Thread: MSVC compile errors

I was egtting a compile error complaining (in effect) abour the lack of
a pid_t type when attempting a compile with MSVC. The attached patch,
which move the definition up in the file win32.h before it is used,
seems to cure the problem.

I am also seeing non-fatal warnings about redefinition of SIG_ERR
SIG_DFL and SIG_IGN which I will investigate.

cheers

andrew


Index: win32.h
===================================================================
RCS file: /home/cvsmirror/pgsql/src/include/port/win32.h,v
retrieving revision 1.41
diff -c -r1.41 win32.h
*** win32.h    1 Dec 2004 23:42:26 -0000    1.41
--- win32.h    26 Dec 2004 18:24:36 -0000
***************
*** 44,49 ****
--- 44,53 ----
  #define DLLIMPORT
  #endif

+ #ifdef WIN32_CLIENT_ONLY
+ typedef int pid_t;
+ #endif
+
  /*
   *    IPC defines
   */
***************
*** 210,219 ****
  #endif
  typedef long key_t;

- #ifdef WIN32_CLIENT_ONLY
- typedef int pid_t;
- #endif
-
  /*
   * Supplement to <sys/stat.h>.
   */
--- 214,219 ----

Andrew Dunstan wrote:

>
>
> I am also seeing non-fatal warnings about redefinition of SIG_ERR
> SIG_DFL and SIG_IGN which I will investigate.
>
>

The cure for this seems to be to add

  #include <signal.h>

near the top of src/include/port/win32.h.

cheers

andrew

Andrew Dunstan <andrew@dunslane.net> writes:
> I was egtting a compile error complaining (in effect) abour the lack of
> a pid_t type when attempting a compile with MSVC. The attached patch,
> which move the definition up in the file win32.h before it is used,
> seems to cure the problem.

I applied a somewhat more extensive patch that tries to put the contents
of win32.h in a saner order (function externs at the bottom).  Got the
<signal.h> problem too.

            regards, tom lane