Re: [musl] getusershell should ignore comments and empty lines.

[Rich Felker <dalias@libc.org>]

On Fri, May 17, 2024 at 08:17:54PM -0700, Collin Funk wrote:
> Hello Musl maintainers,
> 
> The getusershell function behaves differently from Glibc and FreeBSD.
> I believe that it should follow those implementations by ignoring
> comments and empty lines.
> 
> I wrote a test for Gnulib that catches this issue. You may find it
> helpful for testing [1].
> 
> On FreeBSD I have the following etc/shells:
> 
> ==============================================
> # List of acceptable shells for chpass(1).
> # ftpd(8) will not allow users to connect who are not using
> # one of these shells.
> 
> 
> 
> /bin/sh
> /bin/csh
> /bin/tcsh
> /usr/local/bin/bash
> /usr/local/bin/rbash
> /usr/local/libexec/git-core/git-shell
> ==============================================
> 
> And I run the following in a Gnulib checkout:
> 
>   $ rm -rf testdir1 && ./gnulib-tool --create-testdir --dir testdir1 getusershell
>   $ cd testdir1
>   $ ./configure
>   $ make
>   $ ./gltests/test-getusershell
>   /bin/sh
>   /bin/csh
>   /bin/tcsh
>   /usr/local/bin/bash
>   /usr/local/bin/rbash
>   /usr/local/libexec/git-core/git-shell
> 
> GNU libc behaves the same way. I have not checked the other BSDs but I
> assume they use the same code derived from 4.3BSD or 4.4BSD.
> 
> Using an Alpine Linux virtual machine with Musl Version 1.2.4_git20230717
> and a few packages installed I have the default /etc/shells:
> 
> ==============================================
> # valid login shells
> /bin/sh
> /bin/ash
> /bin/bash
> ==============================================
> 
> Using the same commands listed earlier I run:
> 
>    $ ./gltests/test-getusershell 
>    test-getusershell.c:54: assertion 'ptr[0] != '#'' failed
>    Aborted
> 
> And after adding an empty line before the comment:
> 
>   $ ./gltests/test-getusershell 
>   test-getusershell.c:55: assertion 'ptr[0] != '\0'' failed
>   Aborted
> 
> Let me know if you have any questions. The FreeBSD shells(5) man page
> is pretty good and might be helpful [2]. Here is a link to their
> implementation incase that helps too [3].

It says:

    "A hash mark (``#'') indicates the beginning of a comment;
    subsequent characters up to the end of the line are not
    interpreted by the routines which search the file."

This isn't very clear whether # is only a comment on the beginning of
a line (after potential whitespace?) or whether # appearing in a line
with a shell pathname is a comment or part of the pathname. If it's a
comment, it's not clear if whitespace before it is part of the shell
pathname -- e.g. does "/bin/sh # best shell" define "/bin/sh" or
"/bin/sh " as the shell pathname?

It sounds like nobody ever thought about whitespace, quoting, or
rigorous comment syntax here...

Rich