Idiomatic way to check whether `cygpath` exists in the test.sh scripts

[Rangi42]

fa22b78 switched from using which cygpath to using type -t cygpath, due to which being non-conformant on "some platforms" Mac OS X 10.4 Tiger and returning 0 even if the command was not found. It might be preferable to use which -s cygpath ("No output, just return 0 if all of the executables are found, or 1 if some were not found") or command -v cygpath (that's used by pokecrystal's Makefile to check whether sha1sum exists). (I think we should also comment in the test.sh scripts why we're using whichever particular approach we decide on, since future maintainers might not realize there are any differences besides style.)

[Rangi42]

According to sources like https://unix.stackexchange.com/questions/85249/why-not-use-which-what-to-use-then, we should avoid which and use command -v. We do want to identify cygpath even if it exists as an alias, because if the user did alias cygpath=/some/where/else for some reason and manually ran ./test.sh, that should Just Work.

[ISSOtm]

Bash's documentation appears to imply that command -v would ignore functions, which made me prefer type; but, actually testing it (at least with Bash 5.2) showed that it doesn't, which conforms to POSIX.1-2024 (even without POSIXLY_CORRECT). So we can probably safely rely on it.

[Rangi42]

We "probably" definitely can rely on it, since POSIX command doc confirms it handles "shell functions", and bash's command obeys that (whether or not POSIXLY_CORRECT is defined, i.e. non-POSIX bash doesn't somehow break command -v).

[quinnyo]

Hi, I have opinions/thoughts/notes on this, hopefully this is helpful!

The bash documentation is correct in that command asdf will not call the shell function asdf. command -v asdf does return success if asdf is (only) defined as a shell function.

Either of type or command -v will work, but I prefer type for this use case:
The singular purpose of the type builtin is to provide information about (potential) commands. The only reason that type is in a script is as a query about a command name.
Whereas command is firstly a means to direct the shell to interpret a line in a particular way, and only with the -v/-V option does it act like type. So command is less clear, although it's a very well known utility anyway.

Because it's a bash script, there's no particular reason to prefer a POSIX solution. POSIX compliance is there for POSIX compliant scripts to run on, but this isn't a POSIX compliant script.

(You don't need the -t option (type -t) but that's not a big deal.)