test(env): exclude _omp_*

[scop]

Ref https://ohmyposh.dev,
https://github.com/JanDeDobbeleer/oh-my-posh/blob/572c99b2ec6a57a18d4a911cf8d577805002811e/src/shell/scripts/omp.bash#L11-L18

[akinomyoga]

I'm curious about the situation where this matters during the test process. In my understanding, we initialize the Bash sessions in our test with test/config/bashrc, where we do not source external settings. Is there a case where OMP affects the test Bash sessions (e.g., through exported functions, BASH_ENV, ENV & POSIXLY_CORRECT, etc., though I'm not sure if they can actually be used to inject settings in the test sessions)? Or is there a future plan to load external settings during the test process?

[akinomyoga]

Also, to reduce the size of the output, I think environment variables are better to be filtered in _comp__test_get_env (test/config/bashrc):

bash-completion/test/config/bashrc

Line 79 in c4a8fd0

unset -v $(compgen -W '"${!_comp_@}"' -X '!_comp_@(_test_+([0-9])_*|*_mut_*)')

_, PPID, BASH_REMATCH, BASH_LINENO, and LINENO are excluded in conftest.py because they are special variables that cannot be unset. Technically, BASH_REMATCH can be unset in recent versions of Bash, but it can't in Bash < 5.1.

[scop]

Re the original question, TBH I did not research the "why" nor can I answer it offhand. It's just something I saw when doing a pytest test/t/unit/test_unit_compgen_filedir_xspec.py while having the branch related to that change checked out and oh-my-posh active, every time. I can't reproduce right now on main though, so drafting for now and will reopen if I see it again.

[akinomyoga]

OK, I know that apparently strange things sometimes happen. There could be a specific setup that OMP affects the test sessions, though I don't have an idea for now.