perf(builder): reuse unchanged withdrawals roots

[BrianBland]

Summary

Avoid recomputing the L2ToL1MessagePasser storage root when building a post-Isthmus payload whose MessagePasser storage is unchanged relative to its parent. Both the standard assembler and flashblocks builder now use a shared helper that reuses the parent's withdrawals root.

Reuse requires a post-Isthmus parent with a withdrawals root, no changed storage slots, and no storage destruction. Otherwise, use the existing calculation. Flashblocks retain their current published-root behavior and inspect cumulative changes relative to the parent, not just the latest flashblock. No global or flashblock-specific cache is introduced; the fast path also benefits Denim's 200ms blocks.

Includes regression tests for unchanged storage, net-zero writes, cumulative flashblocks, destruction, activation boundaries, different parents, and provider errors. Also imports alloc::Box explicitly to fix an existing no_std compilation error exposed during validation.

Validation

Rebased onto base/base/main at 380b55db3; all checks passed after rebasing:

• cargo test -p base-execution-consensus --lib — 41 passed.
• cargo test -p base-execution-evm --lib — 25 passed.
• cargo test -p base-builder-core --lib flashblocks::payload::tests — 9 passed.
• cargo clippy -p base-execution-consensus -p base-execution-evm -p base-builder-core --lib --tests -- -D warnings
• cargo check -p base-execution-consensus --no-default-features --lib
• Nightly rustfmt on changed Rust files and git diff --check.

No latency improvement is claimed without benchmarking.

Generated with Toshi

[cb-heimdall]

🟡 Heimdall Review Status

Requirement	Status	More Info
Reviews	🟡 0/1	Denominator calculation Show calculation 1 if user is bot 0 1 if user is external 0 2 if repo is sensitive 0 From .codeflow.yml 1 Additional review requirements Show calculation Max 0 0 From CODEOWNERS 0 Global minimum 0 Max 1 1 1 if commit is unverified 0 Sum 1

[depot-code-access]

✅ All benchmarks green — 14 within ±2% (deterministic instruction counts). View run

Benchmark details (14)

Benchmark	Base (target)	Head (this PR)	Δ instructions
batch_queue/drain/drain_cached_span_batches	242,027	242,027	+0.0%
batch_transaction/encode_in_place/encode_in_place	4,199,759	4,199,759	+0.0%
batch_transaction/temporary_frame_buffers/temporary_frame_buffers	8,408,350	8,408,350	+0.0%
flashblock_decode/decode/brotli	3,296,484	3,296,484	+0.0%
flashblock_decode/decode/plain_json	2,280,194	2,280,194	+0.0%
flz/compress_len/real_contract_call	43,148	43,148	+0.0%
flz/compress_len/synthetic_0	38,205	38,205	+0.0%
flz/compress_len/synthetic_1	54,682	54,682	+0.0%
flz/compress_len/synthetic_2	147,976	147,976	+0.0%
flz/data_gas	43,059	43,059	+0.0%
flz/tx_estimated_size	43,056	43,056	+0.0%
frame_parse/decode/single_4kib	1,031	1,031	+0.0%
frame_parse/parse_frames/few_large	1,053,062	1,053,062	+0.0%
frame_parse/parse_frames/many_small	154,763	154,763	+0.0%

[depot-code-access]

Base Std historical fork tests

Fork	Result	Passed	Failed	Skipped	base/base	base-anvil	base-std
Beryl	pass	616	0	13	a3139b7e	98e7839c	520d069b
Cobalt	pass	725	0	14	a3139b7e	98e7839c	4571b325

View run

[github-actions]

Review Summary

PR: perf(builder): reuse unchanged withdrawals roots

This PR adds a fast path in WithdrawalsRoot::compute that skips the L2ToL1MessagePasser storage root recomputation when the predeploy's storage hasn't changed since the parent block. Both the standard BaseBlockAssembler and the flashblocks build_block path now use this shared helper.

Analysis

Correctness of the optimization gate (withdrawals.rs:29-36): The three conditions — post-Isthmus parent, parent has a withdrawals_root, and no changed/destroyed MessagePasser storage — are sound. StorageSlot::is_changed() in revm's BundleState compares the original DB value against the present value, so net-zero writes are correctly detected as unchanged. The Isthmus activation boundary is properly handled by checking the parent's timestamp, avoiding reuse of the pre-Isthmus withdrawals root (which had a different semantic meaning). The was_destroyed() guard prevents reuse after SELFDESTRUCT.

Flashblocks path: build_block calls merge_transitions before computing the withdrawals root, so state.bundle_state correctly contains cumulative changes relative to the parent across all flashblocks. The new test (flashblocks_preserve_cumulative_withdrawals_root) validates that once a MessagePasser write occurs in any flashblock, subsequent flashblocks don't revert to the parent root.

Fallback consistency: When the MessagePasser account is absent from the bundle, the fast path returns the parent root. The existing isthmus::withdrawals_root fallback also returns the parent root in this case (via HashedStorage::default() with wiped: false), so the two paths are consistent.

H: BlockHeader bound on assemble_block: This tightens the generic bound, but the BlockAssembler trait impl already constrains H through BlockAssemblerInput, so this is not a breaking change in practice.

Block production risk: No critical findings. The optimization is a pure fast-path skip of an expensive storage root computation, with a correct fallback to the existing calculation when any condition is not met. Error propagation from the fallback path is preserved.

Minor observation

lib.rs:41-42 re-exports withdrawals::tests as withdrawals_root_test_utils under #[cfg(test)], but this alias is not consumed by any code in the crate or workspace. It's harmless dead code — consider removing if it's not intended for future use.

No blocking issues found.