Diagnostic code consumer audit
Versioned draft snapshot
RFCs, reference catalogs, and evidence reflect Spec 3b2e3e7a114fb826b1445d39a48756b8fff629cd. Counts and pending/status notes belong to that snapshot. Read the dated project update for later progress.
Source repositories are private during staging. Links marked “repository access required” need access to the linked repository.
This is a trace and gap record from the pinned draft specification. It does not establish proof or implementation conformance.
Owner: spec and the affected compiler/control-plane maintainers. Local
source audit on 2026-10-02. The owner reports that nothing is deployed, so
the first release needs no live diagnostic migration. CI secrets and
third-party source integrations have not been inspected.
The reviewed diagnostic family is M013xx–M016xx, M019xx, M14800/01,
and L01400–L01402. Searches covered the current gnomon, graph,
expression, site, and brand repositories and the original
savvifi/aion, savvifi/graph, savvifi/expression, and savvifi/studio
checkouts. Generated RDF, copied spec, source RFCs, and Lean catalogs carry
many definitions and references. The observed operational consumers are:
| Surface | Observation | Migration consequence |
|---|---|---|
gnomon/controlplane/deploy/src/gate.rs | StubMigrationGate emitted M01600 at INFO severity for a structural-lint placeholder. RFC-0016 defines M01600 as an error for RemoveField with live non-null data and no drop: true. | This was an actual meaning and severity collision. The stub emitter was removed; it performed no structural validation and now emits only a DRIFT finding when the plan hash differs. |
gnomon/proto/gnomon/deploy/v1/deploy.proto | GateVerdict.findings comments name the M01600–M01606 family; GateFinding.code documents M01601 and symbolic DRIFT/PROVENANCE. | Keep the wire field stable. The comment names future normative diagnostics; a current stub verdict must not claim they were evaluated. |
| Original Aion control plane and proto | The same emitter and comments were found in the readable original checkout. | Preserve the original repository and account for any deployed consumer separately; a new-repo source fix does not repair an existing deployment. |
Original gnomon/devdocs | The reading guide uses M01502 as an example but says the first two digits identify the RFC owner, which would assign it to RFC-0001 rather than RFC-0015. | Do not republish that explanation unchanged. The new site should describe registered legacy exceptions explicitly. |
graph, expression, site, brand, and their original checkouts | No exact source matches for the reviewed code family outside the copied spec/reference material. | This is only a source-search result, not proof that deployed or stored data lacks the codes. |
The focused //controlplane/deploy:unit Bazel target passed after the stub
emitter was removed. The admission test now asserts an empty finding list for
a matching plan hash; the existing drift test still requires a DRIFT error.
The proto field and descriptor were not changed.
The inherited narrative/decisions/diagnostic-codes.md decision record
says code identifiers are permanent and explicitly reserves M0xxxx as a
legacy allocation. That evidence favors preserving assigned historical IDs
as their canonical identities while recording exact owners and meanings in
a registry. The decision record itself has inconsistent examples and a stale
RFC-0000 section reference; it must be reconciled with the current normative
RFCs before publication.
Acceptance for the remaining consumer audit: inspect the actual compiler and linter emitters, MCP diagnostic transport, registry publish gate, destination CI, and any source integration discovered before first deployment. Verify exact code, severity, and trigger for each occurrence. Search results do not authorize silently reusing an identifier with a conflicting meaning.