Francesco, Andrew:Attached is a patch to add filter_nextsub. It creates two new files, extsub/filter_nextsub.c and extsub/nextsub.h. It highjacked the filter_extsub2 Makefile entries for building. This is a re-work of the existing filter_extsub code, cleaning it up, and (my first attempt at) getting it into NMS format, (and somewhat style compliant, still missing function headers). The new functionality is described (fairly adequately I think) in the help string.
I made an "astyle -l" pass on the code before generating the patch. (I would use "astyle -Bb" when I need to make further changes).
I have only tested the code on a couple of short clips. I'm about to run some full length films with it and will let you know. I would have distributed this earlier, but I had a problem running it on amd64. I'm convinced now that the problem is not mine. It is dying somewhere in tcv_convert() at an sse2 optimization. I'm screwed because I can't build 32-bit variants for all the support libraries required by transcode. So, no 32-bit transcode, and no subtitling for me on amd64.
Still more threading problems as well. Running an identical process can produce different results, given exactly the same input data! It must be some kind of race condition. I am a gentoo nptl only system (glibc 2.4), and have no downgrade option available to glibc 2.3 (old linuxthreads). So, I'm just going to have to wait for your threading changes. Any time estimate on that? Within 6 months? Allan
ps: Now that I have filter_nextsub out of the way, (I can finally subtitle correctly), I'm resuming my efforts on filter_yait.
pps: I'm also looking into another interesting problem, an inverse 5:6 pulldown telecine filter. Yes, they mangle PAL dvd's to NTSC for distribution here in North America. Bastards...
filter_nextsub.patch.gz
Description: application/gzip
