Re: [hatari-devel] Removing SDL use for byte order checks / conversions

I would conditionally define M_PI most accurately in a future-proof way.

M_PI has always been there with the C compiler (UNIX98 Borland ...),

thus relying on a separate package (SDL) is weird.

https://www.gnu.org/software/libc/manual/html_node/Mathematical-Constants.html

I think MSFT is trying to be a trouble maker here, so use the force.

On Mon, Oct 3, 2022 at 5:24 AM Nicolas Pomarède <npomarede@xxxxxxxxxxxx> wrote:

Le 03/10/2022 à 14:18, Thomas Huth a écrit :
> Am Mon, 3 Oct 2022 11:11:42 +0300
> schrieb Eero Tamminen <oak@xxxxxxxxxxxxxx>:
>
>> Hi,
>>
>> On 2.10.2022 22.05, Christian Zietz wrote:
>>> Thomas Huth schrieb:
>>>
>>>> Without that, compilation was broken with VisualStudio:
>>>>
>>>> https://cirrus-ci.com/task/6045603464478720?logs=main#L752
>>>
>>> See:
>>> https://learn.microsoft.com/en-us/cpp/c-runtime-library/math-constants?view=msvc-170
>>
>> Ok, Thanks!
>>
>> So better solution than the SDL include, would be adding this to
>> "main.h", before <math.h> include:
>> #define _USE_MATH_DEFINES // for MSVC
>>
>> Thomas, could you try that in your CI?
>
> Not sure if that's really better - if I've got that right, M_PI and the
> other constants are not part of POSIX or the C standard, so there might be
> other systems where M_PI is not defined. OTOH, SDL_stdinc.h always makes
> sure that it is defined, so unless we really really need to make dmaSnd.c
> independent from the SDL headers, using SDL_stdinc.h is maybe the easier
> choice here.

Or if we really just need M_PI, maybe we can define it ourselves if it's
not already there with our own prefix ?

#define HATARI_M_PI 3.14xxxxx

That's just in case we really want to get rid of as many SDL includes as
possible ; if not, we can keep SDL_stdinc.h if it solves cases for all
possible combination of architectures / compilers.