Magnus Hagander wrote:
>>Recent changes to interfaces/libpg/fe-connect.c have broken
>>MSVC builds,
>>which I am reliably informed are definitely still required - at least
>>some uses of libpq.dll made with gcc apparently cause the resulting
>>builds or applications to blow up.
>>
>>
>
>Can you get some more specifics on that? We know it was a fact with
>early betas because of the conflicting exports, but I was under the
>impression that these issues were all fixed by now.
>(not saying the MSVC build shouldn't work, just that the gcc should work
>too...)
>
>
TJ (see addresses) can speak to this.
>
>
>
>>The trouble seems to come from these 2 lines:
>>
>>+ #define _WIN32_IE 0x0400
>>+ #include <shlobj.h>
>>
>>First, on MSVC, _WIN32_IE is already defined at that point. If you get
>>around that then processing the include file causes errors.
>>
>>
>
>Interesting - mingw was supposed to use compatible headers, I thought
>:-)
>Can you tru moving the #define up to the top of the file and see if that
>fixes things? So it's pulled before *any* windows include file. (My msvc
>has completely stopped building fromt he commandline. Need to fix that,
>but it's not up right now)
>
>
>
>
I surrounded the #define with #ifndef ... #endif. If you agree that
would be the first fix.
But most errors remained (see my follow up email yesterday)
The error appears to be on line that uses NEAR and complains about it
... I see that the MinGW windef.h defines NEAR as empty, while the MSVC
windef.h defines it as "near". Don't know if that makes a difference.
cheers
andrew