Philippe Gorley authored 6 years ago and Philippe Gorley committed 6 years ago

Adds ability to retrieve the shm renderer info from a call id, in case
the client is launched mid call (else it can't show the call).

SinkClient now keeps track of its width and height.

Change-Id: Ie43c196c60de5e22825fc71ff404e99bbfbe9402
Gitlab: #59
Reviewed-by: Sebastien Blin <sebastien.blin@savoirfairelinux.com>

Philippe Gorley authored 6 years ago and Philippe Gorley committed 6 years ago

Change-Id: I5f8fb9802290974a9cc74f535ff22365b56e50c6

Change-Id: Iab6a12cca382c628dc7d3e983259fe2ce1010851

Improve loading performance when having more than one
Ring account by loading Ring accounts configuration
in parallel.

Change-Id: Ic39bc24551a20be2e98ee89221e8eb6e66cb09fd

Change-Id: Ieb4821822ecfb1bacfee5ad4abec96f861b63950

Adds support for single- and double-precision floating point samples in
the WavWriter.

Change-Id: I62f7dac3989b176a39d77882bafcfcae38a733f1

Can be used to call setInterruptCallback on encoders or decoders.

Change-Id: I7b17aa93f211936f67ca237e5fea3f266ae6ca83

Audio time bases are always the inverse of the sample rate, meaning the
increment is always the number of samples in a frame. Video usually has
a time base inverse of its framerate, but this is not always the case.

Change-Id: I50d2d84d073052f8b3a832e8b99725b9d66b12a8

Needed for audio streaming because the audio input needs to be accessed
at call level so it can tell the input to switch to a file or a
different input.

Getting a shared pointer to an AudioInput should only be done during
setup, so as to not walk the whole map in audio processing loops.

Change-Id: I49be1cb3c641b50e6f70356f330d40e1c27bef61

Philippe Gorley authored 6 years ago and Philippe Gorley committed 6 years ago

Most of the time, the frame resolution is unchanged, and the scaler only
needs to convert pixel formats.

Also avoids this crash when size is the same: https://trac.ffmpeg.org/ticket/5356



Change-Id: I4c1c3e0a8b5bba8c937317074e8a9c5652fca407
Gitlab: #58
Reviewed-by: Sebastien Blin <sebastien.blin@savoirfairelinux.com>

Philippe Gorley authored 6 years ago and Sébastien Blin committed 6 years ago

Don't bump version, as API is not broken. Breaks ABI, but we do not set
the SONAME tag, meaning we don't have an ABI version.

Change-Id: I5bc436f00e20eca7a789b0d8f5863d6622f82d0f
Reviewed-by: Sebastien Blin <sebastien.blin@savoirfairelinux.com>

AudioInput now manages audio frames, not AudioSender.
Makes AudioInput observable.
AudioSender subscribes to AudioInput.

LocalRecorder does not need to observe AudioInput or VideoInput because
these already record frames. Will be done in future work on the recording
system.

Change-Id: I011d742063386498d59b2962f7c333b999d0921c

Change-Id: I951cc45a6dda6c13c10e2dedb41fb2d19e8b0206

Hugo Lefeuvre authored 6 years ago and Philippe Gorley committed 6 years ago

There is currently no distinction between the "peer busy" and "we
replied busy after timeout" states, since both end in the BUSY state.

Add a new PEER_BUSY state allowing such a distinction:
* PEER_BUSY is set when peer replied busy
* BUSY is set when we replied busy to an incoming call

Bump daemon API number to major 7.0.0 since this is breaking the
current API. In fact, these changes should not break anything in
any well implemented client because unknown states should be properly
handled, but better check.

Change-Id: Id83f6db3d4524a91951b9945797f5fd2c019ff2f
Reviewed-by: Sebastien Blin <sebastien.blin@savoirfairelinux.com>

Change-Id: Id0fab10de383e71920959f5c025d0f2710f99d86

Adrien Béraud authored 6 years ago and Andreas Traczyk committed 6 years ago

Change-Id: I047f0f1a87466e440cafdb0984b311b8bcb5e5ed
Reviewed-by: Andreas Traczyk <andreas.traczyk@savoirfairelinux.com>

Philippe Gorley authored 6 years ago and Philippe Gorley committed 6 years ago

Change-Id: I86f08805d93c85eafa88c2a4fc61bcd092e1ea57

Philippe Gorley authored 6 years ago and Philippe Gorley committed 6 years ago

Adds unit tests for VideoScaler.

Change-Id: I38fc5edbd37ca11196c72db8aed7927d6371b678

Returns AudioFrame instead of raw AVFRame pointer when calling
AudioBuffer.toAVFrame, and take AudioFrame as parameter when appending
an AVFrame to an AudioBuffer.

No longer need to free AVFrame during encoding/conversions.

Change-Id: I28aa992a5483f84f6cb1a5157718c11c3a69518c

RingBufferPool::getData overrides the format of the AudioBuffer, so
resampling may be required.

Protects AudioFormat with a mutex to avoid data races.

Change-Id: I0a9d8686f142c192f912887b175a42bacb0c1a57

Resolution/codecs won't change without a SIP renegotiation.

Change-Id: Ie6d393caa5a3e631c7be14356a7929c0f11e781c

* Update restbed to latest master 
* Removed dropped Kashmir dependency
* Removed merged/obsolete patches (locale, async read, strand)
* Use branch including necessary findopenssl changes for contrib

Change-Id: I6cdf7de1005f82abeac77eb6bfb0002df83c8017
Reviewed-by: Philippe Gorley <philippe.gorley@savoirfairelinux.com>

Change-Id: I93cc9a7e2c365bf84f64ef6ffbd05131b76ccfae

Philippe Gorley authored 6 years ago and Philippe Gorley committed 6 years ago

Change-Id: I6d01dc501cad775d7561eaf1857393c79dfa2365

Philippe Gorley authored 6 years ago and Philippe Gorley committed 6 years ago

Change-Id: I0f8f9571a715fe26a7db61a43675efc50ce71564

Philippe Gorley authored 6 years ago and Philippe Gorley committed 6 years ago

Change-Id: I7af6bc775b08fc478704dfc5b595677a40c13a8f

Philippe Gorley authored 6 years ago and Philippe Gorley committed 6 years ago

SIP and TURN tests fail sporadically, and all recent additions to the
test folder have been under the unitTest folder.

Change-Id: Icac4827f53da3fbd72492cdfb024a003decbbbec

Philippe Gorley authored 6 years ago and Philippe Gorley committed 6 years ago

Change-Id: I4c70799ce157e97001e85052987835781d5152a7

Philippe Gorley authored 6 years ago and Philippe Gorley committed 6 years ago

Time base for audio should always be the inverse of the sampling rate.

Change-Id: I335a4bab1c29b5f0411dcb38a3bab379b48fa433

Makes it only visible to audio_input's translation unit during
compilation, instead of a class member.

Change-Id: Id4d5d6d134b244aca83d88825c939e10840d2131

Audio will soon be implemented using Observer/Observable, so move these
classes somewhere more sensible.

Change-Id: Iff0ca61a4c73c1ed7fa82ed91f8ca2d11afd65d9

Also avoids race conditions when stopping the recorder

Change-Id: I498b2f9802e647a05410bd5bcb5cf4248a02bcf4

RingBufferPool has all the methods necessary to interface with
RingBuffer.

Change-Id: I04b7cb0ac9b7033e1f1047d7f00708eb987bfdc7

Philippe Gorley authored 6 years ago and Adrien Béraud committed 6 years ago

Makes use of AVFrame's reference counting if possible instead of copying
the data.

Only bumps minor version because MediaFrame et al weren't in the ABI
until now.

Change-Id: I692e76230ed057c1ad4e46ab59ea5cfd163fb98d

AVFrame.linesize may contain some padding, use AVFrame.nb_samples
instead.

Change-Id: I5e46e89ac102b8dfd939c9ccbc3f51e73d995c6e

Philippe Gorley authored 6 years ago and Andreas Traczyk committed 6 years ago

CMake build fails on MinGW when trying to find argon2

Change-Id: Ibe9451cfd324cbfd196434e3cbf8f01e1ae41a1d
Reviewed-by: Andreas Traczyk <andreas.traczyk@savoirfairelinux.com>

Change-Id: I9bf15b7bffb89773a95963f662e05df7040f5f1a

- critical for building with iPhoneOS 12.0 SDK

Change-Id: I89f012d83d0403fb15b9071a28d1bb60a5ffb681

Busy tone is broken by design: since there is no way to play a
tone for a given amount of time, callFailure() was calling
stopTone() just after playATone(), meaning that the tone was only
played for a few milliseconds.

callBusy() was even more broken because it was calling checkAudio()
just after playATone(), meaning that (1) the tone would only play for
a few ms and (2) the tone would start again when the audio layer is
restarted. Even worse, callBusy was not calling stopTone(), meaning
that if an incoming call tone was being playing played then it would
never be stopped.

This feature might be reintroduced for accessibility purposes at some
point in the future, along with a refactoring of the tone system in
the daemon.

Change-Id: I15957e050688bfe5f5ce84f971d5a14b875028b9
Reviewed-by: Sebastien Blin <sebastien.blin@savoirfairelinux.com>

Adrien Béraud authored 6 years ago and Philippe Gorley committed 6 years ago

Change-Id: I21e3a2fade975f03fbdbde84f6fe6e94b8f51926
Reviewed-by: Philippe Gorley <philippe.gorley@savoirfairelinux.com>