Skip to content
Benchmarks

EVALUATION · RELEASE CANDIDATE

Release candidate against the last release

What the benchmark records for the redact-secret commit in development, set beside the last released version. No candidate is recorded in this build, so the last release is shown for reference.

No release candidate is recorded

Nothing is compared in this build

A release candidate is a redact-secret main commit that passed artifact qualification and was then replayed as a product candidate: the same official evidence, engine and configuration as the release, with the candidate's own build. The publication reads that recorded replay; it never measures a candidate itself and never reuses a released observation as one. No replay is recorded for this build's candidate, so the page does not guess and shows no difference.

Production measures the released package only, so it is expected to look like this.

Not recorded

  1. A maintainer registers the candidate's tarballs in benchmarks/product-candidates.json and dispatches the candidate replay (official-runs.yml, input candidate).
  2. qualification:candidate-diff compares the replay with the control at the same pins; the staging publication runs it with --verify-tarballs.
  3. The next build of this page reads the projection and shows its counts.
npm run qualification:candidate-diff -- --candidate <registered id> \
  --candidate-dir <downloaded candidate-run artifacts> --verify-tarballs \
  --out results-output/candidate-diff-from-artifacts.json

The last release, for reference

Last releasepublished

0.1.0-beta.14

Commit
0c62fd3 release pin
Date
Not recorded
Run
Not recorded
Source
No baselines/0.1.0-beta.14.json is saved

Performance cost

No candidate to compare

Not recorded

The accepted performance run measured commit 0c62fd3 (5 repetitions). No candidate is recorded, so no cost is compared and nothing is estimated.

Performance comparison