MJPEG Tools Svn

Fixed bug in flushing pass-1 coded Frames at end of sequence
Add hook for GOP sizing...

Added a patch from the tracker submitted by:
Dean Gaudet
[ 1117718 ] jpeg2yuv: eliminate division in rescale_color_vals

rescale_color_vals includes a division and a multiplication, but it can be refactored so the division can be done at compile time. (compilers are generally not permitted to make this change, but most have compile options which loosen the fp standards compliance which would catch this case.)

in a pipeline with mpeg2enc this yielded a 9% speed improvement.

The "better solution" has arrived ;)
If a system has neither stdint.h or inttypes.h then rather than bail out with
an error about getting those files (which may not be possible for oddball
platforms) simply define the types that are needed to build mjpegtools.

change from the mixture of __inline__ and inline to simply inline. Seems
IRIX doesn't have __inline__ but does have inline. See
http://sourceforge.net/tracker/index.php?func=detail&aid=1116887&group_id=5776
&atid=105776

fix handling of single frame mpegs

fix segfault -- frames were being left in pass1

New lavaddwav from Nicholas Redgrave (baron at bologrew dot demon dot co dot uk)
and Bernhard Praschinger.
A change was made to use the new common reorder_32() routine.
The common reorder_32() function was changed to use a 'big_endian' 2nd argument
and uses of reorder_32() in lav2wav and lavtrans were changed to pass the
flag (intead of doing if/then/else on big_endian).

Create a single instance of reorder_32 rather than replicating it in multiple
places. Also, instead of "unsigned long" it is necessary to use "uint32_t"
when dealing with WAV headers. This is because 'unsigned long' is 8 bytes
(not 4) on X86-64 systems (AMD64, Opteron, etc).

support chroma C422.

fix wrong MJPG_chroma in report.