Heap Corruption Crashes - Multiparty Rooms

[mknippen]

We performed tests with six users, and almost all of us had crashes with the current version (beta 8). Crashlytics noted at least 7 different crashes in the TwilioVideo codebase. I will be happy to provide stack traces for each one.

[mknippen]

Would you like each one as a separate GitHub issue, so it can be more easily tracked?

[idelgado]

@mknippen yes, that would be really helpful. Please add the stack trace for each and we'll start investigating.

[mknippen]

Attached finished demo results. We had 9 users, and 30 crashes. I will update with individual crash logs today.

[ceaglest]

Thanks for filing issues with stack traces @mknippen.

We have customers using the iOS SDK in production who are not seeing the kind of crash rates that you are, so we may contact you for more info on your particular use case. Nevertheless, there are some memory and threading bugs that we are still working to resolve, and will keep you in the loop as we address these issues.

Best,
Chris

[mknippen]

@ceaglest absolutely, and sorry for the GitHub notification overload (I'm sure). I'm here to help in anyway that I can. I'm able to hop on a phone call or screenshare to show you what I'm doing at any point if you would like.

To the best of my knowledge, I'm not hacking anything or using anything outside of the SDK you published. I make no reference to AVFoundation or AudioToolbox at all, and let the framework handle it all directly.

[ceaglest]

@mknippen Other than #78, and #85 it looks the issues have the same underlying problem - heap corruption. Each crashing call is trying to allocate some memory, but the system libraries are detecting corruption and crashing from there.

Edit: To provide a little more info about the use case, these crashes are seen in multi-party Rooms sharing audio.

[ceaglest]

Crash in consume_record.

Biggest crash on the demo. 8 crashes from 6 users in 30 minutes.

TwilioVideo.consume_record.txt

[ceaglest]

Crash in resip::ConnectionBase::preparseNewBytes due to heap corruption.

resip-ConnectionBase-preparse.txt

[ceaglest]

Crash in resip::ConnectionBase::getWriteBufferForExtraBytes due to heap corruption.

4 crashes from 4 users in 30 minutes.

resip-ConnectionBase.txt

[ceaglest]

Crash in cricket::P2PTransportChannel::IsPingable likely due to heap corruption.

2 crashes from 2 users in 30 minutes

cricket-P2PTransportChannel.crash.txt

[ceaglest]

Crash in resip::Helper::makeResponse due to heap corruption.

resip-Helper.crash.txt

[ceaglest]

Crash in rtc::ByteBufferWriter::~ByteBufferWriter() due to heap corruption.

rtc-ByteBufferWriter-crash.txt

[ceaglest]

Crash in sigslot::signal2<cricket::TransportChannel*, rtc::SentPacket const&, sigslot::single_threaded> likely due to heap corruption.

sigslot-signal2.crash.txt

[ceaglest]

Crash in resip::TransactionState::processSipMessageAsNew due to heap corruption.

resip-TransactionState.crash.txt

[ceaglest]

Crash in twilio::signaling::SipTU:process due to heap corruption.

twilio-signaling.crash.txt

[ceaglest]

@mknippen our latest 1.0.0-beta11 release should address the source of these heap corruption crashes. Give it a try if you have the chance.

[ceaglest]

Closing out this issue, since we haven't gotten any new reports of this type of crash since 1.0.0-beta11. Please re-open if you are still experiencing issues.