711953 - [skiplist] Add msvcr.*\.dll.* to prefixSignatureRegEx

Status: RESOLVED WONTFIX

(Socorro :: Infra, task)

Description

[Scoobidiver (away)]

Please add:
'msvcr80.dll@0x.*'
'msvcr90.dll@0x.*'
'msvcrt.dll@0x.*'
to the skiplist as irrelevantSignatureRegEx

Comment 1

[Robert Kaiser]

Not sure if we really want to hide when something happens inside the C libraries. Is libc.so also on the irrelevant list?

Comment 2

[Scoobidiver (away)]

(In reply to Robert Kaiser (:[e-mail]) from comment #1)
> Is libc.so also on the irrelevant list?
Yes. See https://github.com/mozilla/socorro/blob/master/scripts/config/processorconfig.py.dist

Comment 3

[Scoobidiver (away)]

If this bug was fixed, bug 718192 could have been discovered before.

Comment 4

[Scoobidiver (away)]

Add also:
'msvcr100.dll@0x.*'

Comment 5

[Robert Kaiser]

I propose we add msvcr.*\.dll.* to prefixSignatureRegEx here - Ted, Benjamin, does that sound reasonable, or would a different step be good here?

Comment 6

[Andrew McCreight [:mccr8]]

Kyle knows about Windows, maybe he has some idea.

Comment 7

[(not currently active) Ted Mielczarek]

We should be getting symbols for these. They don't change that often, so I don't think we should skiplist them. We did have a bug where Windows system symbols weren't being updated, but that ought to be fixed now.

Comment 8

[Scoobidiver (away)]

Windows 8 has no symbols, and for released Windows, there is still this kind of crash signatures. See https://crash-stats.mozilla.com/query/query?product=Firefox&version=ALL%3AALL&range_value=4&range_unit=weeks&query_search=signature&query_type=contains&query=msvcr&reason=&build_id=&process_type=any&hang_type=any&do_query=1

Comment 9

[Robert Kaiser]

(In reply to Ted Mielczarek [:ted] from comment #7)
> We should be getting symbols for these. They don't change that often, so I
> don't think we should skiplist them. We did have a bug where Windows system
> symbols weren't being updated, but that ought to be fixed now.

Well, we still get enough reports in with those frames as "signatures".

Comment 10

[Robert Kaiser]

Kyle, Ted, what should we be doing here?
We added libc.so even to irrelevantSignatureRegEx, see https://github.com/mozilla/socorro/blob/master/scripts/config/processorconfig.py.dist#L134

Comment 11

[Kyle Huey (Exited; not receiving bugmail, old account, do not use)]

I don't have a problem with ignoring them if we legitimately don't have symbols for them, but comment 7 claims we should ...

Comment 12

[Scoobidiver (away)]

In blocking bugs, two are classified as top crashers: bug 879354 and bug 888614.

Whatever the reason of missing debug symbols, implementing the skiplist will help breakdown this unique signature and see whether they are really top crashers to take care of.

Comment 13

[Scoobidiver (away)]

(In reply to Ted Mielczarek [:ted.mielczarek] from comment #7)
> We should be getting symbols for these.
Unless it's a MS Visual C++ Redistributable DLL that comes bundled with the faulty program such as Flash or Whitesky's ID Vault.

Comment 14

[(not currently active) Ted Mielczarek]

Shouldn't matter, we should be able to get symbols for those from Microsoft's symbol server. The redistributable DLLs are still built by Microsoft.

Comment 15

[Benjamin Smedberg]

FWIW, there is a problem specifically with either the symbols or our stackwalking for the pure-virtual handler; when I load bug 888614 minidumps in Visual Studio you can see that the crash is in that function, but somehow that information is not translated to breakpad. I'll investigate that, but I think overall this should be WONTFIX.