Benefits of this is that the send config previously had unclear locking
requirements, a lock was used to lock parts parts of it while
reconfiguring the VideoEncoder. Primary work was splitting out video
streams from config as well as encoder_settings as these change on
ReconfigureVideoEncoder. Now threading requirements for both member
configs are clear (as they are read-only), and encoder_settings doesn't
stay in the config as a stale pointer.
CreateVideoSendStream now takes video streams separately as well as the
encoder_settings pointer, analogous to ReconfigureVideoEncoder.
This change required changing so that pacing is silently enabled when
using suspend_below_min_bitrate rather than silently setting it.
R=henrik.lundin@webrtc.org, mflodman@webrtc.org, pthatcher@webrtc.org, stefan@webrtc.org
BUG=3260
Review URL: https://webrtc-codereview.appspot.com/20409004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@6349 4adac7df-926f-26a2-2b94-8c16560cd09d
Non-global logging isn't supported, and having a per-call logging
dispatch seems over-eager and adds more complexity than it's worth. The
current implementation is racy on initialization due to missing atomics
support. Besides, logging support should be separate from use of Call.
R=mflodman@webrtc.org
BUG=3250,3157
Review URL: https://webrtc-codereview.appspot.com/12329006
git-svn-id: http://webrtc.googlecode.com/svn/trunk@5971 4adac7df-926f-26a2-2b94-8c16560cd09d
Replaces VideoCodec in VideoSendStream::Config with an EncoderSettings
struct. The EncoderSettings struct uses an external encoder for all
codecs. This means that external users, such as libjingle, will provide
the encoders themselves, removing the previous distinction of internal
and external codecs.
For now VideoSendStream translates to VideoCodec internally. In the
interrim (before the corresponding change is implemented in
VideoReceiveStream) tests convert EncoderSettings to VideoCodecs.
Removes Call::GetVideoCodecs().
Disables RampUpTest.WithPacingAndRtx as its further exposed with changes
to bitrates used in tests.
BUG=2854,2992
R=mflodman@webrtc.org, stefan@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/7919004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@5722 4adac7df-926f-26a2-2b94-8c16560cd09d
Removing assertion that SSRC count should be the same as the number of
streams in the codec. It makes sense that you don't always use the same
number of streams under one call. Dropping resolution due to CPU overuse
for instance can require less streams, but the SSRCs should stay
allocated so that operations can resume when not overusing any more.
This change also means we can get rid of the ugly SendStreamState whose
content wasn't defined. Instead we use SetCodec to change resolution
etc. on the fly. Should something else have to be replaced on the fly
then that functionality simply has to be implemented.
BUG=
R=mflodman@webrtc.org
Review URL: https://webrtc-codereview.appspot.com/3499005
git-svn-id: http://webrtc.googlecode.com/svn/trunk@5123 4adac7df-926f-26a2-2b94-8c16560cd09d
