bug(cast): cast interface generates unexpected library instead of interface

[sakulstra]

Component

Forge

Have you ensured that all of these are up to date?

• Foundry
• Foundryup

What version of Foundry are you on?

9bcfbaa

What version of Foundryup are you on?

No response

What command(s) is the bug in?

cast interface

Operating System

None

Describe the bug

Hey, i think there is a regression in cast interface (or at least an unexpected change).

I have a:

interface ITestContract {
  struct TestStruct {
    address asset;
  }
}

contract TestContract is ITestContract {
  function test(TestStruct memory) external {}
}

Now cast interface ./src/Test.sol:TestContract will yield:

pragma solidity ^0.8.4;

library ITestContract {
    struct TestStruct {
        address asset;
    }
}

interface TestContract {
    function test(ITestContract.TestStruct memory) external;
}

Which I think different to previous versions as i'm pretty sure that before structs were included in the generated interface. Is this change intentional?, as ux wise it's a bit weird.

---

More of a question out of curiosity: Would it be possible to include the natspec on the generated interface?
In most smart contracts i'm reading i have to constantly jump between interface and contract to fully understand the code. Therefore i'm wondering if it wouldn't be a better approach to just have all the natspec on the actual contract (not the interface) and then simply generate the interface.

[zerosnacks]

This is likely related to us moving from Solang to Solar

foundry/crates/cast/bin/cmd/interface.rs

Lines 152 to 167 in 45c3ba1

	/// Converts a vector of tuples containing the ABI and contract name into a vector of
	/// `InterfaceSource` objects.
	fn get_interfaces(abis: Vec<(JsonAbi, String)>) -> Result<Vec<InterfaceSource>> {
	abis.into_iter()
	.map(|(contract_abi, name)| {
	let source = match foundry_cli::utils::abi_to_solidity(&contract_abi, &name) {
	Ok(generated_source) => generated_source,
	Err(e) => {
	warn!("Failed to format interface for {name}: {e}");
	contract_abi.to_sol(&name, None)
	}
	};
	Ok(InterfaceSource { json_abi: serde_json::to_string_pretty(&contract_abi)?, source })
	})
	.collect()
	}

[grandizzy]

@sakulstra I just checked with an (really) old build and there is no regression introduced here, how do you think the output should look?

CC @zerosnacks

[zerosnacks]

Interesting, I would expect it to yield an output like this where ITestContract is an interface, not a library

interface ITestContract {
    struct TestStruct {
        address asset;
    }
}

interface TestContract is ITestContract {
    function test(ITestContract.TestStruct memory) external;
}

[sakulstra]

Trying to figure out which version we were on, but seems like in fact we manually patched the interface because in the version there was some problem with the scoping and i simply forgot.

how do you think the output should look

I would have expected:

pragma solidity ^0.8.4;

interface TestContract {
    struct TestStruct {
        address asset;
    }
    function test(TestStruct memory) external;
}

At least how we currently use cast interface is to build an aggregated interface for contracts with complex inheritance structure. So that you can access Ownable, or AccessControl which might not be defined on the contract itself.
Generating multiple interfaces / libraries kinda defeats the purpose a bit.

[grandizzy]

thanks. ah, I think I was making a mistake and could be a regression at the end, anyways, will look into solving it

[grandizzy]

confirmed it's an regression introduced in d8f6631, most probably in alloy here https://github.com/alloy-rs/core/blob/5b29eb74afb26417704833a0985a2e35bd218ee6/crates/json-abi/src/to_sol.rs#L154-L175

[grandizzy]

after checking the alloy code I think that's something desired and not an issue, @DaniPopes could you pls chime in? thank you!