Seemingly fixes an issue which could cause a few pixels of the remote video in a call to be scaled out of proportion.

Works on preventing an issue with the display of remote video which could scale it down to a tiny number of pixels. Fixes multiple technical/programming errors, warnings, javadocs.

Fixes the display of the remote video in a call which could appear as a single pixel (on Windows at least).

Improves the hotplug response times on Windows which may be noticeably prolonged due to thread scheduling in cases of Windows Multimedia reporting lack of a device driver.

Fixes inconsistent synchronization in RTPSourceStream which could lead to unexpected behavior. Enables debug output in the PortAudio binaries on Windows. Makes sure that failures to load the JNI counterparts of CoreAudioDevice will not break PortAudio and the software volume control. Fixes warnings.

Uses device UID or name (instead of index) for PortAudio CaptureDeviceInfo locator. This fixes the problem of starting/stopping the wrong device after an update of the device list (i.e. a device has been dis/connected.).

Commits work on progress. Prepares to keep video frame memory entirely in the native, non-Java heap.

Refactors AVFrame, ByteBuffer and ByteBufferPool. Prepare to invoke FFmpeg's avcodec_free_frame(AVFrame **) which is the new way to free AVFrames.

Fixes an issue of overlapping videos which could cause the local video in a call to be displayed bellow the remote video. Works on remedying issues with telephony conferences utilizing the Jitsi VideoBridge server-side technology.

Shortens the time it takes the PortAudio integration on Windows/WMME to react to the detection that an audio device has been unplugging.

Fixes freezes in Pa_ReadStream, Pa_StopStream and Pa_WriteStream and incorrect causes of paInvalidChannelCount errors in PortAudio on Windows/WMME. Mitigates an issue with hightened CPU usage (on all supported operating systems) related to hotplug.

- Fixes a crash in Pa_StopStream. - Makes the switching of the audio device while in a call more thorough and resilient.

- Removes the DirectSound support from PortAudio in order to prevent issues with bad audio quality.
- Attempts to recognize USB PortAudio devices on Windows as such, ignores the virtual Microsoft Sound Mapper devices to avoid user confusion. Contributed by Vincent Lucas.

Improves sound device selection: automatically switches to new plugged device, remember user preferences in a list.

- Works on removing the video of a video conference participant from display on the remaining non-focus participants. - Tries to fix a crash in PortAudio (often seen at the end of video conference calls).

Fixes an issue which would leak at least one Thread per call which would endlessly send out RTCP packets.

Fixes a failure to keep the aspect ratio of the video when switching to full screen. Attempts to remedy a multi-video display when switching between camera and desktop streaming. Tries to fix a crash in the QTKit video capture on Mac OS X. Works on addressing a unnecessary remaining of the camera on Mac OS X open.

Re-applies the ideas of r9718 with a refined implementation on the PulseAudio Renderer in order to reflect changes to the user's choice with respect to the default playback device while in a call.

Temporarily reverts to the PulseAudio Renderer implementation to a pre-r9718 state to fix freezes reported by Vincent Lucas and Kertesz Laszlo on the dev mailing list while further looking into the cause of the problem.

Commits an initial version of the libjitsi library which should be fully functional with respect to neomedia/MediaService and should be usable with and without OSGi.