Conformance

How Comline proves its implementations agree, and how it keeps track of their speed, size and health over time. Each subject has its own page. This one shows how they fit together.

The pipeline

flowchart LR
    subgraph producers["Producers (each repository's CI)"]
        rt["runtime"]
        gen["generation"]
        ts["comline-typescript"]
        core["core"]
        cli["cli"]
        sim["simulator"]
    end
    producers -->|gates on pull requests| pr["pass / fail"]
    producers -->|reports on main| rb["reports branch<br/>or default branch"]
    gh["GitHub API"] --> col
    rb --> col["Observatory collector"] --> site["Observatory site"]
    rb --> docs["docs: Languages pages"]

1. Pull requests are gated in each repository's own CI: vectors and goldens must match, support claims must be proven, fuzz targets must survive a short run, and instruction counts must not grow more than 1% without a ledger entry.
2. Pushes to the default branch publish reports, one JSON file per measurement, committed to the repository's reports branch. Conformance reports are committed on the default branch itself.
3. The Observatory collects every report's history, so each published version can be opened and any two compared. It shows results and gates nothing.
4. The docs render the latest conformance, fuzz and bench results into the Languages pages at build time.

Subjects

Schedule