fix(tests): enhance & un-skip Amsterdam ported static tests (Pt. 1) - #3215
Merged
spencer-tb merged 18 commits intoJul 29, 2026
Merged
Conversation
…via work through examples
Enhance and un-skip 10 Amsterdam skip-list entries (234 -> 224): - test_create_empty_contract: CREATE gas via CodeGasMeasure; merge with_balance as a parametrize. - test_create_empty_contract_and_call_it (new): merge 0wei/1wei; CALL gas via CodeGasMeasure with a runtime SLOAD address. - test_copy_offset (new): merge CODECOPY/CALLDATACOPY OOB-offset; drop the delivery CALL. - test_delegatecall_emptycontract, test_deleagate_call_after_value_transfer: gas forwarded via the default Op.GAS; the latter also restores its "after value transfer" intent with a non-zero tx value. - test_static_call_value_inherit_from_call: STATICCALL sees CALLVALUE 0. - test_make_money: value flow as a relationship; drop gas-noise balance. Coverage widened from the ported Cancun-only down to each test's true fork floor. Skill enhance-ported-test.md updated with the validated patterns (CREATE metadata, omit subcall gas, protected=fork.supports_protected_txs, absolute-vs-delta gas, degraded-port restoration). Claude-Session: https://claude.ai/code/session_01FxUr7ixeLr8apwbNzBLgNo
Fold the six stEIP150singleCodeGasPrices RawCreate*Gas fillers (plain, memory, value-transfer, value-transfer-memory, and two insufficient-balance failures) into one CodeGasMeasure parametrize on (create_value, size, fails). Un-skips 6 Amsterdam entries (224 -> 218) and lowers the floor to SpuriousDragon. Confirms the EIP-8037 error-path question: a balance-failure CREATE runs no init code and creates no account, so it charges regular gas only. Success asserts create_code.gas_cost(fork) (regular + state); failure asserts create_code.regular_cost(fork) — one expression correct on every fork. Skill updated: error-path regular_cost split (resolves a known gap), and a new known gap for the 63/64 callee-side *_gas_ask shape. Claude-Session: https://claude.ai/code/session_01FxUr7ixeLr8apwbNzBLgNo
The ported MCOPY_copy_cost filler probed the copy cost indirectly via a razor- thin OOG gas boundary (gas_limit 55697): a flag SSTORE ran, then the MCOPY, and slot 0 stayed set only if the whole thing fit. EIP-8037 reprices that flag SSTORE (zero->non-zero) into ~97920 state gas, which alone exceeds the budget, so every case OOGs on Amsterdam. Reframe to measure the MCOPY copy cost directly with CodeGasMeasure: pre-expand memory past the largest copy region so the measured MCOPY charges only its base + per-word copy cost (never expansion), and assert the fork-derived mcopy.gas_cost(fork). Parametrized over the same (src, size) matrix; a pure execution-gas op, so it is fork-robust. Un-skips 55 Amsterdam entries (218 -> 163). Claude-Session: https://claude.ai/code/session_01FxUr7ixeLr8apwbNzBLgNo
Fold the ten stEIP150singleCodeGasPrices RawCall*GasAsk fillers (CALL / CALLCODE / DELEGATECALL x value-transfer x memory) into one parametrize that verifies the EIP-150 "all but one 64th" gas-forwarding rule. Un-skips 10 Amsterdam entries (163 -> 153). The ported fillers pinned the forwarded gas as an absolute number tied to the tx gas limit, which the EIP-2780 intrinsic change makes fork-fragile. Reframed (approach by @marioevz): an outer call caps the caller frame at a known gas budget so the forwarded amount is deterministic; the callee returns its observed GAS up to the top frame (no lower-frame SSTORE state-gas trap); the expected value is derived from the fork as all_but_one_64th(CALLER_GAS - call.gas_cost(fork)) + stipend - GAS. Note: the cap is available - available // 64, NOT available * 63 // 64 (they differ by 1 off a multiple). Skill updated with the forwarded-gas / 63/64 pattern and the // 64 gotcha. Claude-Session: https://claude.ai/code/session_01FxUr7ixeLr8apwbNzBLgNo
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## forks/amsterdam #3215 +/- ##
==================================================
Coverage ? 93.42%
==================================================
Files ? 625
Lines ? 37032
Branches ? 3385
==================================================
Hits ? 34597
Misses ? 1666
Partials ? 769
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Harness. 🚀 New features to boost your workflow:
|
12 tasks
Merged
2 tasks
Contributor
|
Added some extra tweaks here: marioevz#7 |
This was referenced Jul 29, 2026
Closed
This was referenced Aug 5, 2026
Draft
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Description
Amsterdam's EIP-8037 two-dimensional (regular vs. state) gas model broke a large batch of machine-ported
tests/ported_static/tests, which were temporarily parked inamsterdam_skip_list.txt. These fillers encode gas the wrong way for a fork that reprices: hardcoded absoluteGASsnapshots, tight OOG gas budgets, and per-opcode gas constants baked into expectations. When EIP-8037 moved account creation / storage writes into the state dimension (and EIP-2780 dropped the intrinsic tx cost 21000 → 15000), every one of those pinned numbers shifted and the tests failed on Amsterdam.This PR (Part 1) rewrites a first tranche of them to be fork-robust — they now express gas through framework constructs that track the live schedule (
CodeGasMeasure,bytecode.gas_cost(fork)/.regular_cost(fork)/.state_cost(fork), opcode metadata) rather than magic numbers — so a future fork that re-prices gas no longer silently breaks them. Each enhanced test is verified across its fullvalid_fromrange and on Amsterdam, and its skip-list entry removed.Result: the Amsterdam ported-static skip list drops from 258 → 153 entries (105 un-skipped). Along the way ~40 near-identical sibling fillers are consolidated into a handful of parametrized tests (56 files changed: +1067 / −4621), and most tests gain much wider fork coverage (the ported
valid_fromwas almost always an over-tightCancun; many now run from Frontier / Berlin / SpuriousDragon onward).The remaining 153 skipped cases — dominated by OOG-boundary, revert-depth, refund, and collision families — are left for a follow-up PR (Part 2).
Highlights
SSTOREis repriced into ~97920 state gas on Amsterdam, so every case OOG'd. Reframed to measure the copy cost directly withCodeGasMeasureover pre-expanded memory, assertingmcopy.gas_cost(fork).RawCall*GasAsk(10 → 1). Verifies the EIP-150 "all but one 64th" gas-forwarding rule. An outer call caps the caller frame at a known budget so the forwarded amount is deterministic (independent of the tx gas limit), the callee returns its observedGASup to the top frame (avoiding a lower-frame state-gas trap), and the expectation is derived asall_but_one_64th(caller_gas − call.gas_cost(fork)) + stipend − GAS.RawCreate*Gas(6 → 1). One parametrize over value / memory / failure; confirms an EIP-8037 behavior — a balance-failureCREATEcharges regular gas only (regular_cost(fork)), never state gas.CodeGasMeasure; gas-independent behavior tests freed of tightgas_limits; subcallgasoperands left to default (Op.GAS) instead of hardcoded constants; dynamic addresses throughout;protected=fork.supports_protected_txs()to keep the fork floor at the behavior's real EIP rather than masking a signature floor.Methodology / tooling
enhance-ported-testClaude Code skill (.claude/commands/enhance-ported-test.md) and wires it intoCLAUDE.md— an ordered, empirically-validated methodology for turning a ported filler into idiomatic, fork-robust Python (max out gas, dynamic addresses,CodeGasMeasure, consolidation,valid_fromfloors, re-pinning discipline). This is the playbook Part 2 will follow.Related Issues or PRs
N/A.
Checklist
just static<type>(<area>): <title>, where<type>and<area>come from an appropriateC-<type>, respectivelyA-<area>, label. The title should match the target squash commit message.Cute Animal Picture