Unresolved import for declarative macros 2.0

[fmease]

Since #6016. Importing a declarative macro 2.0 (an experimental feature) is falsely flagged as unresolved.

Minimal example to reproduce:

#![feature(decl_macro)]

use foo as _; // error: unresolved import
macro foo() {}

fn main() {}

If I am not mistaken, RA itself already handles declarative macros 2.0 in some places and it is in fact not rustc which resolves them:

#![feature(decl_macro)]

foo!(); // RA's name resolver can handle this
macro foo() {}

fn main() {}

[fmease]

This is probably low priority for most people but I can try to give it a go in a week, once I have spare time, unless someone else would like to fix this.

[jonas-schievink]

Yeah we don't support them at the moment. We have some hacks in place so that builtin macros from the standard library still work.

[fmease]

If I may ask, do you think there is enough information in the code base already that improving support is relatively easy? Is it even acceptable to implement for RA given it's an unstable/experimental feature?

At first glance, it seems like ModCollector::collect is the place to start, am I correct? Extending it to handle a new ModItem::Macro and going from there.

[jonas-schievink]

Yeah it should be fine to add support for them

[fmease]

FYI, the overarching issue is #2248. However in my opinion, this issue (#6059) should not be closed in favor of the other one since it is specifically about error messages which should not be reported and which weren't reported before.

[matklad]

works on master