fix(interpreter): use IntoAddress in pop_address to avoid const-eval panic

[pjdurden]

Summary

StackTr::pop_address converts the popped stack word with
Address::from(value.to_be_bytes()). Because Address: From<[u8; 20]>, type
inference forces ruint's const-generic Uint::<256, 4>::to_be_bytes::<20>(),
whose const { Self::assert_bytes(BYTES) } requires BYTES == Self::BYTES
(32 for U256).

That const block is unreachable for valid inputs, so the code compiles whenever
the compiler dead-code-eliminates it — but the Rust Reference only guarantees such
a block may or may not be evaluated. Build profiles that do evaluate it
(cargo-fuzz coverage instrumentation, release codegen) hit
error[E0080]: evaluation panicked: BYTES must be equal to Self::BYTES,
breaking downstream builds. Fixes #3728.

Fix

Reuse the crate's existing canonical converter, IntoAddress for U256
(crates/interpreter/src/instructions/utility.rs), via value.into_address().
It does Address::from_word(B256::from(value.to_be_bytes())), whose
to_be_bytes() infers the correct [u8; 32] size — identical EVM word→address
truncation (keep the low 20 bytes) with no fragile const-generic instantiation.
This also removes the inlined conversion in favour of the shared helper. The same
problematic idiom is present on main, so this is not release-specific.

Testing

Added pop_address_truncates_to_low_20_bytes (truncates a full 32-byte word to its
low 20 bytes; returns None on an empty stack). Verified on rustc 1.96:

• cargo test -p revm-interpreter --all-features → 45 passed
• cargo clippy -p revm-interpreter --all-targets --all-features → clean
• cargo fmt --all --check → clean
• cargo build -p revm-interpreter --release and --no-default-features → both build
  (these are the profiles that previously hit the E0080)

Closes #3728

[Copilot]

Pull request overview

Note

Copilot was unable to run its full agentic suite in this review.

Refactors pop_address to use the IntoAddress utility trait instead of constructing an Address directly from big-endian bytes, and adds a unit test verifying the truncation behavior.

Changes:

• Replace inline Address::from(value.to_be_bytes()) with value.into_address() in StackTr::pop_address.
• Import IntoAddress from instructions::utility.
• Add a test covering both the truncation-to-low-20-bytes behavior and the empty-stack case.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
crates/interpreter/src/interpreter_types.rs	Use IntoAddress trait to convert popped U256 to Address.
crates/interpreter/src/interpreter/stack.rs	Add test asserting pop_address returns low 20 bytes and None on empty stack.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[rakita]

No need for this

Suggested change

	#[test]
	fn pop_address_truncates_to_low_20_bytes() {
	use primitives::Address;
	// A full 32-byte word whose high 12 bytes are set and must be discarded,
	// and whose low 20 bytes are the address. This mirrors how the EVM reads an
	// address from a stack word (e.g. BALANCE/EXTCODESIZE/EXTCODECOPY targets).
	let word = U256::from_be_bytes::<32>([
	0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0x11, 0x22,
	0x33, 0x44, 0x55, 0x66, 0x77, 0x88, 0x99, 0xaa, 0xbb, 0xcc, 0xdd, 0xee, 0xff, 0x00,
	0x12, 0x34, 0x56, 0x78,
	]);
	let expected = Address::from([
	0x11, 0x22, 0x33, 0x44, 0x55, 0x66, 0x77, 0x88, 0x99, 0xaa, 0xbb, 0xcc, 0xdd, 0xee,
	0xff, 0x00, 0x12, 0x34, 0x56, 0x78,
	]);
	run(|stack| {
	assert!(stack.push(word));
	assert_eq!(stack.pop_address(), Some(expected));
	});
	// Empty stack returns None rather than panicking.
	run(|stack| {
	assert_eq!(stack.pop_address(), None);
	});
	}

[rakita]

Hey lets remove the test, it is not needed

[pjdurden]

Hey lets remove the test, it is not needed

@rakita , i've removed the test. One small issue - The CI workflows are still showing awaiting approval, would you mind approving the runs so the workflow can run?

[codspeed-hq]

Merging this PR will not alter performance

✅ 176 untouched benchmarks
⏩ 1 skipped benchmark¹

---

_{Comparing pjdurden:fix/pop-address-const-eval-panic (32b4008) with main (16aab94)}

Footnotes

1. 1 benchmark was skipped, so the baseline result was used instead. If it was deleted from the codebase, click here and archive it to remove it from the performance reports. ↩