please consider giving a dedicated exit value for "invalid user name"

[zeha]

Debian has a wrapper around useradd called adduser, which controls a lot of the default policy on Debian etc. It also has its own idea on which usernames are allowable. The adduser maintainer requested the following feature, IMO a sensible request.
I hope upstream is open for this feature.

Quoting the feature request in full:

recently, useradd has grown funcitonality to check user names for
validity, see #1074306 [1]. This kind of clashes with adduser doing the same
thing, and has made testing adduser a bit harder. But still, the efforts
of staying sane upsetream-wise and close to upstream Debian-wise are
appreciated.

I would like to have adduser give more clear error messages like "the
user name is fine with adduser but useradd doesnt like it", and that
would be way easier if useradd's error message "invaid user name" woud
be backed with an appropriate, exclusive exit code.

I understand that this is unlikely to show up in useradd next week, so I
will be tweaking adduser's tests to handle those failures more
graefully, but I have just commented the tests so that we can re-enable
them once there is a more reliable method to detect this kind of useradd
error programatically. Please keep this bug report posted about your and
upstream's decisions.

[1] https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1074306

[alejandro-colomar]

I think adduser(8) should not attempt to check the validity of names. Why not patch useradd(8) to do the checks you want?

[zeha]

I think adduser(8) should not attempt to check the validity of names. Why not patch useradd(8) to do the checks you want?

adduser should still be able to tell somehow that useradd deemed the name to be bad. A dedicated exit code would be a nice way to indicate that.

[alejandro-colomar]

adduser should still be able to tell somehow that useradd deemed the name to be bad.

Could you clarify why adduser(8) needs to be able to tell that useradd(8) deemed the name too bad? Why is it not enough to let useradd(8) fail and print to stderr, and then just fail from adduser(8) too?

What does adduser(8) do different when useradd(8) fails due to ENOMEM, from when it fails due to badname?

alx@devuan:~/src/shadow/shadow/master$ ./src/useradd /
useradd: invalid user name '/': use --badname to ignore
alx@devuan:~/src/shadow/shadow/master$ echo $?
3

[Zugschlus]

adduser might want to be friendly and give a friendly error message. "use --badname" is misleading since adduser's options are named differently.

[alejandro-colomar]

adduser might want to be friendly and give a friendly error message. "use --badname" is misleading since adduser's options are named differently.

How about adding a --badname option for compatibility, as an alias to whatever option you have?

[Zugschlus]

Thank you very much for considering this on your side. Do you want me to contribute docs? At least the man page needs adaption.

[alejandro-colomar]

Thank you very much for considering this on your side.

You're welcome. :-)

Do you want me to contribute docs? At least the man page needs adaption.

Sure, feel free to send a patch/PR to my branch here: https://github.com/alejandro-colomar/shadow/tree/badname. I can then squash it in the PR (crediting you, of course).

[zeha]

@Zugschlus I assume you'll take it from here, ok?

[Zugschlus]

@Zugschlus I assume you'll take it from here, ok?

Depends on what else there is to do.

[zeha]

@Zugschlus I assume you'll take it from here, ok?

Depends on what else there is to do.

You said something about contributing docs. I guess you'll check if the implementation does what you wanted, incl. for groupadd. I'll deal with the Debian side once it's in master, or when we want to run with a draft patch. Feel free to ping me tho!