On Sat, Jun 04, 2011 at 11:30:16AM -0700, Ronald S. Bultje wrote:
> Hi,
>
> On Sat, Jun 4, 2011 at 10:20 AM, Kostya Shishkov
> <[email protected]> wrote:
> > ---
> > libavformat/rmdec.c | 9 ++++++---
> > tests/ref/fate/rv30 | 24 ++++++++++++------------
> > 2 files changed, 18 insertions(+), 15 deletions(-)
>
> 6-14 shows improvement, 26-33 becomes worse. May be vsync related. We
> should make sure that improvements here show up for end users
> otherwise they think we just break stuff.
Well, looks like they have wrong sort of timestamps. While it reports
plausible values, they look more like DTS to me.
Here's debug output ("V timestamp" is what demuxer reports, frame timestamp is
timestamp extracted from frame header, preceeds demuxer timestamp here) from
that RV30 sample:
Frame timestamp 10 seq 0
V timestamp = 1
Frame timestamp 850 seq 1
V timestamp = 85 (was NOPTS)
Frame timestamp 210 seq 2
V timestamp = 86 (was NOPTS)
Frame timestamp 420 seq 3
V timestamp = 87 (was NOPTS)
Frame timestamp 640 seq 4
V timestamp = 88 (was NOPTS)
Frame timestamp 10A0 seq 5
V timestamp = 10A
Frame timestamp A60 seq 6
V timestamp = 10B (was NOPTS)
Frame timestamp C80 seq 7
V timestamp = 10C (was NOPTS)
Frame timestamp E90 seq 8
V timestamp = 10D (was NOPTS)
Frame timestamp 1900 seq 9
V timestamp = 190 (was NOPTS)
Frame timestamp 12C0 seq 10
V timestamp = 191 (was NOPTS)
Frame timestamp 14D0 seq 11
V timestamp = 192 (was NOPTS)
Frame timestamp 16F0 seq 12
V timestamp = 193
Frame timestamp 150 seq 13
V timestamp = 215 (was NOPTS)
Frame timestamp 1B10 seq 14
V timestamp = 216 (was NOPTS)
Frame timestamp 1D30 seq 15
V timestamp = 217 (was NOPTS)
Frame timestamp 1F40 seq 16
V timestamp = 218 (was NOPTS)
Frame timestamp 9B0 seq 17
V timestamp = 29B (was NOPTS)
Frame timestamp 370 seq 18
V timestamp = 29C (was NOPTS)
_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel
