Chrome 42 WebRTC Release Notes.

3,646 views
Skip to first unread message

Vikas Marwaha

unread,
Mar 5, 2015, 10:53:51 AM3/5/15
to discuss...@googlegroups.com

M42


Summary

Chrome 42 is a performance-focused build. The WebRTC team, along with many other Chrome component teams, have made performance improvements a priority for this release.


For example, https://codereview.chromium.org/759753005 improved GPU performance significantly on Windows, although it is outside the scope of the core WebRTC work and therefore not reported as part of these release notes.


In addition, this release turns IPv6 on by default, and provides a mechanism to force all WebRTC traffic through a VPN when desired.


Finally, Chrome 42 introduces the new getUserMedia permission bubble. It’s been a long time coming, but it turned out that a simple revamp for gUM was not possible, and as a result, all browser infobars have been migrated to this new model.  We encourage you to test this new way of asking users for microphone / camera permissions rigorously.


As usual, if you find an issue in Chrome, file a bug immediately and email the discuss-webrtc list.


Performance


  • Issue 450222 : On Chrome for Windows and Mac, we switched from the built-in MJPEG to Chrome's own decoder, as it is significantly faster.  On Mac this also avoids creating an extra decoding process for the integrated camera.


  • Issue 442392: On Chrome for Windows, implemented a non-blocking socket mode. This dramatically reduces the incidence of packets being dropped due to full socket send buffers, which could occur on a few problematic network interface / driver combinations.


  • Issue 4350: Enabled startup bitrate probing by default. This makes the ramp-up for the video bandwidth estimator much faster, and allows senders to reach HD video quality sooner.


  • Removed libpeerconnection.[dll|so], a loadable module, from Chrome.  Reduced Chrome’s binary size on Windows by 1.6MB and shortens initialization time of Chrome’s renderer process since this external module doesn’t have to be loaded.


  • Issue 2822: Several improvements in PeerConnection.getStats implementation makes contention a lot less frequent and significantly reduces memory churn.


Features


  • Gathering of IPv6 candidates is now enabled by default. This is currently in the process of rolling out; applications can still explicitly disable v6 if needed by setting googIPv6 to false in the PeerConnection ctor.


  • Issue 333752 & 4276: Added a mode which disables multiple routes for ICE, and thereby only exposes stun and relay candidates for the default route. While this may reduce QoS in several cases, it will ensure all WebRTC traffic is forced through the default route.

    This gives VPN users a way to force WebRTC traffic through their VPN. Note that this is a first step, and we expect to refine this solution in the future.


  • Issue 1805: Added stats for DataChannel in getStats api.  Additional stats are planned for M43; see issue 4321.


  • Issue 3774: Added support for trying alternate server (STUN 300 error message) for STUN/TCP.


  • Issue 4299: GICE is no longer supported in SDP offers. All applications should now be using standard (RFC5245) ICE.


Bugs


  • Issue 4269 : Fixed issue in sdp generated by Chrome when using ipv6 causing incompatibility between FF and Chrome. Now the default destination used by c line is ipv4 to avoid any incompatibility issues with FF.


  • Issue 3135 : Fixed problem in webrtc code, that might cause us to send NACK packets excessively.


  • Issue 3589: Fixed problem in offer sdp generated by Chrome, it was generating unused parameters for RTX 96 payload.


  • Issue 3868: Fixed problem in webrtc being unable to connect to TURN servers by hostname.


  • Issue 3927: Fixed problem, TURN server access not working for Chrome on Windows with tmproxy.exe running. Did a workaround to allow localhost IP even if it does not match the local TURN port.


  • Issue 4061: Fixed problem where invalid IPv6 candidates could be generated on Windows.


  • Issue 4133: Fixed problem with late audio packets in NetEQ.


  • Issue 4190: Fixed crash in VP8 on iOS armv7 devices.


  • Issue 4223: Fixed issue in Chrome, the signature_algorithm extension  was breaking the DTLS 1.0 handshake. The fix was merged to M41 as well.


  • Issue 4248: Fixed Android video renderer to support video frames with stride > width.


  • Issue 446908: Fixed problem where STUN checks that arrived prior to an answer message were ignored.


  • Issue 454869: Fixed problem in video capturing; on Lenovo machines the camera was using an incorrect colorspace and thereby capturing HD only at 10fps.

  • Issue 457239: Worked around BSOD in RCUVCAVS.sys on Win 8.1

Eric Davies

unread,
Mar 5, 2015, 3:48:55 PM3/5/15
to discuss...@googlegroups.com
Hi Vikas,

I pulled down Canary version 43, ran a webrtc program, and didn't see anything new in the permissions dialog (as compared to chrome 41). Is this something we need to code against? If so, can you supply a pointer to some docs for it?

Thank you,
Eric.

Philipp Hancke

unread,
Mar 5, 2015, 3:52:46 PM3/5/15
to discuss...@googlegroups.com
2015-03-05 12:48 GMT-08:00 Eric Davies <ericthe...@gmail.com>:
Hi Vikas,

I pulled down Canary version 43, ran a webrtc program, and didn't see anything new in the permissions dialog (as compared to chrome 41). Is this something we need to code against? If so, can you supply a pointer to some docs for it?

https://groups.google.com/d/msg/discuss-webrtc/0Oh7JYXyITg/ues6OfcDzT4J has the instructions -- I suppose it's still behind the flag?

Eric Davies

unread,
Mar 5, 2015, 4:15:45 PM3/5/15
to discuss...@googlegroups.com
Thank you Phillipp, now I see it. Looks like an improvement to me, if a bit bland and colorless.

--

---
You received this message because you are subscribed to a topic in the Google Groups "discuss-webrtc" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/discuss-webrtc/PwuzgUypYos/unsubscribe.
To unsubscribe from this group and all its topics, send an email to discuss-webrt...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.



--
Eric Davies

Serge Lachapelle

unread,
Mar 6, 2015, 2:39:16 AM3/6/15
to discuss...@googlegroups.com

It may have been rolled back yet again.... I'll update status here as soon as we find out.


You received this message because you are subscribed to the Google Groups "discuss-webrtc" group.
To unsubscribe from this group and stop receiving emails from it, send an email to discuss-webrt...@googlegroups.com.

Philipp Hancke

unread,
Mar 6, 2015, 3:08:14 AM3/6/15
to discuss...@googlegroups.com
2015-03-05 23:39 GMT-08:00 'Serge Lachapelle' via discuss-webrtc <discuss...@googlegroups.com>:

It may have been rolled back yet again.... I'll update status here as soon as we find out.

https://code.google.com/p/chromium/issues/detail?id=400285 (related to the permission bubble -- first bug filed) looks fixed btw. Even though my 42 is an outdated canary so check again please.

Victor Pascual Avila

unread,
Mar 13, 2015, 5:17:04 AM3/13/15
to discuss...@googlegroups.com
Any plans for H.264 support?
> --
>
> ---
> You received this message because you are subscribed to the Google Groups
> "discuss-webrtc" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to discuss-webrt...@googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.



--
Victor Pascual Ávila
Reply all
Reply to author
Forward
0 new messages