MJPEG Tools Svn

No problem with gcc3. BUT on OSX 10.4 the warning (with caution it will be
fatal eventually):

bsad.c: In function 'bsad_altivec':
bsad.c:160: warning: target of assignment not really an lvalue; this will be a hard error in the future

on Fedora Core4 Test3 it's a fatal error.
https://sourceforge.net/tracker/?func=detail&atid=105776&aid=1206626&group_id=5776

Fix is to cast the return value to the variable type. Intead of

vu8(fH) = vec_mergeh(zero, vf);

use something like:

fH = vu16(vec_mergeh(zero, vf));

No problem with gcc3. BUT on OSX 10.4 the warning (with caution it will be
fatal eventually):

bsad.c: In function 'bsad_altivec':
bsad.c:160: warning: target of assignment not really an lvalue; this will be a hard error in the future

on Fedora Core4 Test3 it's a fatal error.
https://sourceforge.net/tracker/?func=detail&atid=105776&aid=1206626&group_id=5776

Fix is to cast the return value to the variable type. Intead of

vu8(fH) = vec_mergeh(zero, vf);

use something like:

fH = vu16(vec_mergeh(zero, vf));

No problem with gcc3. BUT on OSX 10.4 the warning (with caution it will be
fatal eventually):

bsad.c: In function 'bsad_altivec':
bsad.c:160: warning: target of assignment not really an lvalue; this will be a hard error in the future

on Fedora Core4 Test3 it's a fatal error.
https://sourceforge.net/tracker/?func=detail&atid=105776&aid=1206626&group_id=5776

Fix is to cast the return value to the variable type. Intead of

vu8(fH) = vec_mergeh(zero, vf);

use something like:

fH = vu16(vec_mergeh(zero, vf));

Not a problem with gcc3. BUT on OSX 10.4 the warning (with caution it will be
fatal eventually):

bsad.c: In function 'bsad_altivec':
bsad.c:160: warning: target of assignment not really an lvalue; this will be a hard error in the future

on Fedora Core4 Test3 it's a fatal error.
https://sourceforge.net/tracker/?func=detail&atid=105776&aid=1206626&group_id=5776

Fix is to cast the return value to the variable type. Intead of

vu8(fH) = vec_mergeh(zero, vf);

use something like:

fH = vu16(vec_mergeh(zero, vf));

check if -maltivec is accepted and if so add it to ALTIVEC_CFLAGS. Later
versions of gcc (gcc4) won't generate vectorized memset/memcpy unless both
-faltivec and -maltivec are given.

Use strcasecmp instead of strcmp to shorten the extension checking.

Tidy up a little during the Release Candidate phase. The chroma-type member
is not just for MJPG (could be 'yuv') so shorten the name from 'MJPG_chroma'
to simply 'chroma'.

Need to use "head -n 1" instead of the deprecated "head -1" form - some
systems moan&groan about the traditional form.

assumptions about chroma subsampling should not be coded in as explicit
"divide by 4" (which assumes 4:2:0). Use the Y4M API to dynamically compute
the chroma plane size. I.E. do not do 'w * h / 4' but instead use
y4m_chroma_ss_x_ratio (and 'y').

Fix RPM build