Programming Language

Nocter

A self-contained systems language built around simplicity, encapsulation, and foolproof design.

Documentation GitHub
/development/milestones/v0.34.0-release-preparation.md

v0.34.0 Release Preparation

Status: published and externally audited (2026-09-04). Corrective commits after the first qualification changed compiler and standard-library content. The replacement candidate below passed the complete qualification contract and was published without rebuilding. Public evidence is frozen in the release audit.

Release Boundary

v0.34.0 adds one distinct Unicode-scalar model across source literals, constants, built-in type identity, checked ownership, MIR, runtime shape, target layout and ABI, ARM64 execution, borrowed UTF-8 iteration, owned String mutation, formatting, hashing, diagnostics, structural output, and semantic tooling. Unicode properties, case mapping, normalization, grapheme segmentation, display width, locale behavior, and scalar indexing remain outside this release.

The milestone and Phase 0 through Phase 4 reviews define the closed implementation. Public release notes are ../../releases/v0.34.0.md. Root download links and the public release index identify v0.34.0.

Candidate Identity

The release inputs agree on 0.34.0:

  • development/packaging/VERSION, the sole authored release-version input;
  • the #package version in development/std/index.nct;
  • the archive name and installed manifest derived from VERSION and RELEASE.json;
  • compiler and language-server reporting derived from the validated installed home.

The qualified release-content commit is b876301201105e061000e2d48c8d18246d11814d. Documentation-only qualification evidence recorded after this commit is not an artifact input. The superseded candidate came from 3834ab6602ea3e4f3d821a937da958b6a77ac337 and is no longer retained.

Qualification Contract

The clean release-content commit must pass:

  1. 1. two independent complete locked workspace test runs in distinct external Cargo targets;
  2. warnings-denied all-target Clippy in both external targets;
  3. formatter, no-default-features checking, documentation generation and reproducibility, repository whitespace, and packaging-input validation;
  4. the explicit public-HTTPS dependency-acquisition test;
  5. checks and representative native execution for every public example against an isolated candidate home;
  6. two independent optimized package builds with byte-identical archives and recursively identical installed homes;
  7. fresh extraction checks for version, doctor, help, initialization, locked/offline check and test, deterministic graph, exact public example process contracts, Unicode scalar native behavior, explicit build and execution, and framed LSP analysis;
  8. rejection after either the installed compiler or standard-library content is changed, plus proof that ordinary qualification commands do not mutate the installed home.

Qualification Evidence

The exact replacement candidate passed two independent complete locked workspace test runs and two warnings-denied all-target Clippy runs in distinct external Cargo targets. Every enabled test passed; the public-HTTPS test was ignored in the workspace runs and passed explicitly. The candidate also passed no-default-features checking, formatter and repository validation, deterministic 480-page documentation generation, and packaging-input syntax checks.

Two independent optimized builds produced byte-identical archives and recursively identical homes. Fresh extraction passed version and integrity diagnosis, help, package initialization, locked/offline check and test, every public example, deterministic package graphs, exact process contracts, native run and build, direct execution, framed LSP analysis, installed-home immutability, and compiler and standard-library tamper rejection.

Performance measurements compared separately extracted v0.33.0 and v0.34.0 release archives on the same arm64-darwin machine. Seven alternating process-cold checks of examples/hello.nct had medians of 0.83 s and 0.84 s. Five alternating LSP samples, each containing 100 sequential full-document body edits followed by checked hover queries, had medians of 23,294.533 ms and 23,352.154 ms. A v0.34.0 native workload traversing six mixed-width scalars one million times had a median wall time of 0.49 s and median user time of 0.43 s over seven successful processes. These measurements are evidence, not stable performance contracts.

The retained replacement candidate is:

  • archive: dist/nocter-v0.34.0-arm64-darwin.tar.gz;
  • size: 8,339,944 bytes;
  • archive SHA-256: db0751b6e869a9d1dfb911a93b299d122b6cdb73ca9eb6d3cf6e97efcd7507f9;
  • compiler SHA-256: 4c45f0aa57bb1824babe77e284cc370759c45b1a277680c50bd976abf5822ab1;
  • standard-library tree SHA-256: 29955cd25f368e3d7bec3b92608994a1c246ad49e3914dcf7e46b77ffe861389;
  • standard-library regular files: 153.

Publication Boundary

Publication reused the retained archive without rebuilding, created one annotated v0.34.0 tag, uploaded exactly one asset, and verified the public download byte for byte. The release, tag, and asset are immutable by project policy; corrections require a new version.