[FFmpeg-devel] [PATCH 2/2] avformat/img2dec: add GEM Raster image demuxer

[Peter Ross]

On Sun, Jan 23, 2022 at 01:01:58AM +0100, Carl Eugen Hoyos wrote:
> Am Fr., 17. Sept. 2021 um 10:39 Uhr schrieb Peter Ross <pross at xvid.org>:
> 
> > +static int gem_probe(const AVProbeData *p)
> > +{
> > +    const uint8_t *b = p->buf;
> > +    int ret = 0;
> > +    if ( AV_RB16(b     ) >= 1 && AV_RB16(b    ) <= 3  &&
> > +         AV_RB16(b +  2) >= 8 && AV_RB16(b + 2) <= 779 &&
> > +        (AV_RB16(b +  4) > 0  || AV_RB16(b + 4) <= 8) &&
> > +        (AV_RB16(b +  6) > 0  || AV_RB16(b + 6) <= 8) &&
> > +         AV_RB16(b +  8) &&
> > +         AV_RB16(b + 10) &&
> > +         AV_RB16(b + 12) &&
> > +         AV_RB16(b + 14)) {
> > +        ret = AVPROBE_SCORE_EXTENSION / 4;
> 
> Without running probetest, this score looks too high to me.
> (1, 2 or 3 followed by anything not 0 with some restrictions?)

The probetest warning is only triggered when the score is greater than
AVPROBE_SCORE_EXTENSION / 4

This score is consistent with other image2 demuxers like bmp and pcx.

How much lower does it need to be?

> Why is the above sufficient? Are there files in the wild
> without the fourcc's below?

Correct, not all files have the fourcc's below.

> > +        if (AV_RN32(b + 16) == AV_RN32("STTT") ||
> > +            AV_RN32(b + 16) == AV_RN32("TIMG") ||
> > +            AV_RN32(b + 16) == AV_RN32("XIMG"))
> > +            ret += 1;
> 
> Should be EXTENSION + 1

Agree.

-- Peter
(A907 E02F A6E5 0CD2 34CD 20D2 6760 79C5 AC40 DD6B)
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 195 bytes
Desc: not available
URL: <https://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20220123/0629f10f/attachment.sig>