MJPEG Tools Svn

Security fix from IBM (found on http://oss.software.ibm.com/linux/patches/misc/zoran-memcpy.diff (I'd love if people would send these to me, too)

Two of my changes that apparently weren't needed, it's really only the VStart that causes all this... Weird stuff going on in that chip, I'll want to have a look at how the "other" driver does this...

Some funny twinkling things in here.

* dc30+ works (woohoo! yay! hold your breath a while longer, now comes the bad things)
* only at decimation 1
* not tested with anything except default options
* takes a long time (+/- 2 s.) before lavrec 'loads' until lavrec 'starts'
* still times out ocassionally
* the first +/- 64 lines of the JPEG picture are noise
* however, if I increase VStart (zr36016), it times out again
* so in short, more hacking is needed here.

However, we get data+signal, so the worst part should be over now. :).

Some updates for the Optimizing the stream section (checked by Steven M. Schultz)
And the SMP encoding section , how can we keep both (ore more) CPUs busy.

* Add changelog entry for a change some time ago
* Fix sync() handling in v4l2 (we didn't increase a counter)
* Formatting fixes in zr360{16,50}.c
* Add .cvsignore file with 2.6.x kernel object files in them

default of 3800 in usage() statement was incorrect, the defautl is actually 7500

Forgot to update the manpage

correct typos in Usage()

Rewrite quantization table selection logic, add TMPGEnc, kvcd matrices as well as the ability to load user defined matrices.

yuvscaler implementation of bicubic algorithm should now be correct.
In particular: downscaling using BICUBIC has better anti-aliasing.
+ various minor corrections