Make run-pass/env-home-dir.rs test more robust

[malbarbo]

Remove the assumption that home_dir always returns Some.

This allows the test to be executed with cross.

[rust-highfive]

r? @nikomatsakis

(rust_highfive has picked a reviewer for you, use r? to override)

[nikomatsakis]

Why did you change to now construct a relative path instead of some random string?

It seems like before we never even looked at the value we got back, is that right?

[malbarbo]

I could have only changed the line that assume the home_dir returns Some, that is

- assert!(home_dir().is_some());+ assert_ne!(home_dir(), Some(PathBuf::from("/home/MountainView")));

but that could fail if getpwuid_r returns "/home/MountainView" (or other chosen string). Setting newhome to a relative path avoids this situation.

[nikomatsakis]

You're saying that could fail if set_var had no effect, but it just happens that the directory name is /home/MountainView? I suppose that's true. It just seems a bit weird for the windows test (below) to be asymmetric.

[nikomatsakis]

I think I don't quite understand what is being fixed here =)

We're still asserting that home_dir returns Some -- how was the test failing before?

[nikomatsakis]

(Anyway, the patch seems fine, just trying to understand. =)

[malbarbo]

You're saying that could fail if set_var had no effect, but it just happens that the directory name is /home/MountainView? I suppose that's true. It just seems a bit weird for the windows test (below) to be asymmetric.

I changed the unix case because that's was failing for me, but in the end, I just need that the test does not assume home_dir return Some.

We're still asserting that home_dir returns Some -- how was the test failing before?

The test was failing because home_dir was returning None. Now we are using assert_ne!, so it pass if home_dir returns None.

[nikomatsakis]

Ah, I missed the assert_ne thing somehow. Can you add a comment on top saying something like:

/// When HOME is not set, some platforms return `None`, but others return `Some` with a default.
/// Just check that it is not the `newhome_path`.

Also, I think it'd be best if the unix + windows tests match up -- I'm ok with either your new version or the old one, but I guess it's easier to just revert those commits.

[malbarbo]

Updated.

[nikomatsakis]

@bors r+ -- thanks!

[bors]

📌 Commit adeb0ae has been approved by nikomatsakis

[kennytm]

@bors rollup