TPEPlatform/2013-08-22

Media

Web Audio

• bug 894249 - Regression: Youtube running in Browser app cannot interrupt background playing content / JW
  • trace code - AudioChannelService
  • coding - add a channel type to interrupt background content channel
  • AudioChannelService refactoring

• bug 903279 - Keep consuming power when the device is in sleep. / JW
  • debug wake lock

• bug 907303 - AudioNodeEngineNEON.cpp doesn't build on GCC 4.4.3 / JW
  • check build error

• bug 899935 - Stopping the MediaRecorder does not recieve a TRACK_EVENT_ENDED event from callback / Shelly
  • In the call of MediaRecorder::Stop(), we stop the recording by removing the encoder from the track union stream of recorder. Originally, we thought we should receive a TRACK_EVENT_ENDED from TrackUnionStream, but it turns out that removing the listener doesn't fire the track end event.
• bug 903758 - Media Recording - setting the recorder state to inActive on the stop method / Shelly
  • I don't think the fix is correct.
  • Please find more details about final solution of stopping/starting a recorder in bug 899935, comment 5.
• bug 888177 - [meta] Implement ImageCapture API / Shelly
  • Create WebIDL.

Streaming Media

WebRTC

• bug 904025 - gUM user story / slee
  • study MediaStream spec
• bug 902856 - Should MediaEngineDefaultAudioSource generate real data? / slee
  • Updated patch and waiting for review.
• bug 869869 - e10s for UDP socket / SC
  • writing test case for NrSocketIpc
• OMX codec for WebRTC
  • successfully encode/send/receive/decode VGA frames between 2 Peaks /John Lin
  • glitches
    • high frame dropping rate
    • low quality frame
    • test page stop responding

MediaRecorder

• bug 891704 - [MediaEncoder] Implement MP4Writer, migrate stagefright mp4 muxer / Alfredo
• bug 853356 - [B2G getUserMedia] Display camera/ microphone permission acquisition prompt, disucss with fabrice/alive. Implementation needs to be changed for the deicuss result. / Alfredo

Graphic

Performance Analysis

• bug 905589 - Enable hw composer for flatfish / Morris
  • Try several setting for hw composer but not work
• bug 905589 - Measure performance for homescreen on flatfish platform /Jerry
• bug 907055 - Trigger b2g composition without content transaction update /Jerry
  • Try to post a composition task into compositor message loop

Nexus 4 JB support

• bug 871364 - Support GonkNativeWindow on gonk-JB (Android 4.3) / Morris
  • try patch was sumitted by Sotaro, video and camera work now but not very stable (crash or freeze).

• SkThreadCanvas - separate rendering thread from main thread / Peter
  • Able to debug on chrome and check SkDeferredcanvas usage on chrome

• bug 899798 - [B2G] With Skia/GL, canvases get re-composited with B and R channels swapped
  • import patch from QCT CAF patch

• bug 907109 - [B2G]LayerTree analysis for b2g homescreen
  • check layer tree to see any performance improvement spots

Free Agent

InputMethod

• bug 847763 - Prevent virtual keyboard iframe from getting focus / Kanru
  • Study -moz-user-focus, nsFocusManager::CheckIfFocusable
  • Implement -moz-user-focus for HTML elements
    • Realized that -moz-user-focus was partially removed intentionally
    • Reviewers have different opinions.
  • Fix in-process iframe -moz-user-focus checking
    • Recursively check parent iframe if focusable
    • r+
  • Default shell.xul elements are -moz-user-focus:ignore.
    • Fixed for system app html:iiframe
    • Pending review.
• bug 891763 - Browser API: mozbrowserwindowresizeto and mozbrowserwindowresizeby events / Kanru
  • Add new events to mozbrowser.
    • Need to fire events from nsGlobalWindow
    • Reviewing
    • What is nsIDocShellTreeOwner? How to get the frame's inner size?
• bug 905567 - Expose navigator.MozInputMethod to desktop build / Kanru
  • Move b2g/components/MozKeyboard.js et al. to dom/inputmethod
    • Stole by bug 906096
    • Not sure if these two bugs are aiming for the same thing or not.

Misc

• bug 907117 - IonMonkey: build failed with DEBUG turned on / Kanru
  • Fix landed

• bug 859149 - <iframe mozbrowser>s' processes don't seem to get killed if the iframe is removed from the DOM too early / Shelly
  • Discussed with :cyu, but seems like we couldn't reproduce the zombie process anymore?

Study

VpX codec /rlin

• study and also study ffmpeg encoding flow
• Performance could be issues

webM 2.0 /rlin

• Can leverage source code from libvpx
• Seek still the issue for our design. but can solve by decoder player

Audio Capture

• gUM audio path: GetUserMedia/VoiceEngine/AudioDeviceModule/Encoding/Packetization / SC