[PATCH 2/2] include: Use long for LONG type on Windows targets iff not building Wine.

[Jacek Caban]

On 29.04.2020 22:44, Alexandre Julliard wrote:
> Jacek Caban <jacek at codeweavers.com> writes:
>
>> Signed-off-by: Jacek Caban <jacek at codeweavers.com>
>> ---
>>
>> Right now we use _MSC_VER to choose between long and int type. We
>> could, however, make any other choice. With patch 1, it's possible to
>> use int with MSVC (or at least clang MSVC target). It's also possible
>> for any other compilers except __LP64__ to use long.
>>
>> This raises some questions and I wouldn't mind changing the patch. For
>> example, I could change headers to always use int. The problem with
>> that is that it makes us incompatible with the rest of Windows world.
>>
>> In the past, the choice of int for LONG was motivated by port to
>> Win64, where we can't use long on UNIX side and it's nice to have the
>> same time on all targets.
>>
>> This patch is a compromise: it always uses int for Wine itself, so as
>> far as building Wine is considered, it's consistent among
>> compilers. Outside Wine, it uses long on Windows targets, where the
>> code is most likely to depend on underlying type. And it keeps using
>> int for non-Windows targets, so nothing changes for non-PE winelib.
>>
>> If there is a need for int LONG on Windows targets, we could introduce
>> a macro to change that choice. I could also imagine revisiting the
>> choice for PE parts of Wine itself, but it would be a lot of work.
> All of these options are of course possible, but it doesn't seem to me
> that any of them has a clear advantage over what we are doing now. Are
> there real problems with the current approach that we need to address?


The main problem that I tried to address is that with current approach, 
building Wine itself in MSVC mode causes a lot of format warnings and 
some incompatible pointer type warnings. That's because MSVC is #ifdefed 
to use long. To fix that problem, we probably could just remove MSVC 
ifdef and always use int, but entirely removing support for using long 
doesn't seem right to me. Maybe change #ifdef _MSV_VER with some WINE_* 
opt-in macro?


Thanks,

Jacek