[FFmpeg-user] missing a selected frame

[Ulf Zibis]

Am 19.07.19 um 11:22 schrieb Stéphane Chauveau:
> On 7/19/19 9:18 AM, Moritz Barsnick wrote:
>> On Fri, Jul 19, 2019 at 00:21:53 +0200, Ulf Zibis wrote:
>>> ./ffmpeg -y -v warning -i debug/CYD_1.5m_x264.mp4 -vf
>>> select='eq(t\,10.16)+eq(t\,10.2)+eq(t\,10.24)+eq(t\,10.28)+eq(t\,10.32)+eq(t\,10.36)+eq(t\,10.4)+eq(t\,10.44)+eq(t\,17.52)+eq(t\,47.96)+eq(t\,49.08)+eq(t\,49.2)+eq(t\,55.72)+eq(t\,83.0)'
>>>
>>> -q 5 -c:a copy -vsync vfr debug/outsides/CYD_1.5m_x264_%02d.png
>> I'm not sure how ffmpeg's expression evaluation calculates 't' (means I
>> couldn't find it in the source), but seeing that your av_ts2minutestr()
>> certainly does it differently, I may point out: floating point equality
>> is a bitch. ;-) I'm saying: 10.2 may not be equal to 10.2,
>> computer-mathematically speaking.
>
> Using eq() on floating points is indeed a bad idea because of floating
> point rounding.
In the code of the select filter I've found:
#define TS2D(ts) ((ts) == AV_NOPTS_VALUE ? NAN : (double)(ts))
    select->var_values[VAR_T  ] = TS2D(frame->pts) *
av_q2d(inlink->time_base);

So there rounding happens twice.

This could be avoided with:
    select->var_values[VAR_T  ] = (frame->pts == AV_NOPTS_VALUE) ? NAN :
frame->pts * inlink->time_base.num / inlink->time_base.den;

But after applying this I got even worse results, so there must be
another rounding counterpart somewhere which "heals" most rounding
issues, but not all as I can see with "10.2" from my example.

-Ulf