Skip to content

ci(bench): add JVM and WASI binding nightly lanes (#858) - #911

Open
adityamukho wants to merge 2 commits into
v3from
feat/858-jvm-wasi-lanes
Open

adityamukho wants to merge 2 commits into
v3from
feat/858-jvm-wasi-lanes

Conversation

@adityamukho

Copy link
Copy Markdown
Collaborator

Summary

Adds the JVM (14 IDs) and WASI (3 IDs) lanes to the binding nightly system, part of #858.

  • bench.yml, bindings.sh, bindings-report.cjs: jvm and wasi lanes with their own testbeds (binding-jvm-ubuntu24-v1, binding-wasi-ubuntu24-v1), exact-result validation, selected = executed = uploaded reporting, and the existing regression notification.
  • The validator runs from the measured checkout, so the workflow is identical on main, which has no benchmarks/bindings839/. The companion main PR carries the same workflow files.
  • benchmarks/bindings839/nightly.py: WASI adapter (pinned Wasmtime 49.0.2 by version and checksum, no preopen).
  • The regression repeat lookup searches back for the latest prior run with matching production history.
  • bench-report.yml also triggers on benchmarks/bindings839/**.
  • Docs and Bench: timing and release sign-off for JVM, Android, Swift/iOS and WASI #839 status overlays; spec copied to docs/superpowers/specs/.

The JVM adapter is in project-minigraf/minigraf-java (bench/nightly.sh).

Scope

No change to src/, Cargo.toml or the file format. v3-only measurement; main runs skip the lanes.

Verification

  • Node reporting and workflow tests (35) and Python tests (13) pass, including negative controls for wrong-but-plausible results.
  • Not yet verified: hosted runs, Bencher baselines, live notification scenarios.

🤖 Generated with Claude Code

adityamukho and others added 2 commits November 18, 2026 15:23
Generalize the binding nightly job and reporting to cover a JVM lane
(14 IDs) and a WASI lane (3 IDs) with their own testbeds, provenance,
exact-result validation and notification paths. v3-only, like the
existing binding lanes.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Run the binding validator and hash the harness from the measured checkout so
the workflow is identical on main, where benchmarks/bindings839 is absent.
Skip the Python harness tests when it is absent. The regression repeat lookup
now searches back for the latest prior run with matching production history.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
@codecov

codecov Bot commented Nov 18, 2026

Copy link
Copy Markdown

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

This branch has not been deployed

No deployments
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.

1 participant