Romain Beauxis (HE12026-04-27): > This might not be in the scope of the changes you're considering at > the moment but I wanted to report a complicated issue we had to debug > recently with filters re-configuration in the `ffmpeg` CLI: > > When using `trim` with a max number of frames, a resolution change > would trigger the filter reconfiguration which would, in turn, reset > the number of frames the filter had already seen, resulting in > incorrect output.
Hi. Indeed, now you point to it the issue is quite obvious. I expect we have similar issues with filters that allow formulas that depend on the frame number or on STARTPTS. It would be fixed by proper subgraph reconfiguration, but as I said, “not for tomorrow”. For resolution changes we could just feed the frames to buffersrc, but only for filters that do not care, and not when expressions that depend on the resolution are evaluated at init time. For format changes, it would not work at all. I do not have an idea on how to fix it right now, but now that I am aware of the issue I can think on it in the background. PS: My email you replied to contains “Reply-To: [email protected]”, please do not override it. Regards, -- Nicolas George _______________________________________________ ffmpeg-devel mailing list -- [email protected] To unsubscribe send an email to [email protected]
