feat: module-level globals

[george-cosma]

Pull Request Overview

This pull request adds module-level support for globals.
Note: not all constant expressions are implemented. The missing ones are:

• ref.null
• ref.func
• global.get (need to properly implement imports first)

Testing Strategy

This pull request was tested by running cargo test + a lot of manually testing in wat2wasm. Details of my findings are in the comments

Help Wanted

I feel like the code is everywhere. A little advice as to how to organize all the new additions would be welcome!

Formatting

• Ran cargo fmt
• Ran cargo check
• Ran cargo build
• Ran cargo doc
• Ran nix fmt

Github Issue

This pr depends on #77

[codecov]

Codecov Report

Attention: Patch coverage is 58.40708% with 47 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/execution/const_interpreter_loop.rs	36.36%	35 Missing ⚠️
src/validation/globals.rs	70.73%	8 Missing and 4 partials ⚠️

Files with missing lines	Coverage Δ
src/core/reader/types/global.rs	25.00% <ø> (+25.00%)	⬆️
src/execution/mod.rs	96.07% <100.00%> (+4.75%)	⬆️
src/validation/code.rs	67.68% <ø> (+3.49%)	⬆️
src/validation/mod.rs	62.71% <100.00%> (+4.98%)	⬆️
src/validation/globals.rs	70.73% <70.73%> (ø)
src/execution/const_interpreter_loop.rs	36.36% <36.36%> (ø)

... and 2 files with indirect coverage changes

[george-cosma]

As a note, besides what we discussed I also moved validate_value_stack to validation/mod.rs

[florianhartung]

I'm not sure whether it is the right approach to have duplicated validations and implementations for all const expressions (see execution::const_interpreter_loop::run_const).

[george-cosma]

I'm not sure whether it is the right approach to have duplicated validations and implementations for all const expressions (see execution::const_interpreter_loop::run_const).

While I agree that having both another validator and an execution loop for const expressions is a bit inelegant - tons of repeated code! - the alternative would be to integrate these expressions into the regular loops. While possible, this would take a lot of work and I'd wager that the end result would be less readable than what we have now. Just to name a few reasons, the const expression loop needs to return the length of its code (it isn't mentioned in the binary, like how it is for code blocks); a merged execution loop would still need to differentiate between regular execution and const execution, because we want to error out if we try to execute a non-const instruction in a const expression; etc.

[george-cosma]

In the last commit, I've done a bit more rework so that all validation of globals, including init expressions is done entirely within the validation module.