On Tue, 16 Feb 2016, Carl Eugen Hoyos wrote:
Marton Balint <cus <at> passwd.hu> writes:
+ /* The purpose of this is demuxer is to detect DVB teletext
streams in
+ * mpegts, so we reject invalid buffer sizes */
+ if ((p->buf_size + 45) % 184 != 0)
+ return 0;
I don't think this is ok (although I may miss
something):
The file could end in the middle of a teletext
patckage or the end of the probe buffer could
be within a teletext package.
Usually you'd be right, but not this time I think because the reason for
this demuxer is not to actually handle RAW dvb teletext files (I doubt
those exists in the real world), but only to detect DVB teletext streams
in mpegts. Therefore it matches for a single PES payload, not a bitstream.
A RAW dvb teletext file would be a dump of PES payloads without separation
where one would constantly have to find synchornization because there is
no way of knowning how big is a PES payload and when the next one starts.
On the other hand DVB teletext PES payloads have a strict length
requirement, and they are at most 1437 bytes long, so the full payload
should always be probed even if the probe function truncates at 2048 bytes
or so (I don't know if it does or not).
+ if (!ff_data_identifier_is_teletext(p->buf[0]))
+ return 0;
+
+ for (buf = p->buf + 1; buf < end; buf += 46) {
+ if (!ff_data_unit_id_is_teletext(buf[0]) && buf[0] != 0xff)
+ return 0;
+ if (buf[1] != 0x2c) // data_unit_length
+ return 0;
+ }
I suspect this should count frames.
Not really, a PES payload is only one frame, the loop match data units
(practically VBI lines, or stuffing).
+ return AVPROBE_SCORE_MAX / 2;
Is this ok for one frame?
I don't know, I figured even for the smallest PES payload which is 139
bytes we are matching 1+2+2+2 mostly different bytes, which should be a
pretty safe detection, therefore AVPROBE_SCORE_MAX / 2. I can use anything
here you prefer.
+++ b/libavutil/internal.h
Please create a new teletext header in libavcodec.
Why not in libavutil then? It is used in a codec in libavcodec and a probe
function in libavformat.
Regards,
Marton
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
