APG40 preserves and adopts the exact APG39 authoring phase, independently reviews both replacement candidates, and dispositions them separately:
| Candidate | Corrected-state result | Disposition |
|---|---|---|
matryer-is-test-profile |
Exact-version probes corrected one-sided nil rendering and exposed further caller, diagnostic, writer, and equality details. The single correction pass still misstated the final equality mechanism. | deferred-material-defect; current leaf, specification, fixture, test, route, and projection absent |
nix-test-profile |
Exact Nix and Nixpkgs/NixOS source review corrected phase gating, no-op phases, associated tests, flake outputs, effective configuration, sandbox fallback, cache qualification, and structural accounting. Fresh review passed. | retained-provisional |
ADR 0027 is Rejected. ADR 0026 remains Accepted and controlling for the two
retained Go component owners. ADR 0025 remains Rejected.
go-testing-stack remains absent.
APG39-IS-07 said a one-sided nil failure rendered type information for both
operands. Canonical source and isolated runtime probes in both operand orders
showed the exact contract: comparison is unequal, the nil-like operand renders
as <nil>, and the non-nil operand renders with type and value. APG40 recorded
that corrected outcome before terminal review. The candidate was later
deferred for a different remaining equality-mechanism defect; APG39’s authored
outcome remains preserved in history.
The retained profile begins from an exact claim and selects one project-owned testing surface. It distinguishes evaluation, build, package phases, package-associated derivations, flake output-specific validation, helpers, containers, virtual machines, activation, deployment, and external services without treating them as a scalar ladder. Effective client and daemon configuration, sandbox prerequisites and fallback, builder, cache, system, platform, and actual outcome qualify every runtime claim.
No Nix parse, evaluation, build, flake check, store/cache operation, container, virtual machine, activation, deployment, or external operation ran. Retention is source-reviewed and provisional; it is not runtime compatibility evidence.
Independent reviewers applied both categorical structural models to bounded source and synthetic cases. Size, wrapper depth, failure count, node count, matrix size, and disclosed cost did not manufacture Orange or Red; every escalation named attribution, causality, ownership, truthfulness, authority, safety, exposure, or capacity risk.
matryer/is is MIT. Nix is LGPL-2.1-or-later. Nixpkgs/NixOS are MIT subject to
component exceptions. APG text is independently written factual synthesis.
Public artifacts contain no publication-excluded paths, development Git
identities, credentials, personal details, private topology, copied upstream
expression, or dependency on private/.
Development is 28 canonical skills, 28 catalog rows, and 28 flat projections: 14 stable and 14 provisional rows, 26 general-router entries, one ChatGPT-local entry, and 27 checked route edges. Public and active v0.3.0 remain 19/19/19. No retained owner is promoted and APG40 makes no target repository change. Unrelated target worktree edits appeared concurrently and were left untouched. No readiness, smoke, release, publication, deployment, or successor phase begins.
APG41 subsequently retained nix-test-profile as
ready-provisional-with-limitation after claim-relative static dogfood and
explicit source-to-runtime qualification. It did not run Nix or alter APG40’s
original retained-provisional disposition. The complete APG41 surface reached
ready-for-publication-with-provisional-limitations without reconsidering or
restoring matryer-is-test-profile.