[FFmpeg-devel] [PATCH] avformat/bintext: Implement bin_probe()

[Carl Eugen Hoyos]

2018-02-01 2:19 GMT+01:00 Michael Niedermayer <michael at niedermayer.cc>:
> On Wed, Jan 31, 2018 at 05:43:24PM +0100, Carl Eugen Hoyos wrote:
>> 2018-01-31 15:22 GMT+01:00 Michael Niedermayer <michael at niedermayer.cc>:
>> > Fixes misdetection of sbQ9.bin
>> >
>> > Signed-off-by: Michael Niedermayer <michael at niedermayer.cc>
>> > ---
>> >  libavformat/bintext.c | 49 ++++++++++++++++++++++++++++++++++++++++++++++++-
>> >  1 file changed, 48 insertions(+), 1 deletion(-)
>> >
>> > diff --git a/libavformat/bintext.c b/libavformat/bintext.c
>> > index 12e3bfde4d..722a40c3e5 100644
>> > --- a/libavformat/bintext.c
>> > +++ b/libavformat/bintext.c
>> > @@ -126,6 +126,53 @@ static void predict_width(AVCodecParameters *par, uint64_t fsize, int got_width)
>> >          par->width = fsize > 4000 ? (160<<3) : (80<<3);
>> >  }
>> >
>> > +static int bin_probe(AVProbeData *p)
>> > +{
>> > +    const uint8_t *d = p->buf;
>> > +    int magic = 0, sauce = 0;
>> > +    int invisible = 0;
>> > +    int i;
>> > +
>> > +    if (p->buf_size > 256)
>> > +        magic = !memcmp(d + p->buf_size - 256, next_magic, sizeof(next_magic));
>> > +    if (p->buf_size > 128)
>> > +        sauce = !memcmp(d + p->buf_size - 128, "SAUCE00", 7);
>> > +
>> > +    if (magic)
>> > +        return AVPROBE_SCORE_MAX / 2;
>>
>> This seems too low to me or am I wrong?
>>
>> > +
>> > +    if (av_match_ext(p->filename, "bin")) {
>> > +        AVCodecParameters par;
>> > +        int got_width = 0;
>> > +        par.width = par.height = 0;
>>
>> > +        if (sauce)
>> > +            return AVPROBE_SCORE_MAX / 2;
>>
>> Same here.
>
> was thinking the same but that was what the filename based detection
> used before.
> I kind of just left it where it was, the main goal of the patch was
> just to fix the misdetection not to raise the score.
> But we can surely raise these.
> WHat would you think is a good value ?

I suggest to make it EXTENSION + 1 to have a still low score
but better reflect the high probability.

No more comments, Carl Eugen