Re: Suspicious warning in W64 build

[Richard Copley]

On 16 September 2017 at 09:19, Richard Copley <address@hidden> wrote:
> On 16 September 2017 at 07:40, Eli Zaretskii <address@hidden> wrote:
>>
>> If you look at the beginning of w32.c, you will see that we include
>> string.h before we #undef strerror.  The header nt/inc/ms-w32.h
>> #define's strerror to sys_strerror, so the inclusion of string.h
>> should have supplied the prototype for sys_strerror.  Why doesn't it
>> do that in MinGW64?  Does some system header #undef strerror too
>> early?
>
> I see, thanks. No, there's no #undef. We end up including string.h before
> ms-w32.h (w32.c includes io.h and io.h unconditionally includes string.h).

Also beware fcntl.h includes io.h. But string.h is not included before
ms-w32.h otherwise than via fcntl.h and io.h.