Hi lsb-spec lsb-test and debian-lsb, Tonight I encountered a 3rd party application with a post-install script that was calling useradd with a '-r' option. This broke on Debian because it does not support that option.
Spec ==== Checking the LSB spec, http://www.linuxbase.org/spec//booksets/LSB-Core-generic/LSB-Core-generic/us eradd.html we see that, -r creates a system account, that is, a user with a User ID in the range reserved for system account users. If there is not a User ID free in the reserved range the command will fail. So the LSB specifies it. * Debian's useradd is provided by their 'passwd' binary package (4.0.12-6) which comes from the 'shadow' source package, it does not support '-r' (and there's no bug in the BTS). The source package is based on the upstream source available at, ftp://ftp.pld.org.pl/software/shadow/ * The latest version of that upstream source (4.0.13) does not support the '-r' flag. * Red Hat rhel4's useradd is provided by the shadow-utils binary/source package and supports '-r'. That source package is based on the same upstream source, but forked from a much older version, 4.0.3 released in March 2002. So Red Hat added the '-r' to their version but either did not contribute it upstream, or upstream did not accept it and Red Hat kept it, effectively forking from upstream. I could not find the source for the package (they tend to make that hard due to their business model for rhel). # Fedora is newer, based on 4.0.7, and also supports '-r'. I found the shadow-utils source package and it adds '-r' support via a patch named 'shadow-4.0.7-redhat.patch' applied to the upstream tarball as part of the package. * SuSE sles9's useradd is their own creation, from the pwdutils package written by Thorsten Kukuk. It supports '-r'. Had it been in place at the time, the LSB criteria should not have prevented the addition of this option since it wasn't universally available in the distros and from the canonical upstream provider. But it's in now. If upstream can be convinced to add it, I'm sure Debian and anyone else missing it can adopt it and then it can stay in the LSB. Otherwise it needs to go. I don't think this is a fixable bug in Debian until upstream is willing to accept it. Testing ======= It occurred to me I've never seen this fail in my LSB testing. It looks like the tests don't cover '-r'. I assume it should be in tests/lsb-runtime-test/usersgroups/tset/LSB.usersgroups/commands/misc/comm and_tests.sh in cvs at, http://cvs.gforge.freestandards.org/cgi-bin/cvsweb.cgi/%7echeckout%7e/test s/lsb-runtime-test/modules/usersgroups/tset/LSB.usersgroups/commands/misc/co mmand_tests.sh?rev=1.28&contenttype=text/plain&cvsroot=lsb So the tests need to be improved. I don't think any new test should be used for certification until the above spec issues are resolved. FYI - This same post-install script used useradd with '-M' as well, but that's not in the LSB spec. So that's the developer's fault for using a non-portable option, but maybe '-M' should be considered (assuming it meets the LSB criteria). Thanks, -- Matt Taggart Open Source & Linux Organization R&D [EMAIL PROTECTED] Hewlett-Packard -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
