2.99.34 on AIX (problems). Was: Z-mailer timezone oddity

["Andrew P. Kokarev" <andrew@skfgvc.pyatigorsk.su>]

David Nugent writes:
> On Fri, 2 Aug 1996, Matti Aarnio wrote:
> 
> >	... Several thoughts and tests latter ...
> >	Now I have autoconfig test for timezonal information, and
> >	do use it.  Works at OSF/1, and Solaris 2.5.
> >	It will report an "Aiee..." if it can't figure out the
> >	system timezone method.
> 
> I get this message from configure under Linux using libc 5.3.12.
> The "timezone" method in this case uses the extern long timezone
> in <time.h> with no altzone, but right now zmailer has no timezone
> sense at all. :-(
> 
> How can I get it to work?
> 
> This is with 2.99.34, btw.

I tried to make 2.99.34 on AIX 4.1.1 RS6000/40P (ppc 601) with gcc 2.7.2.
The same "Aiee" message. I fixed it with the following patch:






*** rfc822date.c.orig	Fri Aug  2 12:25:42 1996
--- rfc822date.c	Sun Aug  4 15:35:18 1996
***************
*** 32,37 ****
--- 32,55 ----
  	  offset = -timezone;
  	offset /= 60;				/* Offset in minutes */
  #else
+ #ifdef HAVE_GETTIMEOFDAY	/* On AIX 4.1.1 it works, and depends from TZ */
+ 	{                  	/* On BSD/OS 2.0.1 it always gives UTC */
+ 	  struct timeval s__tv;	/* but BSD/OS has tm_gmtoff */
+ 	  struct timezone s__tz;
+ 	  if (gettimeofday(&s__tv, &s__tz) == 0) {
+ 	    offset = - s__tz.tz_minuteswest;
+ 	    switch (s__tz.tz_dsttime) {
+ 	    case DST_USA:
+ 	      offset += 60;
+ 	    case DST_NONE:
+ 	      break;
+ 	    default:		/* XXX: Best guess */
+ 	      offset += 60;
+ 	    }
+ 	  }
+ 	  else /* Should never get here - gettimeofday() never fails with
+ 	  	correct pointers */
+ #endif
  	{		/* This is fallback stuff, beware! */
  	  time_t tm;
  	  struct tm ts2;
***************
*** 40,45 ****
--- 58,66 ----
  	  tm     = mktime(&ts2);
  	  offset = (tm - *timep) / 60; /* Offset in minutes */
  	}
+ #ifdef HAVE_GETTIMEOFDAY
+ 	}
+ #endif	
  #endif
  #endif
  	sign   = offset >= 0;









> Some other things that should probably be fixed; these are with the

For example, AIX doesn't have <sys/fcntl.h> ,
and sysprotos.h includes it:

#ifdef STDC_HEADERS
#include <sys/fcntl.h>

But configure never checks for existence of sys/fcntl.h !

And configure thinks that --without-bundled-libresolv is
--with-bundled-libresolv.

In libsh/interpret.c , line 1917 there is declaration of wait()
(SYSV one, with int parameter, not union). Why???
Then, in libsh/builtins.c there is 
	int status = 0;    /* line 1041 */
and that status later used in WEXITSTATUS macros, which in AIX
with -D_BSD (and, probably, in real BSD) wants union.

Ok, let's try to include -D_BSD only in libresolv/Makefile.in ...

Hmm... smtpserver still doesn't link...

-- 
/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\
Joint stock company "GVC Energeticy", North-Caucasian Branch, Pyatigorsk
Andrew Petrovich Kokarev, postmaster         andrew@skfgvc.pyatigorsk.su