Add a framebuffer implementation based on clutter instead of "raw"
OpenGL. The performance is slightly worse, but it might bring some
platform abstraction and other benefits for the future.
Current libavcodec does no longer always return S16 sample format, but
the conversion needs to be done explicitly. Use libswscale for that.
Not tested on raspbian, only on yocto/ffmpeg.
* remove ilclient
* use a decoder implementation similar to pidvbip instead,
see https://github.com/linuxstb/pidvbip
* video is decoded in hardware (videocore)
* audio is decoded by libavcodec
* add a AVDec wrapper class for both audio and video decoders
* very raw, needs more polishing. But decodes audio and video :-)
* only tested with h264
ffmpeg audiodecoders after version 1.1 deliver audio samples in planar
formats by default instead of packed as before. Additionally, the AC3
decoder delivers now in planar float.
Use libswresample for sample format conversion, this will work with old
(where it hopefully does nothing) and new ffmpeg versions.
Later on, also sample rate and channel layout conversion could be
implemented if desired.
Tested with ffmpeg versions 1.0.6 and 1.2.1.
* build intermediate libraries for each subdirectory
* link those libs in main directory instead of single objects
* ugly hack in configure.ac to disable dynamic lib for now
This is mostly a dummy implementation except for the dmx class which
should be working. It is intended for testing on PCs with budget
DVB cards which don't have a decoder anyway.
TODO: some code is very similar to SPARK (record and pwrmngr
are just symlinked, dmx is almost identical). Reduce duplication
by factoring out DVBAPI code into an extra directory.
* what works: audio, video, demux
* what probably doesn't work or is untested: record
* what very likely doesn't work: playback
Playback is just copied over from aztrino and made to compile.
If you are lucky, it just segfaults :-)
