Minor improvements to bootstrap

[ljedrz]

• prefer Path-specific methods to String ones
• don't add file extensions if they are removed right afterwards

[rust-highfive]

r? @alexcrichton

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

[alexcrichton]

@bors: r+ rollup

Thanks!

[bors]

📌 Commit db303c1 has been approved by alexcrichton

[kennytm]

@bors r-

distcheck failed with

[00:11:29] tar (child): /checkout/obj/build/dist/rustc-1.29.tar.gz: Cannot open: No such file or directory
[00:11:29] tar (child): Error is not recoverable: exiting now
[00:11:29] tar: Child returned status 2

[ljedrz]

I investigated this and it appears that adding extensions with the PathBuf API might result in some weird results; this is totally expected:

let mut p1 = PathBuf::from(format!("{}.tar.gz", "rustc-1.29.0-dev-src"));
p1.set_extension(""); // removes .gz
p1.set_extension(""); // removes .tar

But this... less so:

let mut p2 = PathBuf::from("rustc-1.29.0-dev-src").with_extension("tar.gz");
p2.set_extension(""); // removes .gz
p2.set_extension(""); // removes .0-dev-src.tar !

In the second case I would either expect set_extension("") to first remove .gz and then .tar or remove .tar.gz with a single call, but I definitely didn't expect it to remove .gz and then the whole .0-dev-src.tar bit... Or am I missing something?

[kennytm]

@ljedrz Path::with_extension replaces the original extension:

use std::path::{Path, PathBuf};

let path = Path::new("foo.rs");
assert_eq!(path.with_extension("txt"), PathBuf::from("foo.txt"));

Therefore PathBuf::from("rustc-1.29.0-dev-src").with_extension("tar.gz") returned rustc-1.29.tar.gz.

[ljedrz]

In that case it seems too risky to use it in case the path might contain any dots; the added value isn't too high, as it's just the bootstrap; I'll close the PR, then.