[FFmpeg-devel] [PATCH]lavf/mpegts: Do not print the character coding as part of service name

[Marton Balint]

On Thu, 7 Feb 2019, Carl Eugen Hoyos wrote:

> 2019-02-07 21:40 GMT+01:00, Marton Balint <cus at passwd.hu>:
>>
>>
>> On Thu, 7 Feb 2019, Carl Eugen Hoyos wrote:
>>
>>> 2019-02-07 16:34 GMT+01:00, Moritz Barsnick <barsnick at gmx.net>:
>>>> On Thu, Feb 07, 2019 at 14:40:40 +0100, Carl Eugen Hoyos wrote:
>>>>> +    if (p[0] < 0x20) {
>>>>> +        p++;
>>>>> +        if (--len < 0)
>>>>> +            return NULL;
>>>>> +    }
>>>>
>>>> If I understand section "A.2 Selection of character table" of ETSI EN
>>>> 300 468 correctly, you need to drop an additional byte if the first
>>>> byte (p[0]) is 0x1F, or an additional two bytes if the first one is
>>>> 0x10.
>>>
>>> New patch attached.
>>
>> Don't simply drop charset information.
>
>> You should convert the strings to UTF-8 based on that.
>
> (Yes we should)
> Do you have a sample that needs this?

There are tons of samples that needs this, two examples:

- samples/ffmpeg-bugs/roundup/issue1871/dvbt.ts
- samples/MPEG2/res_change_ffmpeg_aspect.ts

> If not, a patch that fixes a user-reported issue would rot
> because a better (unneeded) fix is possible.

In this case I think an incomplete fix hurts more, because after it the 
API user will have no way to determine the encoding. Using Iconv to 
convert between the code pages should not be too much work.

Regards,
Marton