bug#37158: 27.0.50; Build failure (MSYS2, master)

[Eli Zaretskii]

> From: Richard Copley <rcopley@gmail.com>
> Date: Fri, 23 Aug 2019 13:44:22 +0100
> 
>  The immediate problem should be fixed now, but what worries me is how
>  on earth did your MinGW build ended up compiling the
>  !HAVE_GAI_STRERROR branch?  This shouldn't happen, by virtue of this
>  part of nt/mingw-cfg.site:
> 
>    ac_cv_func_gai_strerror=yes
> 
>  Doesn't your src/config.h has this line:
> 
>    #define HAVE_GAI_STRERROR 1
> 
>  ?  If not, we should find out why.
> 
> Yes, that line is there, but also, in nt/ms-w32.h, there is this:
> 
> /* Unlike MS and mingw.org, MinGW64 doesn't define gai_strerror as an
>    inline function in a system header file, and instead seems to
>    require to link against ws2_32.a.  But we don't want to link with
>    -lws2_32, as that would make Emacs dependent on the respective DLL.
>    So MinGW64 is amply punished here by the following:  */
> #undef HAVE_GAI_STRERROR

Ah, okay.  It's clear now.