Distribute rustc sources as part of rustc-dev

[jonas-schievink]

They can be used to provide IDE features when working on rustc plugins/backends/etc without having to locate a separate Rust checkout.

r? @Mark-Simulacrum

[Mark-Simulacrum]

One question I guess is whether these are ever useful without the 100+ MB of rlibs that rustc also ships. It might also be the case that if we're going to suggest this then you should be using the "whole source" vendored tarball...

I am inclined to say that those are perhaps interesting to think about but that we should just do this for now though.

[jonas-schievink]

One question I guess is whether these are ever useful without the 100+ MB of rlibs that rustc also ships. It might also be the case that if we're going to suggest this then you should be using the "whole source" vendored tarball...

Hmm, yeah. Since having just the compiler sources won't allow actually building a useful compiler (as you'd need the libstd sources too), I'm not sure what such a use case would look like. And having IDE support only seems useful when you can also link against the compiler crates.

[Mark-Simulacrum]

Well, that's not quite true -- you could build a compiler, because the sources are shipped with a nightly (or whatever channel) std that perfectly matches that compiler tree it should be fine to build the compiler src directly with that std.

But it certainly seems really unlikely that someone would want to do that, when pre-built binaries are available, and I'm not inclined to support that use case (yet at least).

I'm going to go ahead and @bors r+ rollup=never this -- it seems like we'd want to be sure that this is the cause if there are any regressions.

[bors]

📌 Commit ee1e934 has been approved by Mark-Simulacrum

[bors]

⌛ Testing commit ee1e934 with merge ac19c3b...

[bors]

☀️ Test successful - checks-actions, checks-azure
Approved by: Mark-Simulacrum
Pushing ac19c3b to master...