Figure out how to check that `minicore` matches signatures of real `core`

[jieyouxu]

Currently, there's no safeguard to guard against minicore signatures and type definitions drifting/diverging from that of the real core, which is very much not ideal. We should figure out a way to check if the signatures (and type definitions, where applicable) match between minicore and real core.

One related issue that may help with this is to use host core when available but that in itself is insufficient (and can cause complications where something only fails on targets where host core isn't available).

[lolbinarycat]

is there any way to use cargo-semver-checks for this?

[jieyouxu]

I don't think so, not easily at least. Minicore has privileges of being in the same tree as the compiler and library, and so can use unstable features that c-s-c won't be able to catch up or even use.

[lolbinarycat]

@jieyouxu since this involves analyzing function signatures, perhaps it could reuse some of the tooling used for rust-lang/compiler-team#790 ?

[jieyouxu]

@jieyouxu since this involves analyzing function signatures, perhaps it could reuse some of the tooling used for rust-lang/compiler-team#790 ?

AFAIK that isn't implemented, like at all (that scheme also has potential soundness issues that needs experimenting)

[lolbinarycat]

Yeah, I should be more clear, what I meant is "perhaps we can develop tooling/library functions that will be useful for both goals".