vlc_threadobj() is no use otherwise. People may want not to enable
reference checking all the time also...

with an ugly green in packed YUV).

Ugly hack to make the filter chain work correctly for the video output core. This will be removed once all the different picture release methods have been merged into one coherent method.

a lot of code everywhere a filter_t chain is used ("video filter2":
vout core, transcode, chain.c video filter; "sub filter": spu_Init;
"audio filter2": transcode; the other types of filter_t objects are
never chained)

Btw, why aren't "audio filter2" filters used anywhere in the aout core?

Note that it seems to suck badly, but as I get the same awful results
with RTSP+live555, I have to assume eitehr the whole synchro is broken
or the stream output is

Use tabs in Makefile

Currently, only MP2T payload and UDP transport work.

Do not align height and width on multiples of 16 in XVideo output. This fixes #1604. If someone thinks that this fix isn't correct, please revert this commit ... and explain why it was wrong :)

Only link avcodec with libavcodec, avformat with libavformat, swscale with libswscale, imgresample with libavcodec (small hack include to fix AC_CHECK_LIB code), postproc with libpostproc.

Some more seperation of the different ffmpeg based modules. They're now completely disjoint except for the common chroma.h and fourcc.h headers.

Img resample based module (functionality untested ... but compilation and loading works). Remove unneeded stuff from swscale.

Add fallback to img_resample based plugin if swscale isn't available on the system. img_resample will be built as a seperate module than avcodec to prevent avcodec from refusing to dlopen if img_resample is unavailable (debian...).