Le decadi 20 fructidor, an CCXXIV, Marton Balint a écrit : > The work needed fixing this is definitely not comparable to the work I did > creating this patch, or the work anyone will have to do maintaining it. > Asking me to go in this direction, seems an unreasonable amount of extra > work for me, and most probably I will not do it, and rather abandon the > patch, or maintain it privately.
I understand that, but the patch as is goes in the wrong direction. I am sure we can discuss this a bit further and find a solution that goes in the right direction but is still acceptable with regard to the efforts you can propose. > I don't trust the input channel layouts, I want to specify that all my input > channel layouts are unkown. Can/will I do this with -channel_layout 0 or > something like that? I think the tool should allow to override all values from the input, as long as it makes sense, and this applies to the channel layout without doubt. In other words, I strongly support an option to override the channel layout detected from inputs, including to set it to unknown whenever it makes sense. I do not know if "-channel_layout" is the right option for the job. At a quick glance, it seems to do a lot of things. A simpler option that checks the number of channels and overrides the layout may be a better idea; and naming all the overriding options consistently could be a nice extra feature. But... > Sure, I plan to use uncompressed audio on the output side. ... I do not think that allowing unknown layouts too easily is a good idea. Sometimes, it makes sense to have an unknown layout (at least until we have a more expressive system, if it happens); for example, dual-language-disguised-as-stereo as found in some TV broadcasts can not be expressed as a layout. But most of the time, there is a channel layout, and the correct solution is to make sure the tools and library know it as early as possible. I am pretty sure you are able to make the difference. But even like that you may prepare yourself trouble for the future: maybe at some point you will update your code to change uncompressed PCM to Opus, forget you neglected the layout issue, and thus lose inter-channel correlation. Unfortunately, our average users may not have that clear an understanding. I suspect a lot of them, subjected to an issue, would try setting an unknown channel layout as suggested by an anonymous on some barely-related stackexchange question, see it cuts right through their problem and not realize they are creating slightly invalid files. Regards, -- Nicolas George
signature.asc
Description: Digital signature
_______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
