populate parent directories of process outputs locally to match remote execution api

[cosmicexplorer]

Problem

Resolves #7084.

Solution

straightforward

Result

We now match this corner of the remote execution API in local process executions in the v2 engine.

[illicitonion]

You should be able to test this in the test module lower down in this file, by trying to run something like:

touch foo/bar

and seeing that it succeeds because foo exists (whereas before this change, it would fail because file not found)

[cosmicexplorer]

You should be able to test this

Thanks! I have done so in 2ab4633, wasn't thinking too hard last night. Also had to fix a couple other tests with now-unnecessary mkdirs!

[illicitonion]

Thanks! One small restructure and I think we're good to go :)

[Eric-Arellano]

I'm not sure how this relates to #7048. Did you mean a different issue number in the problem section?

[cosmicexplorer]

Thanks, changed to #7084! Typical 2am hack sesh...

[cosmicexplorer]

Thanks! One small restructure and I think we're good to go :)

@illicitonion Sorry, could you describe the restructure necessary? I can't seem to find it in the PR comments, my eyes could be missing something.

[illicitonion]

Ugh, github dropped my comment somehow :(

[illicitonion]

I think this would be more clear as:

let parent_paths_to_create: HashSet<_> = stuff.collect();
for path in parent_paths_to_create {
  create_dir_all(&path).map_err(...)?;
}
Ok(())

rather than using a chain of two iter-collect pairs.

[cosmicexplorer]

github lost my reply somehow! I wanted to say thank you because as much as I love .collect() this is significantly easier to write and read and I wasn't really remembering the utility of ?!

[illicitonion]

LGTM - thanks!