I"m a little wary of piling too much information onto build script environment variables. Right now we added exposing RUSTC unconditionally as the compiler being used, would that be enough?

Link to the crate for the lazy: https://crates.io/crates/rustc_version

@alexcrichton that"s what I"m using currently to shell out to rustc. Just seems like not particularly the best idea to parse compiler output manually for something that I would assume is a common problem. Alternatively one could make the argument that rust itself should provide a macro maybe that exposes the version?

Yeah I guess a macro would perhaps be better, one day Cargo may not always run rustc ahead of compiling crates, so it may not even have this information on hand at all time.

#4408 tracks the more general idea of setting an env var for the rust version.

I wonder if rust-lang/rfcs#2523 would be more appropriate. I feel like we should at least have a cohesive plan between cargo/lang on how we should be handled version detection.

Considering rust-lang/rfcs#2523 is an approved RFC that is meant to target these kinds of use cases, I"m going to close in favor of that. If there is a reason we should reconsider this, let us know!

Note that we still have #4408 for gathering the rustc version for reporting purposes (like a --bugreport flag)./