MozillaWiki

TPEPlatform/2013-07-18: Difference between revisions

< TPEPlatform
 
(21 intermediate revisions by 7 users not shown)
Line 1: Line 1:
==Media==
==Media==
'''Web Audio'''
* {{bug|878015}} - WebAudio ASSERTION: Something went wrong with rounding to block boundaries: 't == aTo' / JW
** debug the test case
** trace MediaStreamGraphImpl code
** submit a patch
'''Media Recorder'''
* MP4 muxer / Alfredo
** study MP4 container format
** survey ffmpeg


==Streaming Media==
==Streaming Media==
<b>RTSP {{bug|831645}}</b> / Bechen
'''RTSP {{bug|831645}}''' / Bechen
*  RTSP {{bug|831645}}
*  RTSP {{bug|831645}}
** waiting for review
** waiting for review
Line 8: Line 17:
** We found that the first few packet's timestamp is negative
** We found that the first few packet's timestamp is negative
*** We should parsing the RTP-Info in RTSP packet instead of the timestamp is RTCP sender report packet.
*** We should parsing the RTP-Info in RTSP packet instead of the timestamp is RTCP sender report packet.
'''WebRTC'''
* memcpy from GraphicBuffer to system memory / Steven
** run on a single test case with multiple memcpy - result is good
** Lock with WRITE permission - still slow
** try memcpy on camera capture thread - processing
** try stop composite thread when copy - processing
* {{bug|869869}} - e10s for UDP socket / SC
** update patch according to 5th review comment, part 1 got r+.
* checking latest version of [http://www.w3.org/TR/mediacapture-streams/ gUM spec] / SC
** studying the difference between current implementation and latest spec
* inject dummy codec into WebRTC processing pipe / John Lin
** create dummy VideoEncoder/VideoDecoder implementation and register it as external codec to WebRTC codec DB.
** pass WebRTC module intergration test.
* {{bug|853356}} - Permission check for microphone acquisition / Alfredo
** connect to B2G ContentPermissionPrompt in C instead of JS


==Graphic ==
==Graphic ==
Line 33: Line 58:
*Study compositor rendering flow/ Morris
*Study compositor rendering flow/ Morris
*Study TextureClient/TextureHost data flow /Jerry
*Study TextureClient/TextureHost data flow /Jerry
'''MemShrink'''
* {{bug|771765}} - [Nuwa] Study browser-element test cases /Cervantes
** Mochitest appears to be broken on device.
** Investigate the usage of marionette instead.
'''OTA'''
* {{bug|891009}} - [OTA] If the download of an update is stopped due to an instable network it cannot be continued until the device gets restarted / SC
** The selection state is not reset after severe network error, therefore, passing the same Update object when manually re-download be treat as an inoperable update.
** patch submitted and waiting for review.
* Try establishing the FOTA build environment for inari/ikura, which can help QA to test the data migration. / SC
** ongoing, need some time to figure out the makefile problem in B2G.
'''perf'''
* perf porting / Steven
** I can run perf on peak and helix - will have an simple introduction
'''Event Handling'''
* {{bug|845169}} - Target events "properly" at overlapping app frames /Kanru
** Three method to attack this
*** Send touchable region after each composition
*** Use compositor to decide which process to dispatch
*** Synchronously ask remote process for hit detection
** Currently trying method #1
* {{bug|847763}} - Prevent virtual keyboard iframe from getting focus /Kanru
** Investigating why the keyboard couldn't receive the touch events when OOP
** It turns out we need to fix {{bug|845169}} first for OOP case
'''Study'''
* read doc / JW
** [http://martinfowler.com/bliki/RoleInterface.html role interface]
** [http://martinfowler.com/bliki/HeaderInterface.html header interface]
** [http://www.objectmentor.com/resources/articles/isp.pdf The interface segregation principle]
** CGDB debugging
* code base study / John Lin
** WebRTC A/V coding module
** stagefright (GB & JB version)

Latest revision as of 02:48, 29 July 2013

Media

Web Audio

  • bug 878015 - WebAudio ASSERTION: Something went wrong with rounding to block boundaries: 't == aTo' / JW
    • debug the test case
    • trace MediaStreamGraphImpl code
    • submit a patch

Media Recorder

  • MP4 muxer / Alfredo
    • study MP4 container format
    • survey ffmpeg

Streaming Media

RTSP bug 831645 / Bechen

  • RTSP bug 831645
    • waiting for review
  • bug 892395 - Rtsp: calibrate rtp timestamp in the beginning of media streaming.
    • We found that the first few packet's timestamp is negative
      • We should parsing the RTP-Info in RTSP packet instead of the timestamp is RTCP sender report packet.

WebRTC

  • memcpy from GraphicBuffer to system memory / Steven
    • run on a single test case with multiple memcpy - result is good
    • Lock with WRITE permission - still slow
    • try memcpy on camera capture thread - processing
    • try stop composite thread when copy - processing
  • bug 869869 - e10s for UDP socket / SC
    • update patch according to 5th review comment, part 1 got r+.
  • checking latest version of gUM spec / SC
    • studying the difference between current implementation and latest spec
  • inject dummy codec into WebRTC processing pipe / John Lin
    • create dummy VideoEncoder/VideoDecoder implementation and register it as external codec to WebRTC codec DB.
    • pass WebRTC module intergration test.
  • bug 853356 - Permission check for microphone acquisition / Alfredo
    • connect to B2G ContentPermissionPrompt in C instead of JS

Graphic

Canvas 2D/SkiaGL bug 858237

  • bug 875168 Integrate SurfaceTexture / Peter
    • prepare design diagram review

HWC

  • bug 884188 High power consumption of youtube streaming / Peter
    • Debug image layer on b2g18 branch

Misc

  • bug 858914 Layer refactoring about textureclient/host / Peter
    • check change scope
  • bug 894262 Merge GrallocPlanarYCbCrImage to GonkIOSurfaceImage / Peter (Implementation)
  • bug 894847 WebGL is horrible slow when hardware composer unavailable/ Morris
    • Analysis: Still try to find root cuase of this problem

Stability

Free Agent

  • Study compositor rendering flow/ Morris
  • Study TextureClient/TextureHost data flow /Jerry


MemShrink

  • bug 771765 - [Nuwa] Study browser-element test cases /Cervantes
    • Mochitest appears to be broken on device.
    • Investigate the usage of marionette instead.


OTA

  • bug 891009 - [OTA] If the download of an update is stopped due to an instable network it cannot be continued until the device gets restarted / SC
    • The selection state is not reset after severe network error, therefore, passing the same Update object when manually re-download be treat as an inoperable update.
    • patch submitted and waiting for review.
  • Try establishing the FOTA build environment for inari/ikura, which can help QA to test the data migration. / SC
    • ongoing, need some time to figure out the makefile problem in B2G.

perf

  • perf porting / Steven
    • I can run perf on peak and helix - will have an simple introduction

Event Handling

  • bug 845169 - Target events "properly" at overlapping app frames /Kanru
    • Three method to attack this
      • Send touchable region after each composition
      • Use compositor to decide which process to dispatch
      • Synchronously ask remote process for hit detection
    • Currently trying method #1
  • bug 847763 - Prevent virtual keyboard iframe from getting focus /Kanru
    • Investigating why the keyboard couldn't receive the touch events when OOP
    • It turns out we need to fix bug 845169 first for OOP case

Study

Retrieved from "https://wiki.allizom.org/index.php?title=TPEPlatform/2013-07-18&oldid=686604"