Bug 60

Summary: OpenSSH 3.0.2p1 configure script fails on NEXTSTEP (3.1p1, too)
Product: Portable OpenSSH Reporter: Vic Abell <abe>
Component: Build systemAssignee: Ben Lindstrom <mouring>
Status: CLOSED INVALID    
Severity: major    
Priority: P2    
Version: -current   
Hardware: 68k   
OS: Other   
Attachments:
Description Flags
Example of the bug and supporting information none

Description Vic Abell 2001-12-26 23:19:16 AEDT 
The OpenSSH configure script no longer creates an empty confdefs.h header file,
as it once did as recently as 2.9.9p2.  This causes the configure compiler
test to fail on NEXTSTEP 3.1 with gcc 2.8.

As a workaround it is possible to "touch confdefs.h" before running configure.

Vic Abell
Comment 1 Damien Miller 2002-01-14 19:28:54 AEDT 
Why / where is this needed? Can you show an example of a failed build. This may
be an autoconf bug.
Comment 2 Vic Abell 2002-01-14 23:54:22 AEDT 
Created attachment 9 [details]
Example of the bug and supporting information
Comment 3 Tim Rice 2002-01-31 16:13:51 AEDT 
looks like a compiler bug.
....from your config.log
> In file included from configure:1351:
> confdefs.h:2: parse error at null character
....

confdefs.h does indeed get created before any tests.
But that is done as part of the autoconf macros and not part
of OpenSSH
Comment 4 Vic Abell 2002-01-31 22:54:17 AEDT 
tim@multitalents.net says:
>
> confdefs.h does indeed get created before any tests.
> But that is done as part of the autoconf macros and not part
> of OpenSSH

As distributed openssh-3.0.2p1.tar.gz contains no confdefs.h --
I just checked in a fresh download at 1340Z on January 31, 2002.
If you type "./configure" out of the box as the INSTALL script
says, there will be no confdefs.h.
Comment 5 Damien Miller 2002-01-31 23:27:36 AEDT 
We're not supposed to ship a confdefs.h file. It is configure's job to create
it. This seems to be an autoconf bug.

Could someone who is affected by this problem please raise a bug with the
autoconf developers?
Comment 6 Ben Lindstrom 2002-04-06 08:24:58 AEST 
I'll be doing NeXTStep testing this weekend if all goes well.  So I'll
see what the issue is on 68k hardware.
Comment 7 Ben Lindstrom 2002-04-07 10:40:10 AEST 
I just build --current (with a few patches outstanding) with the default gcc 
compiler and using autoconf 2.5.2 (which I believe is our current build 
standard, right Damien?)  And I could not reproduce this bug.
Comment 8 Vic Abell 2002-04-29 21:24:28 AEST 
Contrary to the resolver's claim, this bug has not been fixed.  It is present
in 3.1p1, too.  Here is script output of three shell commands: 1) a failed
3.1p1 configure attempt; 2) a `touch confdefs.h` remedy being applied; and 3)
a successful 3.1p1 configure attempt.  (I also included some of the context
following the successful configure attempt to show that it did succeed.)

Script started on Mon Apr 29 04:49:22 2002
vic: 4 = ./configure --with-prngd-socket=/etc/entropy --with-tcp-wrappers \
> --with-default-path=/usr/local/openssh/bin:/bin:/usr/bin:/usr/ucb:/usr/local/b
in \
> --prefix=/usr/local/openssh-3.1p1
checking for gcc... gcc
checking for C compiler default output... configure: error: C compiler cannot cr
eate executables
vic: 5 = touch confdefs.h
vic: 6 = ./configure --with-prngd-socket=/etc/entropy --with-tcp-wrappers \
> --with-default-path=/usr/local/openssh/bin:/bin:/usr/bin:/usr/ucb:/usr/local/b
in \
> --prefix=/usr/local/openssh-3.1p1
checking for gcc... gcc
checking for C compiler default output... a.out
checking whether the C compiler works... yes
...
Comment 9 Ben Lindstrom 2002-04-30 00:28:05 AEST 
We are releasing 3.2.1 soon.  Test --current or the snapshots at 
http://www.openssh.com/portable.html

BTW there is a patch required to compile 3.2.1 due to NeXT's embeding of monitor
() in libc.

- Ben
Comment 10 Damien Miller 2004-04-14 12:24:17 AEST 
Mass change of RESOLVED bugs to CLOSED