[FFmpeg-devel] [PATCH] change av_ts_make_time_string precision

[Marton Balint]

On Wed, 13 Mar 2024, Allan Cady via ffmpeg-devel wrote:

> On Tuesday, March 12, 2024 at 02:24:47 PM PDT, Marton Balint <cus at passwd.hu> wrote: 
>> On Tue, 12 Mar 2024, Allan Cady via ffmpeg-devel wrote:
>>> On Monday, March 11, 2024 at 12:11:45 PM PDT, Marton Balint <cus at passwd.hu> wrote:
>>>> On Mon, 11 Mar 2024, Andreas Rheinhardt wrote:
>>>> Allan Cady via ffmpeg-devel:
>>>>> From: "Allan Cady" <allancady at yahoo.com>
>>>>>

[...]

>
>
>> One thing to notice is that you will not need to use the scientific
>> representation at all, because maximum value this function prints is the
>> product of an INT32 and an INT64, and that is 96 bit, which is at most 29
>> chars. Adding the optional sign and the decimal point, that is still only
>> 31. So we can be sure that by using %.6f, the first character of
>> the decimal point is going to be present in the output.
>
>
> I had done some similar calculation and came to a similar conclusion. 
>
>
> Which is great,
>> because that means we only have to
>> - do a single snprintf("%.6f")
>> - calculate last char position by subtracting 1 from the minimum of
>> AV_TS_MAX_STRING_SIZE-1 and the result of the snprintf() call.
>> - decrement string length while last char is '0' to remove trailing 0s
>> - decrement string length while last char is non-digit to remove decimal
>> point (which can be a multiple chars for some locales).
>> - update last+1 char to \0.
>> Ot is it still too complex to keep it inline?
>
>
> I'll give your suggestion a spin tomorrow. Thanks.
>

In the end I posted a patch myself, sorry if you were working on it too, 
it just looked a bit too complex for a new developer, since it touched 
API/build system/etc... I hope you don't mind.

Regards,
Marton