Disagreement between x.py tidy and rustfmt

[Lucretiel]

rust/src/libstd/io/buffered.rs

Lines 1136 to 1140 in 9d09331

	impl Read for ShortReader {
	fn read(&mut self, _: &mut [u8]) -> io::Result<usize> {
	if self.lengths.is_empty() { Ok(0) } else { Ok(self.lengths.remove(0)) }
	}
	}

rustfmt and x.py fmt disagree about the correct formatting of this block. This causes issues with editors configured to "rustfmt-on-save", because you have to keep manually reverting / running tidy. You can see this in the commit log for #72808, which has many commits that unintentionally / accidentally bounce between the two formats:

1bf8ba3
e89e2e4
8df5ae0
f7650fe

[lcnr]

Isn't this somewhat expected (if unfortunate) as ./x.py fmt is pinned to a specific version of rustfmt and rustfmt has
some slightly changes between different versions?

[Nemo157]

Specifically see

rust/src/stage0.txt

Lines 19 to 23 in 9d09331

	# We use a nightly rustfmt to format the source because it solves some
	# bootstrapping issues with use of new syntax in this repo. If you're looking at
	# the beta/stable branch, this key should be omitted, as we don't want to depend
	# on rustfmt from nightly there.
	rustfmt: nightly-2020-04-22

Unfortunately it's pinned to a specific nightly rather than using the previous stable like rustc/cargo do, so using something like a rust-toolchain file to pick the toolchain and integrate with rustup wouldn't fix this problem.

[jyn514]

This has now been documented in rust-lang/rustc-dev-guide#883. @Lucretiel do you think that's good enough to close the issue?

[Lucretiel]

I'll give it a try and see if it fixes

[Lucretiel]

Yes, that's fixed it, thank you!