On 26/11/2018 13:39, Andreas Rheinhardt wrote:
> This will enable us to change e.g. the parameter sets of H.2645 in ways
> that would change the parsing process of future units. An example of
> this is the h264_redundant_pps bsf.

I thought quite a bit more about how the alloc/free/clone should fit together 
here.  Following is a patch series implementing it using tables defining how 
each unit works, which I think ends up being nicer than the previous setup with 
ad-hoc functions for each case.  The existing ad-hoc functions for the hardest 
cases (that is, SEI) stay, but everything else in H.26[45] can be handled 
without any new functions at all.

It ends up implementing pretty the same make_unit_writable() function that you 
proposed, so the last patch is your one to h264_redundant_pps.

- Mark
ffmpeg-devel mailing list

Reply via email to