Is there any way to write a test for this?

In GitoxideLabs/gitoxide@51bbb86#diff-b1f4fc8d8a8a51923a922a937745a1f5665443821b5c4df58efa44c7c4515216 there is a check if the file returned by path() exists. But that isn't really a useful test.

Maybe there could be some sort of integration test that creates a git-repo with hooks that then runs run_hook. But that wouldn't really help as that would only test one setup and this is about making it work on as many setups as possible.

Maybe there could be some sort of integration test that creates a git-repo with hooks that then runs run_hook

we do exactly that in the hooks unittests

running the tests with SHELL=false or something like that, would check that the SHELL variable is ignored, but I'm not sure if that's helpful

Why is -l added to the shell args?

git only adds -c /~https://github.com/git/git/blob/757161efcca150a9a96b312d9e780a071e601a03/run-command.c#L294-L306 and if i see that correctly it doesn't use string concatenation to provide arguments to the script.

I think on non-Windows platforms it's best to just execute the hooks directly. I don't see a reason to use a shell.