Am 08.02.22 um 10:27 schrieb Andreas Rheinhardt:
> Thilo Borgmann:
>> Am 01.02.22 um 00:40 schrieb Andreas Rheinhardt:
>>> Thilo Borgmann:
>>>> Am 31.01.22 um 14:08 schrieb Nicolas George:
>>>>> Thilo Borgman (12022-01-31):
>>>>>>> v10 attached.
>>>>>>
>>>>>> Also going to apply soon if there are no more comments.
>>>>>
>>>>> I think you neglected to attach the file.
>>>>
>>>> omg stupid me. Here it is...
>>>>
>>>> -Thilo
>>>>
>>>
>>> Seems like I misunderstood your code and ignored the outer while. Your
>>> code can leak if there are multiple 'N', because (as I said)
>>>
>>>>
>>>> +
>>>> +            if (fmt_new && fmt_new != argv[0] && fmt_new !=
>>>> fmt_default)
>>>> +                av_freep(&fmt_new);
>>>> +
>>>
>>> does not free fmt if it is already allocated. It is possible to fix this
>>> by adding an char *fmt_allocated = NULL; at outer scope and a fmt_new in
>>> the block that is executed if a 'N' is executed. (You have to free
>>> fmt_allocated immediately after av_asprintf().)
>>
>>> But maybe it would be best to use an AVBPrint to write the new string
>>> instead of allocating a new string for every %N encountered.
>>
>> v11 does it that way.
>>
> 
>>
>> +    av_bprintf(&fmt_bp, "%s", fmt_begin);
>> +    av_bprint_finalize(&fmt_bp, (char**)&fmt);
>> +    if (!fmt)
>> +        return AVERROR(ENOMEM);
>> +
>>       av_bprint_strftime(bp, fmt, &tm);
>> +    av_freep(&fmt);
>> +
> 
> This is not how one uses an AVBPrint: You are loosing the
> small-string-optimization that way.
> Furthermore, you do not check for truncation.

v12 including IRC comments.

-Thilo