Bugzilla Bug 37402

Shadow suite does not allow a dot in usernames. This was originally allowed and
then changed by author of shadow to be denied by default, making it difficult
to migrate from older systems. This also seems to be allowed in other
distributions by default

Reproducible: Always
Steps to Reproduce:
1. Add user test.com
2.
3.

Actual Results:  
useradd: invalid user name 'test.com'

Expected Results:  
Should successfully add user

Portage 2.0.49-r20 (default-x86-1.4, gcc-3.2.3, glibc-2.3.2-r3, 2.4.24-1.9.13)
=================================================================
System uname: 2.4.24-1.9.13 i686 AMD Athlon(tm) Processor
Gentoo Base System version 1.4.3.10
ACCEPT_KEYWORDS="x86"
AUTOCLEAN="yes"
CFLAGS="-march=i686 -O3 -pipe"
CHOST="i686-pc-linux-gnu"
COMPILER="gcc3"
CONFIG_PROTECT="/etc /usr/kde/2/share/config /usr/kde/3/share/config
/usr/share/config /var/qmail/control"
CONFIG_PROTECT_MASK="/etc/gconf /etc/env.d"
CXXFLAGS="-march=i686 -O3 -pipe"
DISTDIR="/usr/portage/distfiles"
FEATURES="autoaddcvs ccache sandbox"
MAKEOPTS="-j2"
PKGDIR="/usr/portage/packages"
PORTAGE_TMPDIR="/var/tmp"
PORTDIR="/usr/portage"
PORTDIR_OVERLAY=""
USE="3dnow apm berkdb crypt foomaticdb gdbm gpm gtk2 libg++ libwww mad mmx
ncurses pam perl python readline slang spell ssl tcpd x86 xml2 zlib"

Created an attachment (id=23244) [details]
Patch to allow '.' in usernames

This is still happening in sys-apps/shadow-4.0.3-r9, has anyone looked at this?

With Gentoo reverting behavior of coreutils back to the previous behavior
allowing . in  username would be a security risk when used with chown etc...

Please don't add this patch to shadow.

from 'man chown'
 chown [options] user[:group] file...

the period is an extension that breaks POSIX, as they clearly admit to under the GNU DETAILS heading.  In fact, the only mention of using a period is in the man page, followed immediately by the statement that it is not portable do do so.  Maybe we should be patching coreutils as well to remove the period, or at least warn users that it is bad?

actually the latest coreutils had to be patched to add the . back in :P
same with the stupid `head -#` syntax

i complained but no one seemed to listen ... maybe we should file a bug about it

I'm fine with breaking POSIX standards here.

If we have to patch coreutils to add the . back to chown because GNU deprecated
it, and it breaks posix to have it, is there an upside that I'm missing?
They've cleaned up several parts of coreutils lately (see
http://bugs.gentoo.org/show_bug.cgi?id=35634 for another example)

way to not be friendly to the future nedd

coreutils posix issues is now at Bug 39632

personally, i suggest we close this bug as WONTFIX/UPSTREAM

Going by the information in the above bug, it sounds like we will shortly end
up with shadow doing one thing and coreutils doing another...

Of course they will. 
shadow will continue it's current behavior of disallowing '.' and coreutils will issue a warning. This warning should stay around for 3-6 months.

It's not the chown one that kills us as much as it's the head/tail problems with a 
ports tree of this size it sure causes alot of errors. We have attempted to work 
around those errors with the portage fixheadtails.eclass but that was far from 
perfect which is why I assume the non POSIX standard has returned to portage.

hi sorry to be late in piping in -- spanks -- I do actually have that on my
todo list to issue the warning about . -- however, I've not yet gotten around
to actually doing it -- if someone has the time and/or inclination to write me
a patch for coreutils which issues a warning about the . syntax, I'm happy to
add it.  As for the patch to revert to old behaviour -- yes that was added back
in because it proved to a be a HUGE pain in the ass to move to the new
behaviour just like that.  There are packages (like ispell) whose make process
relies on the old behaviour, etc etc (in ispell's case, I think also its
runtime behaviour), so the problem is not one that can be solved overnight. 
Additionally, the patch I'd submitted to ispell's author was greeted with
disdain -- the idea in the chap's response was that the old behaviour works on
every other bloody unix, why change it? -- don't shoot me, I'm just the
messenger.

Anyway, solar's right, the new behaviour being suppressed as the default is
intentional, at least for now and the next few months while we attempt to get
everything else fix0red up and patched.  Since coreutils-5.0.91-r4 recognises
both new and old behaviour, we're ok for the moment.  Sorry about this.

Hello, i reported a few months ago the probleme in this thread. You can find a
clean patch to solve the problem in it.

Thanks in advance.

The thread : http://bugs.gentoo.org/show_bug.cgi?id=22920

*** This bug has been marked as a duplicate of 22920 ***