[Libav-user] porting FFusion codec to current FFmpeg libs - ParseContext1 ??

["René J.V. Bertin"]

On Jan 25, 2013, at 17:53, Carl Eugen Hoyos wrote:

> René J.V. Bertin <rjvbertin at ...> writes:
> 
>> Was my request that unclear???
> 
> I am not a native speaker...
> 
> I suggest you study the git history to find out how 
> function usage was replaced inside of FFmpeg / lavc 
> to find out which changes are necessary.
> 

Thing is that I'm neither a "native" FFmpeg contributor nor a "native" video decoding specialist. I gave myself the challenge to do the porting I described thinking it couldn't be that hard to adapt the old FFusion code to the current lavc API, and then get it to build on win32 using prebuilt shared libraries downloaded from the repository.
Without intimate knowledge of FFmpeg's internal 'cuisine' nor  copious comments describing what the code is doing (or rather, why) it's neigh impossible to follow your suggestion without making it a full time job.

As to building lavc52 for win32 - I got it to build, using the mingw32-gcc-4.3.1 crosscompiler from the Cocotron project on Mac OS X 10.6.8. I even got its ffmpeg to run on my Win7 machine, but something goes wrong building the shared library version, even with a very simple test programme that only calls avcodec_version, avcore_version and avutil_version. Built under cygwin (linked with fully optimised ffmpeg ddls), the test runs fine in debug and optimised mode. But when I use MSVC 2010 Express, the optimised version crashes when trying to access avutil_version (the others are fine?!). It's as if the function is mapped to an invalid address.
BTW, something similar but worse happens with the current-version ffmpeg dlls built with the same cross compiler so it does NOT have to do with the older code or Perian's patches to it.

I've been looking at what ffplay does, thinking it could be a more viable approach ... except that ffplay (also) does not explain what it's doing.
I'm going to post a different question on this in hope I'll get some more feedback.

R.