TPEPlatform/2013-11-21

Media

GetUserMedia

bug 940045 - The origin displayed in the persistent notification matches the location bar, not content \ SC
- TabParent send the recording-device-event with the URL of FrameLoader, but not the URL of iframe.
- need to clarify if any other safe way to get URL in b2g process, getting URL from content process seems unsafe.
- provide a proof-of-concept patch for sending request URL directly from content process

WebRTC

bug 938492 - clean-up after bug 869869 is landed \ SC
- follow-up bug for fixing contractID/interface name in PushService.jsm, remove obsolete TestUDPServerSocket.cpp
- r+'ed and landed.

bug 920921 - [Flatfish]: Flatfish has bad performance on Homescreen wiping /Jerry
- Check nsDisplayTransform::ShouldPrerenderTransformedContent()
- Check nsDisplayListBuilder::MarkOutOfFlowFrameForDisplay()

ImageBridge study /Jerry
- When will the buffer return?
  - In ImageBridgeChild::EndTransaction(), it sends buffer to compositor via SendUpdate() and wait the "TReplyTextureRemoved" command to return the buffer.

bug 929432 - [B2G][Helix][Browser][ChenHoulai] Open youtube in the browser, there is an extra horizontal line on the top when in full-screen mode /Peter
- modify the visible region of colorlayer and video layer but didn't solve the problem, debugging...

bug 934912 - QCT CAF patch for RB swap/copybit related patches
- Try to fix the required kernel header change. Due to the the patch that Diego Wilson provided need the HAL libs keep up with current CAF code base, The overall code base upgrading job need Michael Wu and Marco 's advisory to go further

bug 939348 - [b2g][Settings] Gralloc alloc takes 40ms, causes Settings page transition to be janky
- This issue should share same root cause with 939565 - Gralloc Allocation can take 25ms+ ; BenWa suggested to add a flag - MAP_POPULATE for mmap() calls. My experiments shows that does help for gralloc performance(from 25+ ms to about 19 ms). According BenWa's comment it might dues to mmap sets up the page lazily when the allocation is large. In this case we always memset the entire area so the pages fault-in because they aren't setup. Another reason that I am suspecting is the "write on demand" nature of linux kernel's memory management strategy. Diego just take the ownership of 939565 and will following up that issue.