Repository navigation
feat!: Satin review follow-ups: Amsterdam gas table, per-spec limits, catchable stops, validation rules, tools and cleanup - #404
Merged
Conversation
…s, keyless errors, the base-fork schedule check, post-Karst receipts
A matching grep -q closed the diff pipe, so the change check exited 141 and skipped the gate. The same happened when git diff itself failed.
The mutation driver defaults its job count to nproc, which stock macOS does not have. The count now comes from sysctl, then from getconf.
The check watched the five contract files and missed the interfaces they import, so an interface-only edit left the checked-in bytecode unexamined.
Refuse MEGA_FUZZ_CASES=0 and a price comparison that shares no path. Require each counted conditional class to be reached, and pin the first-frame history refund against a non-zero operator fee.
The pre-block calls were hitting empty accounts. Seed their code and check, outside the measured section, that both wrote storage.
The seventeen regular-gas entries pressed back to Osaka are gone: every regular entry is Amsterdam's (EIP-8038 and EIP-8037's creation split), so creating state costs its access and write in regular gas and the new state in state gas alone. State gas and history gas do not move. Tests that summed Osaka prices by hand now read them off the schedule. Pins of the spec's own numbers move with the table: the fresh-SSTORE boundary 162,306 -> 152,306, a value CALL's start 713,055 -> 690,755, the keyless CREATE charge 32,000 -> 12,000, the Satin pools' cold call and CREATE in deposit_charge, the dump fixture's 165,826 -> 155,826. Premises that held only at Osaka prices are re-derived: the writes that cross detention's cap (1,000 -> 1,653 slots, and a reservoir for them), the overshoot and tracer caps (+400 per cold access), the refund-cap settlement (8 -> 16 slots), a failed load short of the value transfer. The recorded-block replays now refuse a transaction only for the block gas it declared. Regenerated and reviewed: the pricing table, satin-differences.md, 28 tracer views, the keyless differential record.
No member inherits those workspace dependencies. Nothing reads the hex or attack-deploy inputs, and the disabled replay-bench workflow has no Satin bench mode.
EIP-8038 also gives EXTCODESIZE and EXTCODECOPY a static base of two warm reads (200), which revm keeps in the instruction table rather than the schedule. Satin's instruction table now carries Amsterdam's static prices (satin_gas_table), which part from the Osaka base's in those two opcodes alone. The neutral configuration still installs the fixture fork's own table, so an Osaka fixture runs at 100. The refused EXTCODECOPY read pays its 200 static gas, and one recorded transaction of satin-differences.md moves by 100.
A bare line suppression matches every mutant with that text. The gate now fails when one of them matches more than one.
Clippy rejects a panic that only checks the count is zero.
…fuzz passes, seeded EIP contracts in block benches, unused workspace dependencies and the Rex-era replay bench removed
…DESIZE / EXTCODECOPY at Amsterdam's static gas, and section 5 restated
The per-transaction, per-frame and per-block KV limits go, with the KV stop kind, the data-size-before-records check order, block KV packing and their genesis keys. Write records are still counted for data size and history gas, and the KV count is still reported. LimitKind keeps its encoding: kind 1 decodes to nothing.
A rejected transaction was the only setup check, so a halt or a revert was still measured. Each transaction is now pinned to an outcome class and to its gas used at the spec's byte prices, and the bench checks the pins before it measures. UPDATE_CORPUS_EXPECTATIONS=1 rewrites the pins from a run at the constants. Other byte prices skip the gas pin and still check the outcome class.
A measurement build can leave a counted class unreachable. The keyless tally and the differential coverage counts stay printed, and those minimums bind only at the constants. A zero case count is still refused.
…zz coverage minimums bound to the spec's byte prices
The per-transaction and per-block state-gas limits go, with every check of them: the state-gas holds before the first frame, at the opcode, at frame start and at the code deposit, the lanes' state-gas tracking, the block's post-execution check and the StateGrowth stop kind. The state ledger and the block's state gas counter stay. Kind 3 now decodes to nothing. The crate README and the runner's docs drop the KV limit too.
The Satin spec, the agent guides, the crate READMEs, the mega-evme docs and the review skill state that the registry is a precondition the legacy forks left (no seed, no v1 upgrade, a block without v2.0.0 is invalid) and that the genesis configuration is satinTime alone.
…e holds SequencerRegistryRoles::account_info is the v2.0.0 account (code, nonce 1) that MemoryDatabase::sequencer_registry installs and the blockchain harness puts in its pre-state and takes out of the root.
…nd two test comments
Satin deploys and upgrades no predeploy. The block in which Satin activates (MegaBlockExecutionCtx::satin_activation_block, set by the caller from MegaHardforks::is_satin_activation_block and the parent's timestamp) reads the six system contracts and the EIP-7997 factory and refuses the block with one error, UnexpectedSystemContractCode, when one does not hold Satin's code; no other block reads them. Removed: the create path, nonce 1, the Oracle and registry upgrades, ZeroFactoryNonce, UsedEmptyAccount, CREATE2_FACTORY_NONCE, check_genesis_account. Tests, the blockchain harness, the block bench and mega-evme start from a Rex6-shaped state (test_utils::rex6_predeploys); mega-evme run/tx installs a predeploy only where the state lacks its code.
Remove FALLBACK_RUNG and all_activated_hardforks; hardfork_schedule returns None for an unknown chain. mega-evme refuses to pick an engine for such a chain unless --genesis or --override.spec is given, and builds a Satin counterfactual's schedule locally: the chain's own with Satin from genesis, or the engine's base forks with Satin from genesis.
…in runs on lacks the registry
…le carries no fork parameters
… the registry read
Merging this PR will regress 1 benchmark
|
| Benchmark | BASE |
HEAD |
Efficiency | |
|---|---|---|---|---|
| ❌ |
empty_transaction/satin |
50.1 µs | 52.8 µs | -5.23% |
| ⚡ | storage_reads/op_revm |
166 µs | 153.4 µs | +8.23% |
| ⚡ | storage_reads/satin |
177.6 µs | 165.7 µs | +7.18% |
| ⚡ | value_calls/satin |
526.7 µs | 499.6 µs | +5.44% |
| ⚡ |
deep_calls/satin |
468.6 µs | 445.9 µs | +5.09% |
| ⚡ | remaining_compute_gas/satin |
309.7 µs | 294.8 µs | +5.08% |
Tip
Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.
Comparing cz/feat/satin-w8 (9617c90) with cz/feat/satin-w7 (798701a)2
Footnotes
-
388 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩
-
No successful run was found on
cz/feat/satin-w7(c04aa53) during the generation of this report, so 5fe3a41 was used instead as the comparison base. There might be some changes unrelated to this pull request in this report. ↩
This was referenced Oct 22, 2026
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
1 participant
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.
Summary
W8 of the Satin line, on top of #400. It carries the follow-ups from reviewing the stack, plus tool hardening and a test cleanup:
The full description will be written together while the stack is reviewed. The stack merges down into #385, which then holds the whole Satin engine at its final state.