Skip to content

ci: add criterion benchmark regression gate#122

Merged
avrabe merged 1 commit into
mainfrom
ci/bench-regression-gate
May 20, 2026
Merged

ci: add criterion benchmark regression gate#122
avrabe merged 1 commit into
mainfrom
ci/bench-regression-gate

Conversation

@avrabe

@avrabe avrabe commented May 20, 2026

Copy link
Copy Markdown
Contributor

Summary

  • Adds .github/workflows/benchmarks.yml — the CI gate for the existing
    criterion verification benchmarks (src/lib/benches/verification_benchmarks.rs),
    which until now had no runner and therefore caught no regressions.
  • PR job (bench-sanity): paths-filtered to src/lib/src/**,
    src/lib/benches/**, Cargo.toml, Cargo.lock — runs a fast sanity check
    (cargo bench -p wsc --bench verification_benchmarks -- --sample-size 10 --warm-up-time 1)
    to prove the bench code compiles and executes without a 10+ minute run.
  • Nightly job (bench-full): schedule: cron runs the full cargo bench
    and uploads target/criterion/ via actions/upload-artifact@v4.
  • Adds a workflow_dispatch: trigger and copies the concurrency: block
    pattern from formal-verification.yml.

Cross-run criterion baseline persistence is intentionally out of scope — it
needs an external artifact store sigil doesn't have; the artifact upload is
sufficient.

Closes #89.

Test plan

  • cargo build -p wsc --bench verification_benchmarks compiles cleanly.
  • Benchmark harness executes locally — all four criterion groups
    (ed25519_verify, dsse_parse_and_verify, merkle_validation,
    cert_chain_validation) run to Success.
  • CI: bench-sanity job runs and passes on this PR.
  • CI: nightly bench-full job runs and uploads the criterion-output artifact.

🤖 Generated with Claude Code

The verification benchmark code (src/lib/benches/verification_benchmarks.rs)
existed with no CI runner, so it caught no regressions. Add a Benchmarks
workflow with a PR-triggered fast sanity run (reduced sample size, proves the
bench compiles and executes) and a nightly full run that uploads the criterion
output as an artifact.

Cross-run criterion baseline persistence is intentionally out of scope — it
needs an external artifact store sigil doesn't have; the artifact upload is
sufficient.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
@avrabe avrabe merged commit b97a008 into main May 20, 2026
2 checks passed
@avrabe avrabe deleted the ci/bench-regression-gate branch May 20, 2026 18:43
avrabe added a commit that referenced this pull request May 20, 2026
Cleanup-cycle release. Minor-versioned (not patch) because the keyless
client constructors now return `Result` — a breaking change for code
using `wsc` as a library.

Contents:
- #89 (PR #122): criterion benchmark regression-gate workflow.
- #95 (PR #123): Sigstore certificate pinning is now fail-closed —
  `FulcioClient`/`RekorClient` constructors return `Result`, the
  unpinned-fallback path is removed, `Default` impls dropped.

Internal path-dependency version requirements bumped 0.8.0 → 0.9.0
(the `^0.8.0` ceiling would otherwise reject the 0.9.0 workspace).

Deferred and documented: #88 (Kani signature-parser harness — blocked
on a BufReader tractability issue) and #46 (SLH-DSA backend — blocked
on release-candidate-only ecdsa/p256 crates). #91 closed (MIRAI repo
archived).

Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Add criterion benchmarks for signature verification paths

1 participant