Describe the bug
Some of the tests on macOS are flaky (they sometimes work, and they sometimes fail unexplicably).
Frequently caused by sockets (socket.io) connections.
Also sometimes it is because of broken pipes (SIGPIPE).
Also sometimes, tests just fail.
This does not happen in GitHub's CI macOS runners, only on local machines.
There might be two different problems: websockets (socket.io) and datagrams (low-level libwebrtc).
The socket function fails inside of libwebrtc (here).

To Reproduce
Run tests on macOS multiple times.

Expected behavior
The tests always succeed.

Desktop (please complete the following information):

• OS: macOS
• Browser: N.A.
• Version: N.A.

Additional context
This could be caused by a security feature of macOS 13, maybe. This is worth exploring.

Add your name to as authors. Add IntRoLab @ UdeS as sponsor.

In the web client it's relatively easy to open and close some track of the stream using MediaStream and MediaStreamTrack :

let localstream = null;
openteraWebrtcWebClient.devices.getDefaultStream().then(stream => {
    localStream = stream;
});

// Mute mic
localstream.getAudioTracks().forEach(track => track.enabled = false);
// Close camera:
localstream.getVideoTracks().forEach(track => track.enabled = false);

A set of function like mute(true) and cameraOpen(true) could be useful particularly for the C++ and the Python client where the implementation of this features could be more complexe.

I tried to build an android project using C++ but found that the library does not support building under the NDK

Keep the joss_paper in a separate branch and remove it from main since it was not published.

Currently, some tests hang or fail on Windows IF they are run all at once. Every test can pass if ran individually.
This could be investigated and fixed.

Partly blocked by #82.

See https://developer.mozilla.org/en-US/docs/Web/API/Navigator/getUserMedia
It's now recommend to use https://developer.mozilla.org/en-US/docs/Web/API/MediaDevices/getUserMedia

When the user instantiate a new StreamClient he can pass a null value in the attribute localStream of the streamConfiguration parameters of the constructor:

const StreamConfiguration = {
    localStream: nullSrcVideoElement.srcObject, // Where nullSrcVideoElement.srcObject = null
    isSendOnly: false
};
streamClient = new openteraWebrtcWebClient.StreamClient(SignalingServerConfiguration, StreamConfiguration, RtcConfiguration, logger);

If the user expect that he had pass a valid value to localStream it can be really hard for him to figure it out what's wrong in his code without any warning.

This would allow full featured Python examples (with and without ROS).

Github action cpp-python-tests.yml should be modified to test Debug version.

https://opensource.apple.com/source/WebCore/WebCore-7609.1.20.0.9/platform/mediastream/libwebrtc/

• CMake option to use GStreamer or not
• Add a parameter to force the hardware codec in the SignalingClient constructor (optional)
• The encoder factory should try to use the hardware accelerated encoders, then fallback to the default webrtc-native ones. The GStreamer encoder class can use encodebin. A GStreamer encoder can be prioritized by changing the rank.
  • nvv4l2h264enc, nvv4l2vp9enc, nvv4l2av1enc for Jetson platforms
  • vaapih264enc, vaapivp8enc, vaapivp9enc for Intel CPUs and some AMD Radeon
  • v4l2h264enc for Raspberry Pi platforms
• The decoder factory should try to use the hardware accelerated decoders, then fallback to the default webrtc-native ones. The GStreamer decoder class can use decodebin. A GStreamer decoder can be prioritized by changing the rank.
  • nvv4l2decoder for Jetson platforms
  • vaapih264dec, vaapivp8dec and vaapivp9dec for Intel CPUs and some AMD Radeon
  • v4l2h264dec for Raspberry Pi platforms
• Add the GStreamer support in the root README.md

Should install all required headers and libraries for external compilation.

Translations should be part of the project.

Audio is required for bi-directional webrtc communications in some use cases. This would allow to use ODAS separated audio directly as an audio input stream. Other applications would include speech-to-text features where an audio stream gets processed and returns corresponding text in the data channel.

When enabling software GStreamer video codecs and forcing VP9, the video streams do not works.

How to reproduce

1. Modify the CppStreamClient example to use software GStreamer video codecs and force VP9.
2. Execute the WebStreamClient example.
3. Execute the CppStreamClient example.
4. Perform a call.

Add description and documentation to the readme file.

• Python tests on version 3.6, 3.8 do not work.
• FunctionTask::callSync seems to block. We should investigate the underling webrtc thread. Lambda functions are never called.

Installing opentera-webrtc-native from https://pypi.org/ would be great and save time for new projects using the library. I think we should explore how to automate distribution from GitHub.

The library needs to be statically linked so we have no dependency on the system's OpenCV library.

Allow to build on Windows

Hey,
I am trying to compile the code on Ubuntu 20.04, I have cloned the code, pulled the submodules then followed the instructions, during the make command I get the following error:

I assume the build should be self contained, so what I am missing ?

Have a look at this :

https://github.com/crow-misia/libwebrtc-bin/releases

This might simplify compilation. CMakeFiles should download and extract the binary packages instead of building libwebrtc? This would facilitate development/deployment on ARM/ARM64 targets.

Fix segfaults in tests on MacOS.

Create a Python package that allows easy import of the opentera_webrtc_native_client library without needing the .so file in the working directory of the running script.

• https://stackoverflow.com/questions/9507184/can-openssl-on-windows-use-the-system-certificate-store
• https://stackoverflow.com/questions/42432473/programmatically-read-root-ca-certificates-in-ios