"found possibly newer version of crate `std` which `cc` depends on" while building rustc

[SimonSapin]

While working on std lately, I semi-regularly end up with an error like below. Unfortunately I don’t know the exact conditions to reproduce this. It possibly involves re-using a build directory after starting a new branch from a new position of git upstream/master.

This looks similar to #44739, which was fixed in #44792. Maybe something there regressed? CC @Mark-Simulacrum

Building stage1 codegen artifacts (x86_64-unknown-linux-gnu -> x86_64-unknown-linux-gnu, llvm)
   Compiling rustc_llvm v0.0.0 (file:///home/simon/rust1/src/librustc_llvm)
error[E0460]: found possibly newer version of crate `std` which `cc` depends on
  --> librustc_llvm/build.rs:11:1
   |
11 | extern crate cc;
   | ^^^^^^^^^^^^^^^^
   |
   = note: perhaps that crate needs to be recompiled?
   = note: the following crate versions were found:
           crate `std`: /home/simon/rust1/build/x86_64-unknown-linux-gnu/stage1/lib/rustlib/x86_64-unknown-linux-gnu/lib/libstd-bde80a95b5831673.rlib
           crate `std`: /home/simon/rust1/build/x86_64-unknown-linux-gnu/stage1/lib/rustlib/x86_64-unknown-linux-gnu/lib/libstd-bde80a95b5831673.so
           crate `cc`: /home/simon/rust1/build/x86_64-unknown-linux-gnu/stage1-rustc/release/deps/libcc-5e2a255eead634fb.rlib

error: aborting due to previous error

For more information about this error, try `rustc --explain E0460`.
error: Could not compile `rustc_llvm`.

[Mark-Simulacrum]

I think I know what the problem here is but it won't be a trivial fix -- we'll need to split Mode into more structs. I think one of my planned refactors for rustbuild will fix this, but it's a couple weeks out probably.

[petrochenkov]

Stuff like this is especially annoying during bisecting, you can't just checkout a commit and test.

[dsciarra]

I can confirm this error still happens. Restarting the build didn't help I had to first delete the generated libraries. @Mark-Simulacrum did the refactor of rustbuild already land?

[Mark-Simulacrum]

It's in progress; there's a PR open which I think might help with this or at least help make a future PR simpler (#52036).