[Libguestfs] shebang usage in bash scripts

Eric Blake eblake at redhat.com
Fri Aug 26 12:29:34 UTC 2022


On Thu, Aug 25, 2022 at 02:35:47PM +0100, Richard W.M. Jones wrote:
> On Thu, Aug 25, 2022 at 03:32:15PM +0200, Laszlo Ersek wrote:
> > Hi,
> > 
> > most shell scripts in the v2v projects start with a shebang like this:
> > 
> > #!/bin/bash -
> > 
> > I *think* I understand the intent of the single hyphen, but (a) it seems unnecessary, (b) even if we insisted, using the double-hyphen separator "--" is much more idiomatic (even though the shell, per POSIX, is supposed to interpret "-" identically to "--").
> > 
> > Regarding why I think the hyphen is unnecessary:
> > 
> > - setuid shell scripts are not a thing on any platform we (should) care about
> > 
> > - the script name is not reinterpreted as an option *anyway*
> > 
> > Consider:
> > 
> > cat >-v <<EOT
> > #!/bin/bash
> > echo hello \$1
> > EOT
> > 
> > chmod +x -- -v
> > 
> > PATH=$PWD:$PATH -v world

More believable as:

(PATH=$PWD:$PATH; -v world)

In your style, you are relying on the shell using the just-modified
PATH setting to look up the location of -v; but POSIX allows both a
system where the shell does its own lookup (and sticks to the old
PATH, only passing the new PATH to the child process of the -v it
found on the old PATH), as well as a system where the shell forks off
a subprocess, the subprocess modifies environ to set the new PATH, and
then relies on execvp() or similar to do the lookup on what is already
the new environ.

But that's a distraction from your main point, that yes, Linux injects
"--" to the interpreter for a shebang, to protect against an
executable script whose name begins with '-'.

> > 
> > So glibc and/or the Linux kernel already inserts the "--" option/operand delimiter!

Shebang interpretation is sometimes by the kernel, and sometimes by
the shell, and it differs by OS.  Some OS's do word-splitting on the
rest of the shebang after the first space, some treat the entire rest
of the line as a single argument.

For a nice summary of some of the warts of shebang lines, see

https://unix.stackexchange.com/questions/29608/why-is-it-better-to-use-usr-bin-env-name-instead-of-path-to-name-as-my/29620#29620

If you are worried about /bin/bash being available on all systems
where a v2v script will be installed, you probably want '#!/bin/env
bash' to find the first bash in PATH.  But once you add your first
whitespace, it is no longer portable to use any others, so while
'#!/bin/bash -' works, '#!/bin/env bash -' does not.  I find that the
issue of writing a shebang to use a program that may not always be
installed in the same absolute location across all OSs (and therefore
where '#!/bin/env name' is useful) is more common than the issue of
writing a shebang that wants to pass an explicit '-' or '--'
end-of-option marker to protect against the script being installed
under a name like '-c ls'.

Note that GNU coreutils recently added '/bin/env -S', modeled after a
similar feature first in BSD systems, which makes it possible to write
'#!/bin/env -S app --with args' parsed in a more similar manner
regardless of whether the OS pre-splits the line or lumps it in one
argument, but '#!/bin/env -S bash -' is also unlikely to work portably
as 'env -S' is not yet in widespread usage.

> > 
> > I intend to contribute a shell script to virt-p2v; do I need to use the hyphen in the shebang? (If so, I prefer the double-hyphen.)
> 
> Eric should be able to give the definitive answer here.

POSIX intentionally does not standardize #!, so the definitive answer
is that there isn't one ;) But unless we plan on catering to someone
sym- or hard-linking an alternative name for your script that begins
with '-' (which I find overkill - no one sane does that), I think
omitting the '-' in the #! line is the best course of action, because
the only thing it does is add a layer of protection on non-Linux
machines against someone creating a stupid alternative name for the
script, which is unlikely to happen.  And omitting the '-' frees us up
to insert /bin/env if we are worried about /bin/bash being
non-portable.

-- 
Eric Blake, Principal Software Engineer
Red Hat, Inc.           +1-919-301-3266
Virtualization:  qemu.org | libvirt.org


More information about the Libguestfs mailing list