GitLab

Change-Id: Ib88d0d20c3ef536eed27e32af3d20153c0ff5462

This patch fix the problems:
- exit full screen mode when call is finished
- clean up last video picture when call is finished
- hide call options if no mouse activity detected
- resize video to use full screen

Change-Id: I7544dd616e88e35cacf3ca8c60d65aedbfe14860
Reviewed-by: Andreas Traczyk <andreas.traczyk@savoirfairelinux.com>

Height and width contained in Video::Frame structure are now used to
determine the size of a frame instead of the possibly incorrect
Renderer::size() values. This prevent some segfault when hardware
acceleration for decoding is used with certain size of picture.

Change-Id: I22e928eb0b9a0889d0b4ff861bb93dfba16d1bc6
Reviewed-by: Kateryna Kostiuk <kateryna.kostiuk@savoirfairelinux.com>

The previous system to display the incoming video frames was relying
on CGImage generation with raw framebuffer which were then set as the
content of the CallView. This way of doing it is not efficient as it
implies buffer copies and is discouraged by Apple for pictures that
change often. Moreover, this process was done by the
VideoReceiveThread from the daemon which was then blocked by those
copies without being able to decode further incoming frames. This is
why a lag was appearing and increasing on high resolution stream.

The new system now isolates frame delivering to the UI and their
rendering. The VideoReceiveThread just update the current frame buffer
and size without copy and another thread send those data to an OpenGL
texture on screen refresh which also enables to automatically skip
frames in case of heavy load.

Change-Id: I0b79ddce66f52a3db1eee19945733ff93e7ce34f
Reviewed-by: Kateryna Kostiuk <kateryna.kostiuk@savoirfairelinux.com>