803744 - PeerConnectionImpl Initialize: StartGathering failed: 2147500037

Status: RESOLVED FIXED

(Core :: WebRTC, defect)

Description

[Christoph Diehl [:posidron]]

Attachment: NSPR_LOG

PeerConnection fails to initialize after some time of fuzzing SDP, although I try to close/destroy PC instances on failures and on success in each testcase. Each testcase gets loaded into Firefox as a new HTML page with PC calls.
This issue makes fuzzing SDP currently not possible. Let me know if you need further information or if you have workarounds.

2082140544[1037e9e80]: PeerConnectionImpl Initialize: StartGathering failed: 2147500037
JavaScript error: file:///Users/cdiehl/Code/Mozilla/mc-asan-opt/debug_build/dist/NightlyDebug.app/Contents/MacOS/components/PeerConnection.js, line 207: NS_ERROR_FAILURE: Component returned failure code: 0x80004005 (NS_ERROR_FAILURE) [IPeerConnection.initialize]

Comment 1

[Eric Rescorla (:ekr)]

Do you have a test case?

The natural workaround would be to terminate the browser between tests.

Comment 2

[Christoph Diehl [:posidron]]

I could perhaps create one with setTimeout(), I will see if that works.

Comment 3

[Christoph Diehl [:posidron]]

Attachment: testcase

Comment 4

[Christoph Diehl [:posidron]]

Click the run button, after a few seconds you get the message:

JavaScript error: file:///Users/cdiehl/Code/Mozilla/mc-asan-opt/debug_build/dist/NightlyDebug.app/Contents/MacOS/components/PeerConnection.js, line 211: NS_ERROR_FAILURE: Component returned failure code: 0x80004005 (NS_ERROR_FAILURE) [IPeerConnection.initialize]

Reload the page and click the run button again and you get:

Assertion failure: 0, at /Users/cdiehl/Code/Mozilla/mc-asan-opt/media/webrtc/signaling/src/peerconnection/PeerConnectionCtx.cpp:114

Comment 5

[Eric Rescorla (:ekr)]

So, what's the expected behavior here? Obviously, we can remove the assert, but the bottom line is that you are allocating a lot of resources with each PC and while they will eventually be garbage collected, it's not clear to me that that should happen in any finite time, so it's not clear to me that this represents a resource leak.

Comment 6

[Christoph Diehl [:posidron]]

Attachment: testcase

Testcase with calling .close() for each PeerConnection.

Comment 7

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

Comment on attachment 674395 [details]
testcase

This calls x.close() only for the last PeerConnection, not for each of the 100 you create.

Comment 8

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

With that fixed, I got this:
Assertion failure: 0, at c:/mozilla/objdir-debug/media/webrtc/signaling/signali
g_ecc/../../../../../inbound/media/webrtc/signaling/src/peerconnection/PeerConn
ctionCtx.cpp:114

This is a assertion due to PeerConnectionImpl being the wrong state.  Ethan?

Comment 9

[Ethan Hugg [:ehugg]]

Not sure yet if this device event that pops up should be valid or not.  Investigating.  I will upload fixed test case and hopefully patch soon.

Comment 10

[Christoph Diehl [:posidron]]

Attachment: testcase

JS != Python. Fix brackets in testcase.

This throws now an assertion failure as discussed above. 
I will look out for a suitable testcase for the initialization problem.

Comment 11

[Ethan Hugg [:ehugg]]

Attachment: Allow signaling startup when already started

Comment 12

[Ethan Hugg [:ehugg]]

Comment on attachment 674896 [details] [diff] [review]
Allow signaling startup when already started


I don't see why this device event should be an error if we are already in kStarted.  Don't think we want to Cleanup() in this case either.

Comment 13

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

https://hg.mozilla.org/integration/mozilla-inbound/rev/c107d4e66ee4

Comment 14

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/c107d4e66ee4