Discussion (non-blocking): I have two reservations about making this suggestion:

1. Wouldn't this depend on the specific cargo version?
2. Basically, what I was trying to get at, is thatin principle I feel likerustc should not make assumptions aboutcargo (that is,rustc should not "know" aboutcargo's existence), becausecargo isn't the only way to invokerustc.

I guess I could be convinced by "well, most usersdo build withcargo", so don't consider this a hard-blocking concern.

1. I'm not sure, there is a similar list for this inrust-analyzer:https://github.com/chenyukang/rust/blob/702bf000a0b5062d8b5ced717a26d70919df6ba0/src/tools/rust-analyzer/crates/project-model/src/env.rs#L9-L49 , I guess there is a better and general way to keep it synced.
2. I agreerustc should not knowing details about cargo, I think it's ok since we need to check the prefix withCARGO_ of env:
   https://github.com/rust-lang/rust/pull/148559/files#diff-0eaa1095d59f2f3d9c015005d15fd8f7cda13121a8404d77b2955fc5fcd40449R213, and it's a note but not code suggestions, it's maybe incorrect for rare scenarios.

Yeah, seems alright, was just curious to hear what you think.