1399951 - rtc::FatalMessage::~FatalMessage in mozilla::MediaEngineWebRTC::EnumerateAudioDevices

Status: NEW

(Core :: WebRTC: Audio/Video, defect, P3)

Description

[Bob Clary [:bc] (inactive)]

Attachment: Ubuntu Debug Log

1. https://login.1688.com/member/signin.htm

2. 
mozalloc_abort | abort rtc::FatalMessage::~FatalMessage webrtc::ProcessThreadImpl::Start webrtc::VoEBaseImpl::Init mozilla::MediaEngineWebRTC::EnumerateAudioDevices

Nightly 57 at least. Happens on my Ubuntu VMs only. May be due to the fact they are VMs without audio. Does not happen on Fedora 26 or Windows VMs.

Comment 1

[Jan-Ivar Bruaroey [:jib] (needinfo? me)]

Is pulse audio installed?

Comment 2

[Bob Clary [:bc] (inactive)]

Yes.

Comment 3

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

This is failing the RTC_DCHECK() that verifies it hasn't called Start() before on the Process thread.  Debug-only check (DCHECK).

Since it's repeatable on linux, rr should make short work of it:
% rr replay
> c
<hit crash>
> <go up to ::Start()>
> watch -l thread_
(watch -l thread_.value?  thread_ is a std::unique_ptr<rtc::PlatformThread>)
> rc
<hits crash, this always happens>
> rc
You have now stopped on the first call to Start()

Or just run in gdb and find the first call to Start() (really you want to find the call to Start() with the same 'this' value as the crash; there might be more than one Process thread, though probably not)