benjamin Smedberg]

This does not appear to have worked: http://crash-stats.mozilla.com/query/query?do_query=1&product=Firefox&version=Firefox%3A3.7a1pre&query_search=signature&query_type=exact&query=Abort is a crash from today, and the signature is still "Abort".

Comment 11

[Benjamin Smedberg]

Or ugh, I meant http://crash-stats.mozilla.com/report/index/0842a0a2-4061-4061-be47-d11592100115

Comment 12

[Benjamin Smedberg]

Unless... is the signature of that "Abort" because there are no other frames to fall back on?

Comment 13

[K Lars Lohn [:lars] [:klohn]]

as it says in the processor notes for that crash: "No proper signature could be created because no good data for the crashing thread (0) was found"