fix(tests): fixes EXTCODECOPY bench test

[jochem-brouwer]

🗒️ Description

In #1864 I realized a problem with the EXTCODECOPY test. It copies to memory to offset 85. This is correct if the CREATE2 hash alignment starts at 0. It however starts at byte 11 (the address starts at 12). This means that the memory region [11,96) is occupied by the bytes to hash to get the correct address. If we thus EXTCODECOPY to byte 85, we overwrite the bytes there (initcode hash is stored there). This means that all EXTCODECOPYs after the first one target incorrect addressess, i.e. they have empty code.

I checked this file if there are more of these tests but this seem to be the only one. (This could be a problem with other tests also using CREATE2 and using operations which writes to memory. It should not overwrite the to-be-hashed data to retrieve the CREATE2 address created from the factory)

CC @jsign

🔗 Related Issues or PRs

N/A.

✅ Checklist

• All: Ran fast tox checks to avoid unnecessary CI fails, see also Code Standards and Enabling Pre-commit Checks:

  uvx --with=tox-uv tox -e lint,typecheck,spellcheck,markdownlint

• All: PR title adheres to the repo standard - it will be used as the squash commit message and should start type(scope):.
• All: Considered adding an entry to CHANGELOG.md.
• All: Considered updating the online docs in the ./docs/ directory.
• All: Set appropriate labels for the changes (only maintainers can apply labels).
• Tests: Ran mkdocs serve locally and verified the auto-generated docs for new tests in the Test Case Reference are correctly formatted.
• Tests: For PRs implementing a missed test case, update the post-mortem document to add an entry the list.
• Ported Tests: All converted JSON/YML tests from ethereum/tests or tests/static have been assigned @ported_from marker.

[jsign]

Good catch!

Just in case, I debugged a run of both the previous and new code in Geth, and can confirm the bug and fix.

[jsign]

Parallel suggestion: maybe something like this can be clearer for readers:

Suggested change

	attack_call = Op.EXTCODECOPY(address=Op.SHA3(32 - 20 - 1, 85), dest_offset=96, size=1000)
	create2_params_offset = 32 - 20 - 1
	create2_params_size = 85
	attack_call = Op.EXTCODECOPY(address=Op.SHA3(create2_params_offset, create2_params_size), dest_offset=create2_params_offset+create2_params_size, size=1000)

[jochem-brouwer]

Yes good point, I'll follow up to somewhat take out these "magic numbers" 😄 👍

[marioevz]

LGTM, thanks for catching this!

I've verified through tracing analysis that the fix works.

[marioevz]

Checking the coverage drop:

• std::memset(&state.memory[dst + num_bytes_copied], 0, num_bytes_to_clear); clears memory when the size argument of EXTCODECOPY is larger than the target contract, which in the previous implementation was true since we were incorrectly targeting empty contracts.

This is basically a memory expansion test case, but I think it's out of scope for this PR.