It fixes at least RGB  blending with opengl on little indian arch.

 Without this patch, all epehemere subpicture with the same start
date are blended over the picture.

dialog_ProgressCreate can return NULL. For this case,
only deal with the result if it is not NULL.
Signed-off-by: Rémi Duraffort <ivoire@videolan.org>

should not crash if font family not found. Thanks to Anthony Loiseau for
spotting this one.

Also set vbv-bufsize if user didn't specify them.
bufsize is bitrate * seconds between keyframes instead of earlier
2*bitrate. Recommended by Dark_Shikari

So now you can stream for mobiles with following:
 sout="#transcode{deinterlace,acodec=mp4a,vcodec=h264,width=320,height=180,venc=x264{no-cabac,bframes=0,ref=3,level=12,keyint=50},vb=180,ab=64,channels=2}...

tested with rtp-output and G1 android. Produces quite stable near cbr
stream.

And file-encoding should be ok with defaults (without any
venc=x264-settings or bitrates), if not targeted to
handhelds/streaming.

change from default 800 to zero, as nowdays most of bitrate transcoding
is for streaming, and 800 ain't good default for that either. Also
thisways it's easier on check different encoding modes (cbr/qp/etc) on
video encoder modules.

I did quick look on different modules, but didn't spot anything obvious
that would break. I'll look ffmpeg and x264 more closely next.

It avoids loosing subtitles with the messages
"main subpicture error: subpicture heap full".

It also avoid useless mdate() when decoding audio.

Dynamic subtitles are used by ssa and kate.
This commit breaks the API for 2 functions used by the transcode module.

Work around the missing PathIsNetworkPathW() function...

done when the module is created.

Clean up syslog handling in logger: Use only one list of facilities, and get the default facility from that list.

With fontconfig freetype-font is font-family to use, not fontfile.
This should solve issue on linux-side where there aren't any one
place and one file thats usually there.

I'm not that familiar with freetype/fontconfig stuff, so this one could
be way wrong. So feel free to mention/fix/revert this if you spot
something odd.

moment) to toggle a boolean with the variable lock taken (which is not the case
when doing a var_SetBool(!var_GetBool)).
This function is also two times faster (only one lookup).

Pointed-out-by: tixxDZ <tixxdz@gmail.com> - DZCORE Labs, Algeria