[FFmpeg-devel] [BROKEN] apad causes infinite hang

[FFmpeg-devel] [BROKEN] apad causes infinite hang

[Paul B Mahol]

Just try with:

ffmpeg -f lavfi -i sine=d=30 -af apad -f null -

Pressing 'q' will not stop it at all, because current ffmpeg code will try
to flush all frames, but because pad filter never receives EOF from next
filter in chain (sink) it will happily produce frame forever.

Tried to fix ffmpeg.c related code but quickly realized rewrite just made
it 10 times worse to debug this.

Most clean solution is adding av_buffersink_close()
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel

To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".

Re: [FFmpeg-devel] [BROKEN] apad causes infinite hang

[Anton Khirnov]

Quoting Paul B Mahol (2024-06-14 14:19:13)
> Just try with:
> 
> ffmpeg -f lavfi -i sine=d=30 -af apad -f null -
> 
> Pressing 'q' will not stop it at all, because current ffmpeg code will try
> to flush all frames, but because pad filter never receives EOF from next
> filter in chain (sink) it will happily produce frame forever.
> 
> Tried to fix ffmpeg.c related code but quickly realized rewrite just made
> it 10 times worse to debug this.
> 
> Most clean solution is adding av_buffersink_close()

I think it would be cleaner to have an API for closing a _source_ (or
any filter that can produce unbounded amounts of output with no new
input).

-- 
Anton Khirnov
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel

To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".

Re: [FFmpeg-devel] [BROKEN] apad causes infinite hang

[Paul B Mahol]

On Mon, Jun 17, 2024 at 9:24 AM Anton Khirnov <anton@khirnov.net> wrote:

> Quoting Paul B Mahol (2024-06-14 14:19:13)
> > Just try with:
> >
> > ffmpeg -f lavfi -i sine=d=30 -af apad -f null -
> >
> > Pressing 'q' will not stop it at all, because current ffmpeg code will
> try
> > to flush all frames, but because pad filter never receives EOF from next
> > filter in chain (sink) it will happily produce frame forever.
> >
> > Tried to fix ffmpeg.c related code but quickly realized rewrite just made
> > it 10 times worse to debug this.
> >
> > Most clean solution is adding av_buffersink_close()
>
> I think it would be cleaner to have an API for closing a _source_ (or
> any filter that can produce unbounded amounts of output with no new
> input).
>

What do you mean by closing a _source_ ?

av_buffersrc_close() already exist.

This bug is fixed in Librempeg.

Adding a call to close random filter is certainly possible, but such
approach is very
frowned upon in multi-threaded environments.

And once you close all buffersinks the EOF will and must propagate backward
to all not-closed filters and their in/out pads.

If there is better internal API for filters, feel free to present it to
wider audience.


>
> --
> Anton Khirnov
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel@ffmpeg.org
> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>
> To unsubscribe, visit link above, or email
> ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
>
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel

To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".

Re: [FFmpeg-devel] [BROKEN] apad causes infinite hang

[Nicolas George]

Paul B Mahol (12024-06-17):
> And once you close all buffersinks the EOF will and must propagate backward
> to all not-closed filters and their in/out pads.

Your analysis here is right, obviously. Closing a source for this bug
makes no sense.

Regards,

-- 
  Nicolas George
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel

To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".