From: tolj AT uni-duesseldorf DOT de (Stipe Tolj)
Subject: SERIOUS B20 BUG: pipelining seems to break (wasB20: weird variables 
         bug (tryed on win98))
5 Nov 1998 11:08:14 -0800
Message-ID: <363F106A.17249244.cygnus.gnu-win32@uni-duesseldorf.de>
References: <363BF5C5 DOT 18FA323 AT bigfoot DOT com>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
To: creative1 AT bigfoot DOT com, Geoffrey Noer <noer AT cygnus DOT com>
Cc: gnu-win32 AT cygnus DOT com

> checking for rename... (cached) yes
> HAVE_RENAME            <-- this is the echo
> checking for getrusage... (cached) yes
> HAVE_GETRUSAGE           <-- this is the echo
> checking for getdtablesize... (cached) yes
> HAVE_GETDTABLESIZE
> checking for srandom... (cached) yes
> HAVE_SRANDOM
> checking for random... (cached) yes
> HAVE_RANDOM     < the echo...
> checking for sigaction... (cached) yes
> HAVE_  <-- mistake... why? as you can see it detect it perfect, i check
> everything
> after this i have a collection of
> HAVE_
>
> and on the file that it writes...
>
> #define HAVE_RENAME 1
> #define HAVE_GETRUSAGE 1
> #define HAVE_GETDTABLESIZE 1
> #define HAVE_SRANDOM 1
> #define HAVE_RANDOM 1
> #define HAVE_ 1
> #define HAVE_ 1
> #define HAVE_ 1
> #define HAVE_ 1

 I have the same problem using cygwin b20 on a Win98 box running the
autoconfig generated "configure" script of the mysql-3.22.9
distribution.
This is defenetly a b20 bug, since I tested it using my old b19 setup
and it
works perfectly. I have located the place within configure which cause
the
problems. The following code

for ac_hdr in fcntl.h float.h floatingpoint.h ieeefp.h limits.h \
 memory.h pwd.h select.h \
 stdlib.h stddef.h \
 strings.h string.h synch.h sys/mman.h \
 sys/timeb.h sys/types.h sys/un.h sys/vadvise.h sys/wait.h term.h \
 unistd.h utime.h sys/utime.h termio.h termios.h sched.h crypt.h
do
ac_safe=`echo "$ac_hdr" | sed 'y%./+-%__p_%'`          <------- here
goes the
BUG!!!!!!!!!!!!
#ac_safe=`echo $ac_hdr | sed 'y%./+-%__p_%'`
echo $ac_n "checking for $ac_hdr""... $ac_c" 1>&6
echo "configure:2745: checking for $ac_hdr" >&5
echo "test \"`echo '$''{'ac_cv_header_$ac_safe'+set}'`\" = set"
if eval "test \"`echo '$''{'ac_cv_header_$ac_safe'+set}'`\" = set"; then
  echo $ac_n "(cached) $ac_c" 1>&6
else
  ...

is working fine, but suddenly the $ac_safe variable stays unset within
the
for do loop, causing the script to assume the header file has already
been
cached into config.cache, which indeed has not.

I have tried both lines, the one commented and the orginal but both have
the
same fault effect. The whole configure script breaks at some time, since
some
libraries are linked to conftest which the script has "cached" (but not
present within b20) like libbind.a libnsl.a etc.

This seems to be a defenite b20 bug. Has anyone experienced it on an
other OS
than Win98?

Help is needed, Geoffrey.

Regards,
Stipe

--
Stipe Tolj <tolj AT uni-duesseldorf DOT de>
Cygwin32 Porting Project
Department of Economical Computer Science
University of Cologne, Germany
http://www-public.rz.uni-duesseldorf.de/~tolj
-
For help on using this list (especially unsubscribing), send a message to
"gnu-win32-request AT cygnus DOT com" with one line of text: "help".