[sferik]

A review of the same Duel workflow in another project turned up a couple of things worth fixing here too, so this brings perf.yml in line with what duel init generates now.

The Duel job checks out and runs the pull request's code, benchmarks included, and it also had pull-requests: write so it could post its report. Now it only reads, and its checkout keeps no credentials. It uploads the report as an artifact, and a separate Comment job downloads it and posts the comment. That job checks nothing out and runs none of the branch's code, so it's the only one that can write. A pull request from a fork still gets no comment, since its token can't write one, but the job's summary has the report.

The other fix is in .duel.rb. Duel only compares a change that touches lib/ or ext/, so a pull request that changed only the benchmarks or the config passed with "nothing to compare", and a broken benchmark could slip through. It now watches duel/, .duel.rb and the workflow too. That also means this pull request runs a real comparison of the code with itself, which checks that the new workflow works.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-26T01:32:04.515508Z	8859c6d	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[github-actions]

Duel passed: none of 28 benchmarks slower

• Baseline: c2bf295, where HEAD left origin/main
• Candidate: a3761ae

28 benchmarks in 14s. A timing is slower when a 99.0% confidence interval for its ratio, the candidate's to the baseline's, is above 1.02.

Benchmark	Verdict	Timings (ratio [interval])	Counts an iteration
adapter: parse in with_adapter	unsettled	cpu 1.006 [0.990, 1.022] undecided
generate: a large set of records	unsettled	cpu 1.020 [1.005, 1.165] undecided

Across the board

• cpu 1.001 [0.998, 1.006] undecided (limit 1.0050, 28 benchmarks)

Unsettled

A timing too close to the limit to settle doesn't fail the run, but one that's unsettled in run after run is noisy: adapter: parse in with_adapter, generate: a large set of records.

26 benchmarks unchanged

Benchmark	Timings (ratio [interval])
adapter: resolve one by name	cpu 1.006 [0.990, 1.019]
options: a cache miss in a full cache	cpu 1.005 [0.993, 1.014]
legacy: MultiJson.load	cpu 1.009 [0.998, 1.018]
legacy: MultiJSON.dump	cpu 0.991 [0.944, 1.019]
generate: a tiny hash	cpu 0.997 [0.972, 1.015]
generate: a small flat object	cpu 1.002 [0.992, 1.019]
generate: a small set of records	cpu 1.000 [0.995, 1.016]
generate: a large array-heavy document	cpu 1.000 [0.990, 1.008]
generate: pretty	cpu 0.974 [0.936, 1.012]
generate: with Oj	cpu 0.970 [0.603, 1.006]
generate: pretty, with Oj	cpu 1.008 [0.995, 1.018]
generate: through fast_jsonparser	cpu 1.000 [0.991, 1.016]
parse: a tiny object	cpu 0.998 [0.985, 1.019]
parse: a scalar	cpu 1.002 [0.995, 1.020]
parse: blank input	cpu 1.003 [0.993, 1.017]
parse: a small flat object	cpu 1.004 [0.997, 1.013]
parse: a small set of records	cpu 0.998 [0.974, 1.005]
parse: a medium set of records	cpu 1.001 [0.987, 1.008]
parse: a large set of records	cpu 1.002 [0.992, 1.017]
parse: a large array-heavy document	cpu 1.000 [0.989, 1.010]
parse: with symbolize_names	cpu 1.009 [1.000, 1.017]
parse: from an IO	cpu 1.001 [0.981, 1.017]
parse: binary-encoded input	cpu 0.998 [0.990, 1.015]
parse: invalid JSON	cpu 0.995 [0.979, 1.016]
parse: with Oj	cpu 0.985 [0.954, 1.007]
parse: with fast_jsonparser	cpu 1.008 [0.979, 1.019]