Re: APR 1.0 released

Peter Eisentraut wrote:
> Bruce Momjian wrote:
>
> > OK, I am wrong above. Coding assumes WIN32 is only for port named
> > WIN32, which is mingw, and for BCC and VCC. I was not aware Cygwin
> > defined it at all. Are we sure it does in a header file?
>
> The problem is that some pieces of Cygwin code include windows.h, which
> it shouldn't do. If you fix those places, then there is no problem.

There are alot of windows.h includes:

/include/port/win32.h:7:#include <windows.h>
/port/crypt.c:56:#include <windows.h>
/port/dirmod.c:45:#include <windows.h>
/port/dirmod.c:405:#include <windows.h>
/port/open.c:16:#include <windows.h>
/port/sprompt.c:35:#include <windows.h>
/timezone/zic.c:22:#include <windows.h>
/utils/dllinit.c:44:#include <windows.h>
/bin/pgevent/pgevent.c:15:#include "windows.h"
/bin/psql/input.c:14:#include <windows.h>
/bin/psql/mbprint.c:18:#include <windows.h>
/bin/psql/startup.c:16:#include <windows.h>
/interfaces/libpq/libpqdll.c:3:#include <windows.h>
/interfaces/libpq/pthread-win32.c:14:#include "windows.h"
/interfaces/libpq/win32.c:30:#include <windows.h>
/backend/port/dynloader/win32.c:3:#include <windows.h>

and I bet some of the system includes that we use call windows.h
themselves.

> > I wonder if we should just call the port mingw and change the proper
> > defines to __MINGW__. We would then create a define called
> > WIN32_NATIVE that is defined for __MINGW__, BVC and VCC.
>
> WIN32 is the correct symbol; see above.

I hope you are right.

--
Bruce Momjian | http://candle.pha.pa.us
pgman(at)candle(dot)pha(dot)pa(dot)us | (610) 359-1001
+ If your life is a hard drive, | 13 Roberts Road
+ Christ can be your backup. | Newtown Square, Pennsylvania 19073