[ net-snmp-Bugs-744637 ] CONFIGURE_OPTIONS missing '\' on openbsd

[SourceForge.net <no...@so...>]

Bugs item #744637, was opened at 2003-05-27 17:39
Message generated for change (Comment added) made by sf-robot
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=112694&aid=744637&group_id=12694

Please note that this message will contain a full copy of the comment thread,
including the initial issue submission, for this request,
not just the latest update.
Category: configure
Group: openBSD
>Status: Closed
Resolution: Fixed
Priority: 5
Private: No
Submitted By: Robert Story (rstory)
Assigned to: Nobody/Anonymous (nobody)
Summary: CONFIGURE_OPTIONS missing '\' on openbsd

Initial Comment:
IRC user, building net-snmp-5.0.8 from source
Host: OpenBSD 3.2 i386

Syska: its running now: ./configure --with-perl-modules
--with-mib-modules=host

Syska: still look likethis in net-snmp-config.h

Syska: #define CONFIGURE_OPTIONS "" --with-perl-modules
--with-mib-modules=host""



This causes ucd-snmp/versioninfo.c fail to compile..
ucd-snmp/versioninfo.c: In function
`var_extensible_version':
ucd-snmp/versioninfo.c:97: wrong type argument to decrement
ucd-snmp/versioninfo.c:97: syntax error before `with'
*** Error code 1


----------------------------------------------------------------------

>Comment By: SourceForge Robot (sf-robot)
Date: 2006-12-21 19:20

Message:
Logged In: YES 
user_id=1312539
Originator: NO

This Tracker item was closed automatically by the system. It was
previously set to a Pending status, and the original submitter
did not respond within 180 days (the time period specified by
the administrator of this Tracker).

----------------------------------------------------------------------

Comment By: Thomas Anders (tanders)
Date: 2006-06-24 04:58

Message:
Logged In: YES 
user_id=848638

This is believed to be fixed in (at least) 5.2.2 and later.

----------------------------------------------------------------------

Comment By: Brian A. Seklecki (bseklecki)
Date: 2003-08-10 06:06

Message:
Logged In: YES 
user_id=500741

...I traced it from the back to the front. From defs.sed, we
see that
it's wrong in the conftest.defines and confdef.undefines. 
So if it's
bad at that point, then it must be bad in confdefs.h.

...and so, it turns out not to be a sed(1) but a cat(1) or
possibly an
sh(1) problem:


NetBSD conftest.defines (correct)
${ac_dA}CONFIGURE_OPTIONS${ac_dB}CONFIGURE_OPTIONS${ac_dC}"\\\\"'--with-defaults'
'--prefix=/usr/local' '--with-logfile=/var/log/snmpd'
'--with-sys-contact=bofh@precursor'
'--with-sys-location=Arlington\\, VA Datacenter
1'\\\\""${ac_dD}



OpenBSD conftest.defines (incorrect):

${ac_dA}CONFIGURE_OPTIONS${ac_dB}CONFIGURE_OPTIONS${ac_dC}""'--with-defaults'
'--prefix=/usr/local' '--with-logfile=/v
ar/log/snmpd' '--with-sys-contact=bofh@precursor'
'--with-sys-location=Arlington\\, VA Datacenter 1'""${ac_dD}


--------------------

NetBSD confdefs.h (correct)

#define CONFIGURE_OPTIONS "\"'--with-defaults'
'--prefix=/usr/local'
'--with-logfile=/var/log/snmpd'
'--with-sys-contact=bofh@precursor'
'--with-sys-location=Arlington, VA Datacenter 1'\""

OpenBSD confdefs.h (incorrect)

#define CONFIGURE_OPTIONS ""'--with-defaults'
'--prefix=/usr/local'
'--with-logfile=/var/log/snmpd'
'--with-sys-contact=bofh@precursor'
'--with-sys-location=Arlington, VA atacenter 1'""



$ac_configure_args gets set early on, it's value is
identical on both
systems:

So at some point, the double quotes people use at the
command line to
pass strings to --with-sys-location, etc, are removed and
the whole flag
is encapsulated in a single quote as seen here:

'--with-defaults' '--prefix=/usr/local'
'--with-logfile=/var/log/snmpd'
'--with-sys-contact=bofh@precursor'
'--with-sys-location=Arlington, VA
Datacenter 1'

(I know both systems are identical at this point as well)


*** So the variable is right at this point, *BUT*, when it
gets appended
to confdefs.h, it becomes corrupted on OpenBSD?  Why, well, for
starters's it's cat(1)'d intead of echo(1)'d for some reason
(performance issue? I'm not a autoconf hacker, but they
always taught us
to use echo(1) or echo builtins over cat(1) if at all possible)

Anyway the line is:

----------------------------------------------------------
#
# save the configure arguments
#
cat >>confdefs.h <<_ACEOF
#define CONFIGURE_OPTIONS "\"$ac_configure_args\""
_ACEOF
----------------------------------------------------------

Now my interpretation of this is:

#define CONFIGURE_OPTIONS "'--switch1' '--switch2' '--switch3'"

...but i guess that's from all those years of echo(1) over
cat(1). 
However since this IS cat(1), I would  be more inclined to
think along
an OpenBSD state of mind

#define CONFIGURE_OPTIONS ""'--switch1' '--switch2'
'--switch3'""

But apparently that's incorrect, so I'm in the minority, as
Linux and
NetBSD versions parse this correctly. I suppose this could
also be a
sh(1) problem with the way control characters are escaped
*BEFORE*
they're handed to cat(1); but, I don't see any major
differences in the
man pages for NetBSD and OpenBSD cat(1) or sh(1), and I'm
not prepared
to sift through five+ years of CVS commits to find out when this
behavior changed.

"you bring new meaning to the terms slackass. I will have to
invent a
new term." -- theo de raadt

Anyway, a thousand regex's later, it's still screwed up when
this all
get applied to include/net-snmp/net-snmp-config.h.in and
then of course
the breakage during the build.

This has certainly been exhilarating ride.  I think I'm in
danger of
breaking a sweat over here.  I am, however, not inclined to
propose a
fix -- that's far beyond me.  I vote for re-writing the
whole thing in
LISP or MONO, but that argument won't fly in the lists, I'm
sure.

-lava


----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=112694&aid=744637&group_id=12694