On Tue, May 17, 2016 at 10:00:38PM +0200, Nicolas George wrote: > Le nonidi 29 floréal, an CCXXIV, Ronald S. Bultje a écrit : > > It's not for us (ffmpeg.c developers), it's for people using the ffmpeg > > libraries on other language platforms (e.g. python). > > I think one of the cruxes of the issue is that we are (ab)using the AVOption > system for two different things: introspection and uniform access to > structure members for libraries and language bindings, and user interaction. > > For introspection to work, all public fields of all structures that have > some need to be accessible. Libraries can make a few exceptions, and > therefore we can get away with one "almost" in the previous sentence, but > not two. We are nowhere near it and we will never achieve it. And we > probably would not want it anyway. Plus, the AV_OPT_TYPE system is way too > limited and the lookup is very inefficient. >
> Michael used the AVOption system as a quick way to give access to fields > that were not present in the fork without breaking ABI compatibility, but I > think this was a mistake, it would have been better to just write an > accessor function. Since we have dropped ABI compatibility with the fork, we there where and are accessor functions: git grep MAKE_ACCESSORS [...] -- Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB I do not agree with what you have to say, but I'll defend to the death your right to say it. -- Voltaire
signature.asc
Description: Digital signature
_______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
