GitLab

Instead of casting the observable object in the attached callback, have
the recorder's owner add the streams manually.

Change-Id: I5c79e32e470364b0c6a4253f50357a4bafade769

Makes the MediaRecorder an observer on AudioInput, AudioReceiveThread,
VideoInput and VideoReceiveThread, so they no longer need a reference
to the recorder.

Makes sure the framerate is set for the recorder by adding a way to get
the stream parameters from the video receiver and input, since AVFrame
does not carry framerate and time base information.

Adds way to detach the recorder from the media sources, so it can be
reattached on subsequent records.

Change-Id: I71bac0825541f6445861f4f949288e1c62416f05

1. Adds constness to streams in MediaRecorder; the latter shouldn't change
stream parameters.
2. addStream is called from recordData to simplify workflow; users need
only check that MediaRecorder is recording.
3. Fixes regression where multiple records couldn't happen in the same
call.
4. Fixes AudioSender not sending starting timestamp to help mix audio.
5. MediaStream::name is now used as key for the streams_ map, so it must
be constant, filter names are no longer hardcoded in the recorder.
6. fromPeer parameter is determined by checking if MediaStream::name
contains the string "remote".
7. Adds copy constructor to MediaStream.
8. Renames incrementStreams to incrementExpectedStreams to better reflect
what it does.
9. Removes keyframe triggers when starting a video record; we are
recording raw frames, so it doesn't matter if they're keyframes or not.
10. MediaRecorder supports more than 2 audio streams, but not video.
11. Move preview to very bottom right corner.
12. Fix potential Unchecked Return Value to NULL Pointer Dereference in
recordData() (CWE-690).

Gitlab: #39

Change-Id: Id2fae4e9bb2072994c065e4843cc3ad832f89efc

It better explains what the method does, as the recorder is not started
at that moment).

Change-Id: Ib536600b663e9ee188257b3b166f497f5a2b2c1e
Reviewed-by: Sebastien Blin <sebastien.blin@savoirfairelinux.com>

MediaRecorder no longer uses VideoSender (encoder) as a source for its
local video and will directly use VideoInput (decoder), as it is earlier
in the pipeline.

This is the first step to support switching inputs while recording.

Change-Id: Ia163efa3b20a349a93fc7b05213ec5e00de1704e
Reviewed-by: Sebastien Blin <sebastien.blin@savoirfairelinux.com>

Change-Id: Ib25362e474af6cecee59c4d3738473bfafcd766f
Reviewed-by: Philippe Gorley <philippe.gorley@savoirfairelinux.com>

If there are 2 video streams, overlays local video over peer video. Peer
video will be upscaled if it is too small, while local video will be
scaled to 1/16th the size of the peer video.

If there is only 1 video stream, uses peer video settings if they are
valid, or fall back on local video settings.

Adds title and description metadata to recorded file.

Change-Id: I182de013e3ac2d18161eadf406fc6494e5f59a4e

Compresses audio to Ogg/Opus.

Refactors audio recording to use MediaFilter (audio mixing) and
MediaEncoder.

Change-Id: Ib8ee63ac71910782fb21d73d7da5fa5d19893b8e
Reviewed-by: Sebastien Blin <sebastien.blin@savoirfairelinux.com>

Change-Id: I1ef88bc640f79790aad2b5285ea2c776aa8cd366

Falling back from hardware to software decoding requires only a
reinitialization of the AVCodecContext. Hardware decoding failure is no
longer bubbled up to the RTP session, which makes it much faster.

Change-Id: I79fdfcfa41f822b3299d74ac654146789fcfd97b

Replaces empty methods and std::bind calls with lambdas.

Change-Id: I0d9c876a9ba0bbdc6a77a85e0e2d75766312f952

Fixes a regression where keyframe requests were sent by the wrong peer.

Corrects artifacts in a shorter amount of time.

Change-Id: I2b981b46c07422a4289f378e2a5deaaba0047a3b
Reviewed-by: Guillaume Roguez <guillaume.roguez@savoirfairelinux.com>

Keyframes will only be sent on request instead of a timely interval.
This may interfere with clients that do not yet have this patch.

Change-Id: Ie786160744257052ef9d103d50fe46cd90138927
Reviewed-by: Olivier Soldano <olivier.soldano@savoirfairelinux.com>

This patch follow-up the commit

[dffb0d6d video: save codec bitrate when quality is not auto]

This patch makes a better version by giving a correct name to the method
that setup the video bitrate and make sure its called when input of
this method (send_.codec) is modified.
Resulting dead code is removed.

Change-Id: I78ea8e318cecf516f4a018a341130c58f68e5845
Reviewed-by: Olivier Soldano <olivier.soldano@savoirfairelinux.com>

Commit 1ed6a0fe revealed a
case where codec quality is not saved and restored correctly
when holding/un-holding if quality is not set to auto.

The videoBitrateInfo_ struct of the VideoRtpSession is left to its
default value where quality is equal to 0. When un-holding a call,
the quality value to be used by the codec is then equal to 0 which
put the codec in CBR mode. As the default bitrate on H.264 codec is
very low (250 kbits/s), very poor quality was noticed on average resolution.

We now initialize videoBitrateInfo_ with the default quality settings
which makes the restored quality consistent on un-hold.

Change-Id: I23ef27cae953bcf60f1f90b18e042c3affd8ecd0
Reviewed-by: Guillaume Roguez <guillaume.roguez@savoirfairelinux.com>

Decrease the manager.h file dependency map by removing
uneeded inclusion and small code refactoring to nicely decoupling
classes.

Change-Id: Ibde69e0a088e7e2fbd842e39301db5b3a4bc8c28
Reviewed-by: Andreas Traczyk <andreas.traczyk@savoirfairelinux.com>

This implementation uses gnutls dtls heartbeat API to test path MTU.
heartbeat allowing messages with automated response in a datagram,
the application is able to guess the MTU via a timeout in the heartbeat.
(timeout on packet sent and no response, implies that the MTU is lower
than the lost payload.)
To minimize false positives (a response is lost for example), each attempt
triggers one retry on the first timeout.
This version ensures a minimal MTU of 512 bytes will be returned in
case of any failure in the procedure.
For retrocompatibility with non heartbeat capable clients,
a fallback MTU is set at 1280.

Change-Id: Ib9a7f63a70e8bdad239d8fc103779a0f2c387e87
Reviewed-by: Andreas Traczyk <andreas.traczyk@savoirfairelinux.com>

Change-Id: I6dbc51dd41c1f44d1446fe7365026af037bc86d9
Reviewed-by: Stepan Salenikovich <stepan.salenikovich@savoirfairelinux.com>

Fixes an issue where falling back to software decoding fails
if quality is not set to auto (no video).

Change-Id: I502558f82cbabdf90990cd9177a36674fb3df4d1
Tuleap: #1088

Hardware acceleration is now correctly enabled/disabled using Ring's
config file. The setting was moved from DeviceParams/VideoSettings,
which are used for local video, to preferences under video.

The video preferences object calls serialize() on the video device
preferences, so as to not overwrite one another (they're under the
same section in the config file).

Change-Id: I8c11ea8813dee5cc514a4fa4c1bae962a14577eb
Tuleap: #1122

This patch also implements a way for the Call object to restart
the media receiver

Change-Id: I1090575f14425d383eb5471a388c4c9e8d888b0e

Fixes spelling-error-in-binary lintian warning

Tuleap: #350
Change-Id: I014ebe76e71f9bae247d4d0726fd3e6cbe6459f2

tools/update-copyright script used for that

Change-Id: I89b13300d9ccbdbcd47dba679ba64be82499e83d
Tuleap: #454

- remove videoManager global instance.
- allocate a VideoManager instance when Manager instance is created.
- add Manager::getVideoManager() method to get this instance.
- force getSignalHandlers() to be called before Manager instance creation.

All of this prevents VideoManager to access to signal handlers map,
already destroyed, causing a memory access violation.

Change-Id: I21ebff08fa8d49e797f99299de12dab023c37769
Tuleap: #413

Based on RTCP information. Video quality was often toggling.
To avoid that, we only allow to decrease quality if a higher quality
was already calculated in of the 5 previous tries.

Tuleap: #158
Change-Id: I9a42a87f7958e4477dbebecc4c24c6a3d000b835

Add codec properties accessible by client and LRC to enable or
disable adaptative algorithm

Change-Id: Ia1baaad8a4fec6713e9657af4759ffe071be4ab5
Tuleap: #215

When several calls were performed, quality and bitrates of previous
call were used. When stopping VideoRtpSession, we make sure to reset default
values.

Change-Id: Ibe7b58f33979fae734dbbe3f7c31bd4d57c5a13f
Tuleap: #156

If a keyframe request is never answered, a timeout is reached
and the call failed.
As keyframe requests are transmit over UDP, they are too easily dropped,
on connection like wifi.

This feature is not mandatory for communication and until we found
a better solution this patch disables the request.

Change-Id: I3d5e8d0e3d0119db52fbfc0f37ce2949224b1806
Tuleap: #167

Dynamic quality change was done by increase/decreasing encoder bitrate.
It led to video artefact issues, specialy at low bitrates.
In addition, it does not make sence to use same bitrate with different
resolutions and fps. With high resolutions and fps, maximal bitrate
need to be higher.

It is possible for H264 and VP8 to set video quality using CRF.
This parameter solves video artefact issues and choose a different
output bitrate depending on input resolution and fps.

Tuleap: #80
Change-Id: I07d37175dfc335b5fc70cf54798f82eafe921ea9

Calling this API causes a race condition with RtpSession::start().
The former needs the latter to be run first, this was not enforced,
causing a crash by nullptr access on socketPair_ member.
This patch checks for a non-nullptr socketPair_ to procceed the restart.

Issue: #81987
Change-Id: I20886fe70398f018621aade1c7bac3ff15ef9055

During a conference, video display blink on a full black image
at each video restart (due to adaptative bitrate algorithm).
This patch fixes this effect due to an unvoluntary mixer's sink
start/stop.

Issue: #82044
Change-Id: I181051de7230189d0a951f65f02b5673f202ffb5

If video_rtp_session is restarted due to bitrate changement
we need to setup again conferencepipeline. Otherwise videomixer sink
would be never attached and video never send.

Issue: #81547
Change-Id: I6de180afab5a7dd320fe3b31a1d61735106d76d1

std::stoi is not available on Android

Issue: #78221
Change-Id: I3cf69dba62c8d4e5bf3f946340cd44a8c08a788e

Warning only if bitrate is effectively changed.

Issue: #78984
Change-Id: I6203d88ad519d448e33aa016941ff92d262427c5

Based on RTCP RR packets info, this patchset add a dichotomous algorithm that
converge to highest bitrate value.
Minimal and maximal bitrates are set by account.
Packet lost rate threshold is defined by call to 1%.

Issue: #78984
Change-Id: I8864a5e83bf93146e8bd0186bab0dde36fcd0348
Signed-off-by: Guillaume Roguez <guillaume.roguez@savoirfairelinux.com>

- remove OpenSSL exception
- fix Savoir-faire Linux naming
- fix common Author: representation

Issue: #80663
Change-Id: I6c3b2ca1ed48ed474a0ecd5a30fe793526e11b00

When bitrate is changed, new RTP session created must rely
on RTP sequence of previous RTP session in order to provide
continous sequence numbers.
This patchset ensures to keep last RTP sequence updated correctly
ensuring that no new packet will be sent when bitrate change is performed.

Issue: #80058
Change-Id: Ia7fbecd19aa0ec648327307de3ce3f8eace163a0
Signed-off-by: Guillaume Roguez <guillaume.roguez@savoirfairelinux.com>

- pass by-ref conference object in private API
- un-pairing enterConference/exitConference
- fix a nullptr access in enterConference

Issue: #79075
Change-Id: I06f0b72c9ee39f264937b9a36ae3d2dcbc418dab

An unwanted behavior of rtp stop method was to nullify conference
pointer, so remove attachement between and prevents video mixing.

Issue: #79075
Change-Id: Ic08dde0759d4116ffcdb6e90310f59b9a96b3b5f

All rtp_session::start() have mostly the same code, only how
the socket_pair is created change.
We also add "override" C++11 keywords that prevents possible
writing errors.

Issue: #79075
Change-Id: I7ff281a7d72661b0d44d867aa7925f769f93e71d