MJPEG Tools Svn

Since I have an Opteron-250 system now it's possible to put the correct
option in the cpuinfo.sh script. Using "-mcpu=opteron" does NOT work
(gcc manpage not withstanding):

gcc -DHAVE_CONFIG_H -I. -I. -I../.. -I../.. -I../../utils -march=opteron -mcpu=opteron -g -O2 -pthread -Wall -Wunused -MT build_sub22_mests.lo -MD -MP -MF .deps/build_sub22_mests.Tpo -c build_sub22_mests.c -fPIC -DPIC -o .libs/build_sub22_mests.o
cc1: error: bad value (opteron) for -march= switch
cc1: error: bad value (opteron) for -mcpu= switch
gmake[3]: *** [build_sub22_mests.lo] Error 1

Using "-mcpu=x86-64" DOES work for the mode=5 case.

use a temporary buffer when reading/writing fields

Since the code that uses the mmx intrinsics is #if 0 do the same for the
xmmintrin.h file (which does NOT exist with gcc/g++ 2.95.3).

split out CPPFLAGS to only those programs that need the particular flags
(on my machine, LIBDV_CFLAGS needed gtk+ 1.2 while glav needed gtk+ 2.4,
causing quite a problem)

fix double reference to streamstate.*

Basic two-pass structure for encoding now in place with dummy pass-2
Tidied up source structure in some of the uglier corners

Allow compilation without SDL (means frame_planar_to_packed in unavailable)

strip_fPIC.sh no longer exists

Added that patch:

Hallo,
there is a severe allocation bug in pipelist.c

The line
if ((pl->segment_count % 32) == 0)
pl->segments = (PipeSegment **) realloc (pl->segments,
sizeof (pl->segments) + 32 * sizeof (PipeSegment *));
}

assumes pl->segments to return the size of malloced memory pointed to by
pl->segments;

myght be
if ((pl->segment_count % 32) == 0)
pl->segments = (PipeSegment **) realloc (pl->segments,
(pl->segment_count + 32) * sizeof (PipeSegment *));
}

use temp var for indexing