821241 - Occasional crash in [@ std::_Tree ... sipcc::PeerConnectionMedia::AddTransportFlow]

Status: RESOLVED WORKSFORME

(Core :: WebRTC, defect, P1)

Description

[Mark Banner (:standard8)]

This bug was filed from the Socorro interface and is 
report bp-52def9e7-42b1-41d3-adb7-3a74d2121213 and bp-f65ba811-5584-4946-b066-803f32121213
============================================================= 

STR:

1) Have two computers hooked up to the social API demo
2) Initiate a call
3) Accept on the other end
4) As the call exchange is going through, close the call at the initiation end

Results:

Occasionally it crashes, at the receiver with the stack reported.

0 	xul.dll 	std::_Tree<std::_Tmap_traits<int,mozilla::RefPtr<mozilla::MediaPipeline>,std::le 	xtree:1742
1 	xul.dll 	std::_Tree<std::_Tmap_traits<int,webrtc::ViEChannel*,std::less<int>,std::allocat 	xtree:1450
2 	xul.dll 	std::map<int,mozilla::RefPtr<mozilla::TransportFlow>,std::less<int>,std::allocat 	map:211
3 	xul.dll 	sipcc::PeerConnectionMedia::AddTransportFlow 	media/webrtc/signaling/src/peerconnection/PeerConnectionMedia.h:311
4 	xul.dll 	vcmCreateTransportFlow 	media/webrtc/signaling/src/media/VcmSIPCCBinding.cpp:2609

Comment 1

[Maire Reavy [:mreavy]]

Ekr -- Will this be fixed when the fix for Bug 820102 lands?

Comment 2

[Eric Rescorla (:ekr)]

I don't know. I need the full stack trace.

Comment 3

[Maire Reavy [:mreavy]]

Mark -- Can you retest with the current changes?  If it is still happening, can you catch a stack trace in gdb (or equivalent)?

Comment 4

[Mark Banner (:standard8)]

The full stack trace is accessible from the links in comment 0.

I can retest this in the next day or two.

Comment 5

[Mark Banner (:standard8)]

I've just been re-trying this, but due to other bugs, webrtc isn't stable enough for me to re-test this with any confidence in the results.

Comment 6

[Mark Banner (:standard8)]

This is WFM now for the steps in comment 0. The only slightly different crash I've found is in bug 835283, but that was crashing on the initiator and with a slightly different stack, hence I've filed a separate bug for it.