Interesting idea. I feel a little weird about committing a tmp dir though. @enebo?

I'll take another look... I didn't look into generating the tmp dir as needed for some reason.

I would almost prefer we just moved this to a build dir somewhere and probably leave them named as-is? Now that we use maven we no longer use build but having yet another scratch dir seems less appealing (even though it is more appealing than n tmp_* files).

Also what is that .gitignore line for?

the .gitignore is so that tmp dir gets saved to the repo. Moving the tmp_* directories to a build dir is equivalent for my purposes... I was just tired of having to manually delete them since they don't clean up properly with certain test failures.

@tduehr Wow I did not notice that is tmp/.gitignore nor know you could have sub directory .gitignores!!!

yup, the common idiom for keeping a directory around is either a .gitkeep file which does nothing but it's content and therefore can be tracked by git. Or add a .gitignore which has the added bonus of being able to keep the contents of the directory out of the repo without having to ignore the directory itself. Unsurprisingly .gitkeep has fallen out of favor.

Has there been any decision on this one?

I still don't like the idea of having tmp committed to the repository with nothing but .gitignore in it. We should just fix the test to use a system-level temp location. That would be good enough, right?

I agree that would be better.

There was an additional problem with the test: it used rm_f instead of rm_rf to delete the @tmpdir in teardown. In other words, it didn't actually delete it. Both issues are fixed now.