const fn support, #[selector = sol("foo()")] attribute

[sbillig]

This adds support for extern and fe-defined const fns, with support for basic code.

For example:

pub const fn sol<T: AsBytes>(sig: T) -> u32 {
    (keccak(sig) >> 224) as u32
}

The rhs of attributes can now be a const expr, so we can do this:

msg Erc20 {
    #[selector = sol("transfer(address,uint256)")]
    Transfer { to: Address, amount: u256 } -> bool,

Limitations (for now):

• const fn must be effect-free (no uses (...))
• supported: literals, if/else, unary/binary ops, indexing, casts, field access, tuple/array/record construction, calls to const functions and methods (including trait-method dispatch when resolvable).
• not supported: loops, match, assignment/+=, with, mutable bindings/patterns
• Array repetition [elem; N] requires N to be fully resolved during type checking (so it fails for abstract consts and const-generic parameters)
• Some extern const fns remain abstract (not evaluated), so they’re rejected in places that require a fully-resolved value (e.g. const items)
• core::keccak can only be used with statically known values for now; it can appear in a normal function body and will evaluate to the statically-computed keccak if possible. Otherwise, it'll result in a codegen error. (This should of course be fixed, so that it does the reasonable thing when faced with a runtime value)

[micahscopes]

Could this be relevant to the optimization stuff happening in #1211 ?

[sbillig]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 52584dc36d

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[sbillig]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 48c71fce9d

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[sbillig]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: a71e322001

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Evaluate const size_of calls in CTFE

core::intrinsic::size_of was made a const fn, but eval_extern_const_fn only handles __as_bytes/__keccak256, so a const item like const N: u256 = core::size_of<u256>() (or a const generic/array length) now type-checks yet fails CTFE and gets reported as “const value must be known.” Since size-of is always computable at compile time, this is a regression for const contexts; consider adding a size_of case here that returns the layout size (similar to the MIR lowering) so CTFE can resolve it.

Useful? React with 👍 / 👎.

[sbillig]

Deferring until later: #1225

[cburgdorf]

Looks good! Let's merge that thang!