Bugzilla Bug 127157

User-Agent:       Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.8.0.1)
Gecko/20060201 Firefox/1.5.0.1
Build Identifier: 

Motivation: I installed Gentoo on several machines already, including x86,
amd64 and sparc flavors. This time I wanted to build as a test an ~x86 system
with gcc-4.1, just for fun ;-)

During "emerge -e system" the process stopped on openssh, stating that it was
unable to find the 'groupadd' command (a pretty basic one). Retrying the build
process using "emerge --resume" always breaks at the same place.


Reproducible: Always

Steps to Reproduce:
1. prepare and mount partitions like described in the Handbook
2. unpack stage1 tarball
3. configure /etc/fstab (I always do this before chrooting for the first time)
4. chroot /mnt/gentoo /bin/bash
5. mount an additional partition on /usr/portage
6. emerge --sync
7. put USE="unicode nptl nptlonly userlocales" in /etc/make.conf
   add ACCEPT_KEYWORDS="~x86" and configure /etc/locales.build
8. unmask "=sys-devel/gcc-4.1*" in /etc/portage/package.unmask
   (to avoid switching gcc profiles)
9. cd /usr/portage
   scripts/bootstrap.sh

Here the stage2 is built wihout problems. So far, so good :-)

10. set USE="unicode nptl nptlonly userlocales pic ithreads bzip2 justify"
    in /etc/make.conf
11. emerge -ave system

Actual Results:  
While progressing from stage2 to stage3 compilation of openssh goes smoothly.
But while installing binaries, ebuild stops with the following:

>>> Original instance of package unmerged safely.
 * Adding group 'sshd' to your system ...
 *  - Groupid: 22
/usr/portage/eclass/eutils.eclass: line 728: groupadd: command not found

!!! ERROR: net-misc/openssh-4.3_p2-r1 failed.
Call stack:
  ebuild.sh, line 1546:   Called pkg_postinst
  openssh-4.3_p2-r1.ebuild, line 148:   Called enewgroup 'sshd' '22'

!!! enewgroup failed
!!! If you need support, post the topmost build error, and the call stack if
relevant.

!!! FAILED postinst: 1


Expected Results:  
The openssh ebuild should be able to find 'groupadd' command,
so enewgroup 'sshd' '22' wouldn't fail.


I checked what other packages should be built:

# emerge -pv --resume

These are the packages that would be merged, in order:

[ebuild  N    ] net-misc/openssh-4.3_p2-r1  USE="ipv6 pam tcpd -X509 -chroot
-hpn -kerberos -ldap -libedit -sftplogging -skey -smartcard -static" 0 kB
[ebuild  N    ] sys-apps/shadow-4.0.14-r3  USE="nls pam -nousuid -skey" 0 kB
[ebuild  N    ] sys-process/procps-3.2.6  0 kB
[ebuild  N    ] sys-apps/module-init-tools-3.2.2  165 kB
[ebuild  N    ] sys-apps/gawk-3.1.5  USE="nls -build" 2,256 kB
[ebuild  N    ] sys-apps/kbd-1.12-r6  USE="nls" 0 kB
[ebuild  N    ] net-misc/wget-1.10.2  USE="ipv6* nls* ssl* -build* -debug
-socks5 -static" 4 kB
[ebuild  N    ] sys-apps/file-4.17-r1  USE="python -build" 543 kB
[ebuild  N    ] sys-apps/net-tools-1.60-r12  USE="nls -static" 102 kB
[ebuild  N    ] app-editors/nano-1.3.10-r1  USE="justify ncurses nls spell
unicode -build -debug -minimal -slang" 1,139 kB
[ebuild  N    ] net-misc/iputils-021109-r3  USE="ipv6 -doc -static" 0 kB
[ebuild  N    ] sys-apps/man-pages-2.25  USE="nls" 1,719 kB
[ebuild  N    ] app-i18n/man-pages-pl-20040401  0 kB

Total size of downloads: 5,931 kB

And the conclusion is that if the 'shadow' package was built before 'openssh'
there should be no problems at all. Just wanted to let you know about it.

I've checked and found a big fat warning about the 'shadow' package in the
/usr/portage/profiles/default-linux/packages.build file, so I'm almost sure
that this is the behaviour that the author of this warning was referring to.
Anyway, I thought that this may be interesting at least to the releng team.

I know several methods of how to deal with that situation, so no real
assistance is needed. :-) I'm using Gentoo for almost three years already
and occasional breakage is not a problem anymore. ;-)

Friendly,

Created an attachment (id=82850) [details]
output of 'emerge --info'

Just in case if someone was interested...

I had the same when I build a glibc 2.4/gcc 4.1 system from 2006.0 stage1

I suppose openssh should have an explicite (build time) depend on shadow.

(In reply to comment #0)

same issue.

*** This bug has been marked as a duplicate of 113298 ***