ср, 23 окт. 2024 г., 18:23 Phyllis Smith <[email protected]>:
may be we better to skip 7.1 completely?
This is disturbing. Usually if something is not working in one version, it is still carried to the next.
I crossposted my frustration with current situation (when ffmpeg team one-sidely breaks not just us but everyone downstream) earlier this day. Of course team of professional developers will change their part (breaking things) much faster and effortlessly than I ever can hope to repair them. Because ffmpeg apparently set to 2-3 release per year, with unknown amount of breakage each time ...I am not really eager to fix something twice. yuv range fix still not agreed upon, as far as I can, may be it will end up as rework and not just fix. I amnot sure if 7.1 makes much sense for us in itself? Of course breakage that come in 7.2/8.0 might be so big that I'll prefer to fix 7.1 instead. My main worry that even outside ff-scale filter and profile vs vprofile change something broke with some vaapi/qsv profiles, and it might be harder to track down. Right now 7.1 *builds* fine if you add renamed patch10 at least, but on runtime ..... we have all those errors Terje reported.
... Snip, Snip, Snip!
Doesn't this correspond to what we previously also discussed for ffmpeg
7.0.2?
ffmpeg as cli does its own option processing, as far as I understand.
Using av_dict_* functions in c++ code was working with libav*/ffmpeg libs up to 7.0 but evidently not after..
I'll look into what ffmpeg cli does now, but if more changes in pipeline may be we better to skip 7.1 completely?