[FFmpeg-devel] [RFC] using ByteIOContext with buffer without reading function

[Baptiste Coudurier]

Hi Michael,

On Fri, Mar 20, 2009 at 04:36:04AM +0100, Michael Niedermayer wrote:
> On Sun, Mar 15, 2009 at 03:53:07PM -0700, Baptiste Coudurier wrote:
> > Hi,
> > 
> > I notice something strange when using ByteIOContext with a supplied
> > buffer without read function, reading only.
> > 
> > aviobuf.c
> > init:
> >     if(!read_packet && !write_flag){
> >         s->pos = buffer_size;
> >         s->buf_end = s->buffer + buffer_size;
> >     }
> > 
> > seek:
> >     if (!s->must_flush &&
> >         offset1 >= 0 && offset1 < (s->buf_end - s->buffer)) {
> >         /* can do the seek inside the buffer */
> >         s->buf_ptr = s->buffer + offset1;
> > 
> > mov.c:
> >     if(init_put_byte(&ctx, moov_data, moov_len, 0, NULL, NULL, NULL,
> > NULL) != 0)
> >         goto free_and_return;
> > 
> > Problem is:
> > when you use url_fskip on the ByteIOContext, for example to skip
> > everything not read (size is 51, url_fskip(51)), you will end up seeking
> > at s->buf_end.
> > 
> > However the condition in "seek" won't allow this since offset1 ==
> > s->buf_end - s->buffer.
> > 
> > Would it be better to change init to set buf_end to s->buffer +
> > buffer_size + 1 or check for <= in seek ?
> 
> i would guess to <= in seek but this should be carefully tested, iam
> not sure if this might have some unexpeted sideeffects
> 
 
Any confidence in the <= case ?

I've checked the buf_end tests in aviobuf.c and they seem ok.

It would be nice to move on on the multiple stsd feature. This should be a
good step in exporting the struct to decoders needing it.

-- 
Baptiste COUDURIER                              GnuPG Key Id: 0x5C1ABAAA    
Key fingerprint                 8D77134D20CC9220201FC5DB0AC9325C5C1ABAAA
checking for life_signs in -lkenny... no
FFmpeg maintainer                                  http://www.ffmpeg.org