MJPEG Tools Svn

# automake 1.9 could use AM_PROG_STDC but 1.10 became picky and needs AM_CC_C_0

Change the date string to something more current (reflecting that a change has
been made recently). Also suppress the version/usage info message if the
verbosity level has been explicitly set to 0.

workaround a crash on OSX 10.5.1 MacIntel systems. Seems the intermediate
buffer was one line too short and while that "works" on many systems (linux
intel, OSX PPC) it is luck due to malloc() implementations and memory layout.

Until someone who understands the program has some time the workaround is to
allocate a larger intermediate buffer.

A updated version of the howto.
This should do the trick for the next release.

That patch seems to fix this Problem:
#10 0x4cc3fffc in png_read_destroy () from /usr/lib/libpng12.so.0
#11 0x4cc4043e in png_destroy_read_struct () from /usr/lib/libpng12.so.0
#12 0x0804939f in decode_png (pngname=0xbffff830 "menu-bg.png", process=0,
param=0xbffff930) at png2yuv.c:405
#13 0x08049671 in main (argc=13, argv=0x8050180) at png2yuv.c:425
---END---

The problem was reported by: Reinald Gfüllner.

The change changes the way the pointer is freed like it is done in the examples. from the libpng

Need to add back, but conditionally, ${X_LIBS} -lX11. Can't use HAVE_X
because OSX 10.5 has X but adding extra -lX11 options causes problems. Use
HAVE_V4L in hopes that only linux systems will be affected.

Two things:
1) HAVE_X is not needed. lavplay builds and runs on both Linux and OSX
(10.4 and 10.5) fine without it.
2) HAVE_X causes a build failure on OSX 10.5:
ld: cycle in dylib re-exports with /usr/X11/lib/libGL.dylib

Removing HAVE_X from configure.ac and lavtools/Makefile.am resolves the error

Fixed: non-clipping of base quantisation in the look-ahead on-the-fly
rate controller.
Fixed: use of MB lum_variance rather than best_me->var for activity
calculation for Pass-1 on-the-fly rate controller causing wildly
inconsistent quantisation values for material with a lot of chroma
activity.

Added the Patch from the Tracker for Yuri :
http://sourceforge.net/tracker/index.php?func=detail&aid=1691505&group_id=5776&atid=305776

Allow more control to advanced MPEG encoding application
Summary: exposed inner steps inside SeqEncoder::EncodeStream() to user
allowing finer control.

SeqEncoder::EncodeStream() currently parses image stream using 'pull'
method, calling reader from inside the long loop.

target_bits can be int instead of int32_t. Seems cygwin's implementation of
min() is broken somehow.