Bug#319136: [Pkg-shadow-devel] Bug#319136: passwd: config script silently fails if invalid group file is present

[Alexander Gattin]

tags 319136 pending
thanks

Hi!

On Fri, Jul 22, 2005 at 07:48:08AM +0200, Marc Haber wrote:
> > > the config script fails with exit code 1 (but without error message)
OK, your main concern, seems to be the          ^^^^^^^^^^^^^^^^^^^^^^^^^,
right?

Then I think I've found a reason -- the shadowconfig
util is run with stdout redirected to /dev/null _and_
because underlying utils (pwck/grpck) direct major
error messages to stdout, you see no error messages.

pwck/grpck use stdout for the mesages like these:
> cherokee:~# pwck -qr
> duplicate password entry
> 
> delete line
> `xxxxxx:x:444:444::/home/builder:/bin/bash'? No
because this is _conversation_ with user.

Tomasz prefers to leave this conversation at stdout,
as he told me today on IRC.

Maybe pwck/grpck could use stderr for final message
when there were errors found/fixed, but this should
be discussed with upstream, and patch developed.
I'll consider this later.

Right now running shadowconfig with stdout redirected
to /dev/null seems strange for me anyway, that's why
I replaced these calls in passwd.config

> > > if shadowconfig fails, for example when a syntax error is found in the
> > > group file.
> > 
> > But, err..., when there's a syntax error in
> > /etc/group or /etc/passwd file, the whole system can be
> > considered broken, at least partially. Couldn't it?
> 
> So have the program say so and then error out. Silently failing is not
> an option.

OK, sorry for my misunderstanding. %)

BTW I would prefer to have a sequence of actions to
reproduce the bug. ;)

AFAIU, you saw this during installing new system
(base-config) or running `dpkg-reconfigure passwd`
after deboostrap, yes?

> > what you mean here?
> 
> if ! shadowconfig bla; then
>   echo >&2 "ERR: shadowconfig failed"
>   exit 1
> fi
>
> Having the -e trip is something to be avoided.

In order for the above code to work, you mean?

-- 
WBR,
xrgtn