Subject: Re: pkg/32031: pkgsrc-2005Q3 problems under NetBSD 2.1
To: None <pkg-manager@netbsd.org, gnats-admin@netbsd.org,>
From: Matthew Mondor <mm_lists@pulsar-zone.net>
List: pkgsrc-bugs
Date: 11/10/2005 20:06:01
The following reply was made to PR pkg/32031; it has been noted by GNATS.

From: Matthew Mondor <mm_lists@pulsar-zone.net>
To: gnats-bugs@netbsd.org, bouyer@antioche.eu.org
Cc: 
Subject: Re: pkg/32031: pkgsrc-2005Q3 problems under NetBSD 2.1
Date: Thu, 10 Nov 2005 15:05:11 -0500

 Hmm after your suggestion that my pkgsrc tree might be broken,
 I decided not to trust cvs not giving any message with a "cvs update -dP"
 and to download the pkgsrc tarball for 2005Q3 and cvs update again from
 there after making sure that CVS/Tags had -rpkgsrc-2005Q3 (it did).
 db4 still appears to fail the same way (it fails in rpc, it doesn't
 if I comment out the --enable-rpc CONFIGUE_ARGS though, except for the
 PLIST obviously).  The same problems as before are experienced with
 nspr, vim, cscope, etc.  And of course I tried without the CFLAGS, and 
 using the /bin/sh shell for both the session and default superuser shell.
 
 I am starting to wonder if I'm not stumbling on some gcc bugs at
 this point...  Although if I remember well I used that same gcc version
 on 2.0 more than a year ago the last time I successfully compiled
 those packages...  My whole system was compiled using -march=i686 -O2,
 including the compiler itself.  But this used to always be the case
 in the past (and had no problems with production servers under such a 
 setup).  However, compiler bugs do exist, such as toolchain/26456
 (although this one isn't related to optimization flags). 
 
 If it was because of a compiler bug, it could be any package which
 got broken such as sh, perl, etc which could perhaps fail some critial
 tests.  I have compiled a fair number of packages without problems, but
 it's odd to see that some which fail only seem to fail in my case...
 
 Consider this case with vim:
 
 [...]
 ===> Extracting for vim-6.3.085
 ===> Required installed package vim-share-6.3.085{,nb*}: vim-share-6.3.085 found
 ===> Patching for vim-6.3.085
 ===> Applying distribution patches for vim-6.3.085
 ===> Applying pkgsrc patches for vim-6.3.085
 ===> Overriding tools for vim-6.3.085
 ===> Creating toolchain wrappers for vim-6.3.085
 ===> Configuring for vim-6.3.085
 [...]
 
 So it didn't try to install ncurses at all, despite it using
 CONFIGURE_ARGS+= --with-tlib=ncurses
 .include "../../devel/ncurses/buildlink3.mk"
 
 So what's wrong, if it's actually supposed to install ncurses?
 (because it's definitely not on my system, not even as a package).
 I thought at first that perhaps the ncurses buildlink stuff verifies
 if the local curses can be used instead and does if so, but you
 suggested that ncurses should have been installed in this case.
 
 I have the following packages installed, if the versions can matter:
 (since several of these are used to build packages)
 
 digest-20050731
 pkg-config-0.19
 x11-links-0.26
 pkg_install-20050318
 pkg_tarup-1.6.5
 xpkgwedge-1.14
 libtool-base-1.5.18nb5
 
 
 So, I installed cscope from the NetBSD repository (so i386 version, as well
 as nspr).  As for db4, since it was only failing in the RPC module I
 disabled that module which I don't need for my applications, and it's fine.
 So I basically have all the packages which I needed after all the fiddling.
 
 As for what's wrong with my system, I have no idea yet, especially that the
 majority of packages compiled fine (most of approximately 400 packages).
 
 If I really am the only one experiencing such problems, you may close the
 PR if wanted, my next serious package building will not be before
 pkgsrc-2005Q4 release (and might be minimal depending on the amount of
 security fixes, my full packages rebuild is unlikely to happen until 12
 months from now)...
 
 
 Thanks,
 Matt
 
 
 -- 
 Note: Please only reply on the list, other mail is blocked by default.
 Private messages from your address can be allowed by first asking.