remove non-sysroot sources from rust-src component

[RalfJung]

See #69592 (comment): these were likely added in #58269 for the sake of compiler plugins, but those are being entirely phased out, so there is no good reason to ship these sources.

OTOH, @eddyb wrote

Yeah, my question is why librustc_plugin specifically? Everything else makes sense.

So maybe there is some good reason to keep these? Then we should have a comment explaining that reason.

Cc @eddyb @taeguk @Mark-Simulacrum

[rust-highfive]

r? @nikomatsakis

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

[eddyb]

Keep in mind that since then, libsyntax and librustc were split into a handful of crates, and something happened to librustc_plugin so it's possible this has been broken for plugin authors for a while.

When I wrote "Everything else makes sense." I probably imagined we could have most of src/ in rust-src but now that we have rustc-dev we should probably have a similar rustc-src component if we want to provide all of those libraries' source.

[RalfJung]

Yeah, having just the sources of these two crates without all their dependencies seems pretty useless. Hence this PR.

[Mark-Simulacrum]

I am in favor of removing these, particularly as I believe it's highly likely that they're not doing too much good since we've stripped so much out of them these days into other crates.

I think the best way to do so would be to land this change now-ish and then if people complain evaluate readding them (perhaps into the rustc-dev component?). I doubt that people are aware that they are using them if they are.

cc @matklad as well, in case IntelliJ or other IDE tooling wants these for some reason.

[matklad]

No, neither IntelliJ nor rust-analyzer needs librustc

[Mark-Simulacrum]

I'm going to approve this in that case. If someone feels we should get wider consensus, though, please feel free to unapprove.

@bors r+

[bors]

📌 Commit cbf52b1 has been approved by Mark-Simulacrum

[RalfJung]

@bors rollup