镜像站点 · 本页由第三方 GitHub 只读镜像提供,非 GitHub 官方站点,不接受任何登录或凭据输入。前往 github.com
Skip to content

tools: keep algorithmic-growth self-test within the supported evaluator budget #1044

Closed

Description

@chrisbbreuer

Problem

zig build benchmark-comparison-test fails under the repository's zig-js home-tool backend because tools/algorithmic-growth.ts --self-test exhausts the evaluator's default step budget after its accepted-artifact phase.

This reproduces locally in isolation in about 50 seconds, so it is deterministic validation work rather than hosted machine speed. The self-test repeatedly rebuilds and revalidates the same complete nested artifacts: successful construction, explicit validation, default validation, and render() each traverse the whole input; every legacy version repeats build + validate + validating render. The later mutation checks also place the changed row after unchanged rows, paying validation work that cannot affect the asserted failure.

This is separate from #933's human-blocked policy choice for the engine's default step budget. The validator should fit the current supported host by removing redundant traversals while preserving every schema, compatibility, report-byte, and fail-closed assertion.

Definition of done

  • The algorithmic-growth self-test completes through the zig-js home-tool backend at the existing default evaluator budget.
  • Successful artifacts are not rebuilt or revalidated solely to exercise formatting; formatting has a validated internal boundary and public render() remains fail-closed.
  • Default v4 schema pairing, legacy v1-v3 construction/report compatibility, the checked historical report, zero-allocation replay, and every mutation failure remain covered.
  • zig build benchmark-comparison-test passes without increasing or bypassing the engine step budget.
  • Direct JSC and zig-js Home self-tests agree.

Activity

  1. chrisbbreuer commented on Oct 8, 2026

    @chrisbbreuer
    MemberAuthor

    Landed in 9d435e3.

    The self-test now reuses already-validated fixtures for formatting, keeps public render() fail-closed, uses the schema-minimum alternating pair count, and moves mutation witnesses to the first width. Matrix V48 re-pins only the validator; evidence schemas, scored workloads, validation rules, and publication acceptance are unchanged.

    Validation:

    • zig-js Home self-test completes at the unchanged default evaluator budget
    • JavaScriptCore Home self-test passes the same contract
    • zig build benchmark-comparison-test -Dhome-tool=/tmp/home-checker-prefix/bin/home-tool: 19/19 validators passed
    • representative matrix structural tests: 84/84
    • documentation: 51 pages, manifest exact, 378 links resolved
    • 0 test262 cases affected
  2. chrisbbreuer commented on Oct 8, 2026

    @chrisbbreuer
    MemberAuthor

    CI exposed one missing dependency classification after the V48 matrix landed: tools/algorithmic-growth.ts was referenced by the new immutable matrix and its validator without both callers being registered. Fixed in 7cc902f0; the local dependency audit now passes with 16 classified edges and 65 repository tools.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions