Re: _WIN32_WINNT redefined?

[Eli Zaretskii <eliz@gnu.org>]

> Date: Wed, 02 Nov 2022 17:20:50 +0000
> From: i.nixman@autistici.org
> 
> On 2022-11-02 16:51, Eli Zaretskii wrote:
> 
> > That's on your system, with your MinGW w32api headers.  But that's not
> > the only game in town.
> 
> I'm sure there is no WINAPI implementation that does not define 
> _WIN32_WINNT. it's just impossible.
> but in line 75 the condition is exactly for that impossible case.

It is not impossible.

I use mingw.org's MinGW, where the header which defines _WIN32_WINNT
is included by every header file.  So as soon as you #include
anything, _WIN32_WINNT is already defined.

Maybe it isn't so in the headers you are using, but that doesn't mean
it is so for everyone.

And I don't see how is that relevant: even if _WIN32_WINNT is
undefined before that place, my suggestion is to change this:

 # else
 #  define _WIN32_WINNT 0x0501

into something that defines _WIN32_WINNT to 0x0600 under the condition
that the gthread patch is being used, and to 0x0501 otherwise.  This
should solve everyone's problem, right?

> What I'm doing now is trying to tell you that it's illogical to check 
> the value of _WIN32_WINNT BEFORE the corresponding header file in which 
> this value was originally defined, has been included.

See above: that is not necessarily what happens, not in all the cases
anyway.