MJPEG Tools Svn

Initial Fix attempt for broken handling of forcing P frames.

Ok - after discussion with James revert some of the argument sanity checking
by putting it back inside the ALTIVEC_VERIFY ifdef. Leave the one check
for 16x16/8x8 blocks present to avoid a repeat of the yuvdenoise debug session.

at Stefan's request (he is on travel for a while) administer first aid to
the program by changing it to use 16x16 search blocks rather than 32x32 blocks.
The accelerated SAD routines have a 16x16 limit. Altivec routines check for
this and will log a message and exit, but the MMX versions will simply crash
eventually.

Enable some of the sanity checks that were hidden inside '#ifdef ALTIVEC_VERIFY'
conditionals. In particular the requirement in sad_00 that 'h' be 8 or 16 was
being violated by yuvdenoise and the program was mysteriously coredumping.

IF "user input" (in this case a parameter from a user program into a library)
can cause a program to segfault THEN that condition SHOULD BE CHECKED and
a meaninful error given.

After enabling the checks the cause of the problem was immediately known:

**ERROR: [yuvdenoise] sad_00: h != [8|16], (32)

Update documentation - lingering references to jpeg-mmx

Aesthetic changes, use 'err' as a local variable rather than 'errno' (system
error), use filen() instead of hardcoding filedescriptor numbers

Add the y4mivtc executable.

getopt() returns 'int' not 'char'. Cleanup the help/usage messages so they
don't double or triple space.

Fix the usage() summary and coincident comments. The proper usage is -v -v
and not -vs to increase the debug/verbosity level.

Move imageplanes.cc from mpeg2enc_source to libmpeg2encpp_source so that
symbols do not become undefined in the library.