Pipeline giga static block processing

[codchen]

Summary

• split giga static transaction preparation from stateful execution
• add an async per-block static processing pipeline so later block stateless work can overlap earlier block stateful execution
• wire giga synchronous and OCC execution paths to consume prepared EVM metadata, keeping Cosmos transactions as legacy fallbacks

Tests

• go test ./app
• make giga-integration-test
• make giga-mixed-integration-test
• make autobahn-integration-test
• make parquet-integration-test

[cursor]

PR Summary

High Risk
Changes consensus-critical giga/OCC block execution ordering and when stateless vs stateful EVM checks run; incorrect static results or pipeline races could diverge tx outcomes from the prior path.

Overview
Giga static pipeline front-loads decode plus state-free EVM work (Cosmos envelope checks via EvmStatelessChecks, eth tx conversion, signature recovery) into gigaStaticEVMTx results, with concurrent per-tx preparation and panic recovery mapped to ABCI errors.

Async block jobs start from ProcessProposal (and again in ProcessBlock) keyed by height/hash, cap concurrent jobs (32 total, 4 per height), prune stale entries, and let ProcessBlock wait and reuse prepared typedTxs/staticTxs before stateful execution.

Execution paths now take optional precomputed static results: synchronous giga and OCC giga short-circuit on static failures, call executeEVMTxWithGigaExecutor with prepared metadata instead of re-running stateless checks inline, and OCC maps static results by tx checksum with on-demand fallback.

Tests cover concurrent static prep, multi-height pipeline retention, and same-height job caps; giga executor Signer.Equal is implemented for type-safe comparison.

^{Reviewed by Cursor Bugbot for commit fd54eac. Bugbot is set up for automated code reviews on this repo. Configure here.}

[github-actions]

The latest Buf updates on your PR. Results from workflow Buf / buf (pull_request).

Build	Format	Lint	Breaking	Updated (UTC)
✅ passed	✅ passed	✅ passed	✅ passed	Jun 26, 2026, 4:50 PM

[codecov]

Codecov Report

❌ Patch coverage is 43.17181% with 129 lines in your changes missing coverage. Please review.
✅ Project coverage is 59.05%. Comparing base (0ff5a52) to head (fd54eac).

Files with missing lines	Patch %	Lines
app/app.go	43.75%	111 Missing and 15 partials ⚠️
giga/executor/internal/signer.go	0.00%	3 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3495      +/-   ##
==========================================
- Coverage   59.16%   59.05%   -0.12%     
==========================================
  Files        2262     2187      -75     
  Lines      187009   181590    -5419     
==========================================
- Hits       110643   107237    -3406     
+ Misses      66430    64750    -1680     
+ Partials     9936     9603     -333     

Flag	Coverage Δ
sei-chain-pr	46.79% <43.17%> (?)
sei-db	70.41% <ø> (ø)
sei-db-state-db	?

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

Files with missing lines	Coverage Δ
giga/executor/internal/signer.go	18.18% <0.00%> (-1.82%)	⬇️
app/app.go	66.37% <43.75%> (-4.91%)	⬇️

... and 488 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[cursor]

Cursor Bugbot has reviewed your changes using default effort and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 5d8b337. Configure here.}

[seidroid]

Cleanly splits state-free EVM tx preparation (envelope checks + sender recovery) from stateful giga execution and pipelines it across blocks via a deduplicated (height,hash) job map; callers, imports, panic recovery, and tests are all in order. No blockers found; a few non-blocking observations.

Findings: 0 blocking | 6 non-blocking | 2 posted inline

Blockers

• None at the file/PR level.

Non-blocking

• Determinism dependency worth a code comment: the static result is started in ProcessProposalHandler (proposal ctx) but deduped-and-consumed in ProcessBlock (which may run under the FinalizeBlock deliver ctx). This is consensus-safe only because prepareGigaStaticTx depends solely on header-derivable values (ctx.BlockHeight/BlockTime/ConsensusParams, chainID, tx bytes) and never touches the KV store. A note on prepareGigaStaticTx / StartGigaStaticBlockProcessing stating 'must remain strictly state-free' would protect against a future change to EvmStatelessChecks or GetVersion (signer-version selection) silently introducing a state dependency and breaking consensus across the two contexts.
• Second-opinion passes produced no material findings: codex-review.md reports 'No material findings' (and notes it could not run go test ./app due to a Go 1.24 vs required 1.25.6 mismatch); cursor-review.md is empty.
• Minor: in ProcessTxsSynchronousGiga the staticTx.err != nil ('[BUG]') branch calls recordGigaTxProcessed() while the analogous downstream execErr '[BUG]' branch does not — a trivial metrics inconsistency on an unreachable bug path.
• Consider asserting len(staticTxs) == len(typedTxs)/len(txs) more defensively at the entry of ProcessTxsSynchronousGiga; today a short non-nil slice that happens to match length but is internally misaligned would index silently, though current callers always pass either nil or a full-length slice.
• 2 suggestion(s)/nit(s) flagged inline on specific lines.

[seidroid]

[nit] Nit: this [BUG] branch calls recordGigaTxProcessed() but the analogous downstream execErr [BUG] branch (the txResults[i] = {Code:1, Log:"[BUG] giga executor error..."} path) does not record the metric. Harmless and on an unreachable path, but worth making the two [BUG] paths consistent.

[seidroid]

[suggestion] Consider a short comment here noting that the static work started with the proposal context is later deduped-and-consumed by ProcessBlock (potentially under the finalize-block context) via the (height,hash) key, and that this is only consensus-safe because prepareGigaStaticTx depends solely on header-derived values (height/time/consensus-params/chainID/tx bytes) and never reads chain state. It guards future maintainers against introducing a true state dependency in EvmStatelessChecks/GetVersion that would diverge between the two contexts.