From: Marius Storm-Olsen <mstormo@gmail.com>
Date: Sat, 22 Aug 2009 11:40:15 GMT
Subject: Re: [PATCH 09/14] Avoid including windows.h in winansi.c for MSVC build
Message-ID: <18cd41840908220440o37cf0909l15ff552abd52e075@mail.gmail.com>
In-Reply-To: <3f4fd2640908211416i489b6e23s27cb9cae6a439ddc@mail.gmail.com>

On Fri, Aug 21, 2009 at 11:16 PM, Reece Dunn<msclrhd@googlemail.com> wrote:
> 2009/8/21 Junio C Hamano <gitster@pobox.com>:
>> Marius Storm-Olsen <mstormo@gmail.com> writes:
>>
>>> From: Frank Li <lznuaa@gmail.com>
>>>
>>> compat/msvc.h includes winsock2.h which conflicts with windows.h.
>>> msvc.h also defines the oldest Windows API version required.
>>
>> The first sentence sort-of makes sense; compat/msvc.h will be included by
>> git-compat-util.h and including <windows.h> here will bring conflicting
>> definitions, so we avoid doing so when on MSC.
>>
>> The second sentence does not make any sense to me.  It may be correctly
>> stating a fact (i.e. "defines required WAPI version"), but it is unclear
>> what relevance it has to this change to stop including <windows.h>.
>
> The way things are configured, windows.h is pulling in winsock.h. In
> git-compat-util.h, winsock2.h is included which has conflicting
> definitions of various functions and data structures.
>
>> Having said that, the first sentence hints me that perhaps you guys should
>> include (conditionally only on windows but not with MSC) <windows.h> not
>> from this file, but from git-compat-util.h?
>
> It would make sense for windows.h to be included in git-compat-util.h.
>
> According to http://social.msdn.microsoft.com/Forums/en-US/windowssdk/thread/4a90b143-1fb8-43e9-a54c-956127e0c579,
> the following will work:
>
> #define _WINSOCKAPI_    // stops windows.h including winsock.h
> #include <winsock2.h>
> #include <windows.h>
>
> Also, if you define WIN32_LEAN_AND_MEAN, windows.h will pull in a
> subset of the Windows header files (which also improves compilation
> times). Adding this may prevent it from pulling in winsock.h. This
> would be a better approach (and would make sense to go into
> git-compat-util.h).
>
> I don't have access to a Windows dev box at the moment, so can't
> verify that this does indeed work.
>
> - Reece
>

I'll check this on Monday, and send out a v3 of the series,
incorporating the pieces commented on, and clarifying some of the
still confusing commit msgs.

--
.marius