Re: [xml] Patch to guard #define WIN32_LEAN_AND_MEAN

[Earnie <earnie users sourceforge net>]

On 11/9/2017 12:42 PM, Nick Wellnhofer wrote:
> On 09/11/2017 15:33, Earnie wrote:
> > Testing for !defined(__CYGWIN__) here is meaningless.  _WIN32 isn't
> > defined in GCC with __CYGWIN__ defined and vice versa.
>
> I think really old Cygwin versions used to set _WIN32 by default. See
> for example this thread from 1998:
>
>     https://www.cygwin.com/ml/cygwin/1998-08/msg00569.html
>     https://www.cygwin.com/ml/cygwin/1998-09/msg00002.html
>
> I assume you're Earnie Boyd who also participated in this thread, so you
> probably know better than me. My guess is that's where the ubiquitous
> `defined(_WIN32) && !defined(__CYGWIN__)` idiom came from.

Yes, that would be me but I don't know that I know better. Yes, really
old Cygwin did.  I don't know which versions of GCC stopped the
automatic define but anyone using anything that old deserve what they get.

> I'd be happy to remove all the `!defined(__CYGWIN__)` checks.

I'd say remove it.  Doesn't make sense to keep them and only convolutes
the code needlessly.

-- 
Earnie