rustc_resolve: don't allow paths starting with ::crate.

[eddyb]

cc @aturon@joshtriplett
r? @petrochenkov

[eddyb]

cc @nikomatsakis If we go down this path, chalk at least needs to be updated.
AFAIK crate::... without a leading :: should work before this PR too.

[petrochenkov]

Implementation looks good to me, but I'd like to wait for other people's opinions.

I've seen an argument that ::crate may be useful in scenarios like

macro m($crate_name: ident){::$crate_name::something();}m!(my_crate);// OKm!(crate);// OK

[joshtriplett]

@petrochenkov Doesn't seem that hard to write m!(::my_crate) or m!(crate), in that case.

I don't have any fundamental objection to ::crate::foo working, but since crate:: is already unambiguous, it seems preferable to have one way to write it rather than two.

[petrochenkov]

@joshtriplett

Doesn't seem that hard to write m!(::my_crate) or m!(crate), in that case.

Yes, this should be possible when #48067 is implemented (paths cannot be easily concatenated in macros right now).

Anyway, prohibiting ::crate is more conservative variant so we can do it now and re-allow again if it becomes necessary.

[petrochenkov]

Do we have a label "ping @rust-lang/lang, wait for a few days and then r+ even if nobody responded"?

[cramertj]

We discussed this in a previous meeting and I got the impression that there was consensus around ::<cratename>::.. being the one and only way to use leading-:: in a path.

[Centril]

I second @cramertj. ::crate should not be permitted; or at the very least I'd want to be conservative here.

[joshtriplett]

This still needs its build failures fixed (s/::crate::/crate::/g everywhere), but as soon as that happens, let's merge this.

[eddyb]

@bors r=petrochenkov

[bors]

📌 Commit 9347bf78aaadf67967bd03b144bce9cf99850203 has been approved by petrochenkov

[eddyb]

@bors r=petrochenkov

[bors]

📌 Commit be90514154a4202ba27a0be50d64ff4fe193c61c has been approved by petrochenkov

[bors]

🔒 Merge conflict

This pull request and the master branch diverged in a way that cannot be automatically merged. Please rebase on top of the latest master branch, and let the reviewer approve again.

How do I rebase?

Assuming self is your fork and upstream is this repository, you can resolve the conflict following these steps:

1. git checkout no-crate-in-root(switch to your branch)
2. git fetch upstream master(retrieve the latest master)
3. git rebase upstream/master -p(rebase on top of it)
4. Follow the on-screen instruction to resolve conflicts (check git status if you got lost).
5. git push self no-crate-in-root --force-with-lease(update this PR)

You may also read Git Rebasing to Resolve Conflicts by Drew Blessing for a short tutorial.

Please avoid the "Resolve conflicts" button on GitHub. It uses git merge instead of git rebase which makes the PR commit history more difficult to read.

Sometimes step 4 will complete without asking for resolution. This is usually due to difference between how Cargo.lock conflict is handled during merge and rebase. This is normal, and you should still perform step 5 to update this PR.

Error message

warning: Cannot merge binary files: src/Cargo.lock (HEAD vs. heads/homu-tmp)
Auto-merging src/Cargo.lock
CONFLICT (content): Merge conflict in src/Cargo.lock
Automatic merge failed; fix conflicts and then commit the result.

[eddyb]

@bors r=petrochenkov

[bors]

📌 Commit 9b1d3c7 has been approved by petrochenkov

[rpjohnst]

Being conservative here is fine, especially now that :: has become far less necessary, but it seems more consistent to me to allow ::crate because crate plays the same role as crate_name.

[eddyb]

@rpjohnst The current interpretation isn't that, but rather that crate is a keyword just like self and super, that only appears at the start of a path, and changes it from the default mode (whichever it may be, Rust 2015 vs Rust 2015, with and without uniform_paths) to the root of the local crate.

[rpjohnst]

I get that interpretation, but the ::crate/"crate is a crate name" mental model is a great way to explain the new system, and it would be a shame to lose it.

[eddyb]

@rpjohnst Note that we can't make that work in Rust 2015, because there is no place (in the "module tree", so to speak) where you can just "access crates by their name".

In Rust 2018 that's the topmost path level, but in Rust 2015 it doesn't exist, and if we allow crate to resolve in the crate root (which is what ::crate means in Rust 2015), then self::crate, super::crate, etc. should resolve too, if the prefix of the path reaches the crate root, to remain consistent (unless we special-case it further).

[rpjohnst]

Yeah, 2015 does make it messier. But perhaps allowing crate to resolve in the crate root makes sense there? That's where all the extern crate declarations go anyway, and crate really only needs to make sense in 2015 insofar as it's part of the transition path to 2018. :P

We could also allow ::crateonly in 2018, by allowing it to resolve in the 2018 root, if that feels less messy. Or maybe it makes it worse via interaction with the new prelude-like mechanism, I'm not sure how that works.

Regardless, :: plays a smaller role in 2018, and that role doesn't intersect with crate, so maybe it doesn't matter to the mental model.