[MPlayer-dev-eng] [PATCH] Try to adapt to recent compilers such as GCC 14

[Reimar Döffinger]

> On 27 Dec 2024, at 19:37, Reimar Döffinger <Reimar.Doeffinger at gmx.de> wrote:
> 
> Hi!
> 
>> On 2 Oct 2024, at 14:09, KO Myung-Hun <komh78 at gmail.com> wrote:
>> 
>> Hi/2.
>> 
>> Ingo Brückl wrote:
>>> Lorenzo wrote on Tue, 1 Oct 2024 14:25:54 +0200:
>>> 
>>>> On Tue, 01 Oct 2024 14:13:36 +0200
>>>> Ingo Brückl <ib at oddnet.de> wrote:
>>> 
>>>>> Lorenzo wrote on Tue, 1 Oct 2024 12:52:48 +0200:
>>>>> 
>>>>>> with this patch applied it still fails[1] to build on i386 (Debian)
>>>>>> for me, with
>>>>> 
>>>>>>  73 |     if (iconv(inv_msgiconv, (const char **)&filename, &filename_len,
>>>>>>     |                             ^~~~~~~~~~~~~~~~~~~~~~~~
>>>>>>     |                             |
>>>>>>     |                             const char **
>>>>> 
>>>>> Which patch? It's okay in current svn.
>>> 
>>>> Apologies, I was not subscribed so I just copied the subject without a
>>>> proper quote; the patch posted in
>>>> https://lists.mplayerhq.hu/pipermail/mplayer-dev-eng/2024-September/074217.html
>>> 
>>> This patch is broken regarding all (const char **) casts for iconv()
>>> which takes char ** as a second argument.
>>> 
>> 
>> Hmm... OS/2 iconv() has `const' qualifier.
> 
> POSIX is quite clear that that's wrong.
> If at all possible it would be much preferable to make OS/2
> behave as POSIX says.
> If not an option, I think it would be better to have
> something that separates out OS/2 behaviour as it is wrong.
> Maybe wrapping iconv or a define.
> Or a #define for the proper type to cast the iconv input buffer
> to with a comment that/why it's different on OS/2

I forgot a 3rd option: maybe even just disabling the warning for OS/2...