1042348 - crash in libsystem_kernel.dylib@0x15866 (mozilla::ipc::GeckoChildProcessHost::PerformAsyncLaunchInternal)

Status: VERIFIED FIXED

(Core :: WebRTC, defect)

Description

[Nils Ohlmeier [:drno]]

This bug was filed from the Socorro interface and is 
report bp-c3fb9e06-34a6-4260-9e61-699c02140722.
=============================================================

Comment 1

[Nils Ohlmeier [:drno]]

This was after starting todays Nightly (34.0a1 (2014-07-22)), wait for the Plugin to become enabled and then try to place an H264 only call via webrtc-landing/pc_test.html

Comment 2

[Nils Ohlmeier [:drno]]

And another one: https://crash-stats.mozilla.com/report/index/5ae24aab-3397-46a1-9fff-1d0222140722

Comment 3

[Ethan Hugg [:ehugg]]

I saw this with today's Nightly (7-22) but only on OSX, the other two platforms worked for me.  Also only on the build from nightly.mozilla.org and not when I build locally without crashreporter.

Comment 4

[Maire Reavy [:mreavy]]

Georg -- I know you're fixing crash reporter now.  Can you look at the crash on this bug and tell me if this looks like an issue you're already fixing?

Comment 5

[Georg Fritzsche [:gfritzsche]]

Those look like hang reports (EXC_SOFTWARE / SIGABRT).
If they are, then they are dupes of bug 1041525 - Ethan?

Comment 6

[Georg Fritzsche [:gfritzsche]]

Hm, actually they are 34.0a1 crashes, so it can't be bug 1041525 exactly as that one is worked around.
Due to the similar stacks i'm suspecting this has a common cause with bug 1041525.

Comment 7

[Ethan Hugg [:ehugg]]

Here's one from my Mac:
https://crash-stats.mozilla.com/report/index/b2b60a27-194e-43de-af0b-119732140722

So it's probably the same cause.  They happen at the same time while setting up the 264 call.  Monday's build just hung with the wait icon indefinitely, Tuesday's build crashes to desktop.  Both were 100% repeatable for me.

Comment 8

[Maire Reavy [:mreavy]]

(In reply to Georg Fritzsche [:gfritzsche] from comment #6)
> Hm, actually they are 34.0a1 crashes, so it can't be bug 1041525 exactly as that one is worked around.
> Due to the similar stacks i'm suspecting this has a common cause with bug 1041525.

Georg -- Since this looks like it has a common cause with bug 1041525, I think it makes sense to assign this to you.

Comment 9

[Georg Fritzsche [:gfritzsche]]

This should be fixed by bug 1041525.

We're crashing on the std::vector<std::string>::push_back() line here:
http://hg.mozilla.org/mozilla-central/annotate/63f44b4968c2/ipc/glue/GeckoChildProcessHost.cpp#l668

Bug 1041525 disabled the crashreporter initialization, so we end up requesting an - uninitialized - pipe identifier (char*):
http://hg.mozilla.org/mozilla-central/annotate/63f44b4968c2/toolkit/crashreporter/nsExceptionHandler.cpp#l2666
http://hg.mozilla.org/mozilla-central/annotate/63f44b4968c2/toolkit/crashreporter/nsExceptionHandler.cpp#l279
... and end up constructing a std::string from it.

Comment 10

[Randell Jesup [:jesup] (needinfo me)]

Can you verify this is now fixed on inbound?  (or if you believe it is, close and if we're wrong it will be reopened or a new bug filed.)  Thanks.

Comment 12

[Liz Henry (:lizzard) (relman/hg->git project)]

Nils, is this something you could verify for us? If not maybe Alexandra can take it on. Thanks!

Comment 13

[Nils Ohlmeier [:drno]]

In all my manual tests I have not encountered a crash again. So I think this seems to be fixed.

Comment 14

[Bogdan Maris, Desktop Test Engineering]

Marking this bug as verified based on comment 13.