feat(spec-specs, tests): charge EIP-8037 account creation at access

[spencer-tb]

Description

Implements the EIP-8037 charge at access model for CREATE/CREATE2 and creation transactions from ethereum/EIPs#11858.

Account creation state gas is now charged conditionally at the destination access, before the 63/64ths split, rather than charged unconditionally before the create frame and refilled. Failing a pre access check (balance, nonce, stack depth) reads no destination and charges nothing, an already alive target charges nothing, the charge is decided by existence alone, so a storage only collision is charged and refilled, and a failed create frame still refills.

Creation transactions follow the same rule: the create component of the intrinsic state gas is seeded into the reservoir rather than pre-consumed, charged at the deployment-address access only if the destination does not exist, and refilled on failure.

The storage only collision is untestable, the fixture loader rejects empty with storage pre state accounts (StateWithEmptyAccount), hence the codecov patch miss on the refill lines.

Related Issues or PRs

ethereum/EIPs#11858

Checklist

• Ran fast static checks to avoid CI fails, see Code Standards & Verifying Changes: just static
• PR title has the form <type>(<area>): <title>, where <type> and <area> come from an appropriate C-<type>, respectively A-<area>, label. The title should match the target squash commit message.

Cute Animal Picture

[codecov]

Codecov Report

❌ Patch coverage is 84.61538% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 93.37%. Comparing base (77e95ef) to head (c1c7140).
⚠️ Report is 9 commits behind head on forks/amsterdam.

Files with missing lines	Patch %	Lines
...ethereum/forks/amsterdam/vm/instructions/system.py	84.61%	1 Missing and 1 partial ⚠️

Additional details and impacted files

@@                 Coverage Diff                 @@
##           forks/amsterdam    #3116      +/-   ##
===================================================
+ Coverage            93.31%   93.37%   +0.05%     
===================================================
  Files                  624      624              
  Lines                36994    36990       -4     
  Branches              3384     3386       +2     
===================================================
+ Hits                 34521    34539      +18     
+ Misses                1693     1677      -16     
+ Partials               780      774       -6     

Flag	Coverage Δ
unittests	93.37% <84.61%> (+0.05%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[spencer-tb]

cc @chfast @misilva73

[spencer-tb]

FYI the storage only collision is untestable, the fixture loader rejects empty with storage pre state accounts (StateWithEmptyAccount), hence the codecov patch miss on the refill lines. Please double check, these lines in the spec.

Opcode-level refill (generic_create collision branch):

execution-specs/src/ethereum/forks/amsterdam/vm/instructions/system.py

Lines 122 to 123 in b5fa770

	if new_account_charged:
	credit_state_gas_refund(evm, StateGasCosts.NEW_ACCOUNT)

Transaction-level refill (process_message_call collision branch):

execution-specs/src/ethereum/forks/amsterdam/vm/interpreter.py

Lines 146 to 149 in b5fa770

	if new_account_charged:
	# A storage-only collision target is non-existent:
	# charged above, refilled here.
	message.state_gas_reservoir += Uint(StateGasCosts.NEW_ACCOUNT)

[gurukamath]

This largely looks good to me. But I have a comment on how we should merge this in.

A lot of the changes to src/ethereum/forks/amsterdam/fork.py seem to be transitional scaffolding to bridge over the fact that the latest version of EIP-2780 isn't already in forks/amsterdam. EIP-2780 gets rid of the intrinsic state gas (see this)
The real meat of this PR lies in the system.py changes and some interpreter.py changes.

So, I would say, we should first get #3126 merged in, then rebase this PR and try to get that in next. @spencer-tb your thoughts?

[spencer-tb]

Sgtm!!

[marioevz]

LGTM on the tests side! I just left one comment, because I forgot to update CREATE/CREATE2 docstrings in the PR I submitted to your branch (sorry!)

[gurukamath]

The specs side looks good to me.