Understand MIR scopes

[virgil-serbanuta]

Mir functions may define variables in nested scopes. We should figure out if these scopes are useful or they can be ignored (i.e. declaring variables in scopes is the same as declaring them at the top level).

Example:
https://github.com/runtimeverification/mir-semantics-compiletest/blob/9251b00e38504a6f1279b0ca9f81b90b4964080d/ui/as-precedence.mir#L40-L79

[geo2a]

The scopes in MIR have something to do with region checking, which seems related to the borrow checker. The scopes are requires to check "lexical lifetimes". It looks like there is an alternative, more powerful lifetime analysis which is called "non-lexical lifetimes" (NLL, https://rust-lang.github.io/rfcs/2094-nll.html) and is based on the control-flow graph. It is not clear if the scopes are still used for lifetime analysis.

References:

• MIR uses (or used to use?) scopes to track lexical lifetimes: https://rust-lang.github.io/rfcs/1211-mir.html#representing-scopes
• Non-lexical lifetimes: https://rust-lang.github.io/rfcs/2094-nll.html

[geo2a]

It seems like NLL is enables in current Rust: rust-lang/rust#43234 (this issue closed as completed)

Does that mean that we can ignore the scopes?

[virgil-serbanuta]

Even if NLL was not enabled, would we be interested in lifetime analysis? If not, we could ignore scopes anyway.