Cleaning of shell scripts

[ElectreAAS]

Shellcheck!

This PR includes several rounds of applying shellcheck to all the scripts in the repo.

Commit details

The first commit (faa02b2) was done by hand, and only adds shebangs to scripts that missed them.

The second & third commits (8273b4f & f3bfaaa) are the result of letting shellcheck auto-fix the problems it found. I still reviewed all changes (and removed some of them) by hand before committing.

The fourth commit (395322d) was done by hand again, applying changes that are potentially breaking. By this I mean that the previous behaviour was ambiguous / using a bespoke feature, and I clarified it, but that can mean misanderstanding.

Special notes

Most changes present here have to do with preventing unexpected globbing and word-splitting, which are classic problems in scripts ("oh this thing breaks when there are spaces in filenames, how odd!"). Some have to do with POSIX sh not defining certain things well or at all.

One change that is notably absent is eval $(opam env) which triggers a warning from shellcheck, but I'm pretty sure the 'weird' behaviour is indented.

Additional notes & context for reviewers

I came across some weird-looking scripts while trying to pinpoint the root cause of #5761. I figured it'd be easier to write correct fish scripts if I knew what the bash equivalents were, but my editor kept saying that everything was ambiguous.

I'll also note that I did not change src/state/shellscripts/prompt.sh at all because I just don't understand it. A comment there says "this script should be portable across all shells in common use.", but fish doesn't use PS1, and the script itself uses local variables that are undefined in POSIX sh.
If that's of interest to maintainers, I could write a prompt.fish to palliate this.

[ElectreAAS]

Note: the only "completely wrong" thing I encountered is this line in sandbox_exec.sh, where the argument to -z is always false due to literal strings. If this was intended to check if the directory existed (my best guess), I can add a fix for it in this PR or make a separate one, as it's a bugfix and not a refactoring.

[kit-ty-kate]

looks good to me overall on quick first glance aside from adding the shebang to the files in src/state/shellscripts/* which i feel a bit icky about are those files are sourced directly anyway and in some cases it's not clear which shell is to be used or even where their locations is (it could be /usr/local/bin/bash on BSDs for example)

[kit-ty-kate]

Note: the only "completely wrong" thing I encountered is this line in sandbox_exec.sh, where the argument to -z is always false due to literal strings. If this was intended to check if the directory existed (my best guess), I can add a fix for it in this PR or make a separate one, as it's a bugfix and not a refactoring.

If you can open a separate PR for that, that would be perfect indeed. Thanks a lot!

[ElectreAAS]

If you can open a separate PR for that, that would be perfect indeed. Thanks a lot!

-> #5780

[ElectreAAS]

in some cases it's not clear which shell is to be used or even where their locations is (it could be /usr/local/bin/bash on BSDs for example)

I agree and could change offenders from /bin/bash to /usr/bin/env bash. This still isn't 100% portable (at least according to SO user CodeClown42), but it's probably better for BSDs as you mention.

In the end I agree with you, and reverted these changes in 735cdc1