Quoting Andreas Cadhalpun (2015-09-27 22:30:58) > On 27.09.2015 21:46, wm4 wrote: > > On Sun, 27 Sep 2015 21:11:37 +0200 > > Anton Khirnov <[email protected]> wrote: > >> Quoting Vittorio Giovara (2015-09-27 02:41:43) > >>> This field was used in combination with coded_frame to compute some > >>> sort of PSNR value. Regardless of its use, this property is tied to the > >>> encoding processing, while AVFrame is a structure suited for decoded > >>> data only. > >>> > >>> The only use of this field (besides the deprecated coded_frame) is > >>> in mpegvideoenc: since it might be used in bframes estimation, add an > >>> additional field to Picture, which is a private structure anyway. > >> > >> While I'm all for removing crap, this commit message is very misleading. > >> The fact is that error is used for _exporting_ some information to the > >> user. This is done through coded_frame, which someone deprecated, but > >> did not provide a replacement for this field. And now the commit message > >> is trying to somehow pretend that AVFrame.error[] has always been an > >> unperson and we were always at war with eastasia. > >> > >> That is not good practice IMO. If you're deprecating crap that nobody > >> should use, just say so, don't cover it under a wall of euphemisms. That > >> way people people can actually see what's going on here and > >> object/suggest alternative ways or whatever. > >> > > > > Now the prize question: does anyone use it? > > It is used to print PSNR information at least in elmerfem [1], kino [2] > and transcode [3]. > > Best regards, > Andreas > > 1: > https://sources.debian.net/src/elmerfem/6.1.0.svn.5396.dfsg2-4/post/src/plugins/savempg.c/?hl=93#L93 > 2: https://sources.debian.net/src/kino/1.3.4-2.1/ffmpeg/ffmpeg.c/?hl=904#L904 > 3: > https://sources.debian.net/src/transcode/3:1.1.7-9/export/export_ffmpeg.c/?hl=1756#L1756
>From what I can see, it's used just to write the numbers to some log, they don't actually do anything with those numbers. So I'd say it's about as useless as what avconv currently does. -- Anton Khirnov _______________________________________________ libav-devel mailing list [email protected] https://lists.libav.org/mailman/listinfo/libav-devel
