fix: Better handling with paths that include spaces

[hyperupcall]

Summary

I saw that there was a ASDF_BATS_SPACE_IN_PATH variable, which seemed to be part of a gradual transition to make the tests work better with whitespace stuff. Probably with help from #1433, it was somewhat trivial to make the proper fixes and remove this variable entirely.

One test was left out because it seems that change requires a fix in the .tool-versions parser (related to the path:~/src/elixir feature)

[jthegedus]

The changes I requested seemed to have introduced a bunch of shellcheck issues we would need to ignore. Are you able to resolve them @hyperupcall ?

[hyperupcall]

The changes I requested seemed to have introduced a bunch of shellcheck issues we would need to ignore. Are you able to resolve them @hyperupcall ?

I wish I mentioned this earlier, but the reason why the Bats test fails is because the newer changes moves the BASE_DIR=... inside of the test function. This means that it has no effect since BASE_DIR is used in setup, which runs before the test function.

[hyperupcall]

I reverted those changes, and the BASE_DIR is now set inside of setup(), so things should work. I know it's kind of ugly since the condition depends on the name of the test and there is a separation of sorts - do you think there is a better way or are the changes OK as they are?

My first thought would be to move the invocation of setup_asdf_dir from within setup(), to each individual function, just for the tests in shim_exec.bats, but it seems that would make all that code harder to read and more repetitive just for the sake of a single test.

I also tried re-calling setup (with the new BASE_DIR) from within the test function, but that doesn't seem to work either.

[jthegedus]

It is a shame we cannot opt-out of using the setup for a single test in the file. The most clean alternative I can think of is having a separate test file for these specific tests.

If someone thinks of something better, we can discuss then, in the meantime we will merge this.

[Stratus3D]

Is there really not a better way of doing this? This seems like quite a bit of code just to find and delete shims containing a certain line.

[hyperupcall]

There was, https://github.com/hyperupcall/asdf/blob/82b9ef01e26f39b375c1b462979aa7ffd069df8a/lib/commands/command-plugin-remove.bash#L24 but it didn't work on Alpine.

I suppose I could have replaced the nullglob stuff with a [ -f "$f" ] in the loop

[Stratus3D]

I guess I'd prefer [ -f "$f" ] in the loop as it feels harder to get wrong - it's easy for someone to forget to unset an option or reset it to the wrong value. In this case it looks correct so not an issue.