[FFmpeg-devel] [PATCH]Add an image3 demuxer with format autodetection

[Carl Eugen Hoyos]

Michael Niedermayer <michaelni <at> gmx.at> writes:

> > > ./ffmpeg -i tests/lena.pnm -f image2 -vcodec png out
> > > cat out out out|./ffmpeg -i -
> > 
> > I agree but isn't this impossible because of ticket #3687?
> 
> for bmp, no pipe variant would work reliable neither image2, 
> image3 nor with or without individual AVInputFormats

Agree.

> in practice that would probably be dealt with by not 
> autodetecting bmppipe, the user still can specify it with 
> -f bmppipe and maybe he is lucky and his images work with 
> the bmp parser

> -f image2/3pipe wouldnt really help here or did you mean 
> something else ?

With my patch, bmps can be auto-detected and -f image2pipe 
can auto-detect the image codec.
I think the code to autodetect pngs as pngpipe but let the 
user decide if he wants bmps autodetected or not will look 
messy (or at least will not be consistent).

But all this doesn't matter as long as somebody implements 
the preferred solution.

Carl Eugen