You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
feat(testing): spec-conformance test for AdcpError code names against error-code.json enum #416
The AGENT_SUSPENDED, AGENT_BLOCKED, REQUEST_AUTH_UNRECOGNIZED_AGENT, and INVALID_BILLING_MODEL codes shipped for months as non-spec (caught in #375 / fixed in PR #393). Drift was only caught by manual review.
Add an automated test that walks all AdcpError(\"CODE\", ...) raise sites in src/adcp/ (AST or grep) and asserts the code is in the canonical static/schemas/source/enums/error-code.json enum (or has the X_ vendor prefix). Catches future drift at PR time.
Acceptance Criteria
New tests/test_error_code_conformance.py
Walks all AdcpError(\"CODE\", ...) raise sites in src/adcp/
Asserts code is in static/schemas/source/enums/error-code.json enum or has X_ vendor prefix
Classification: Feature request Bucket(s): client, handlers Status: drafting-pr Milestone: (omitted — no target version specified)
What the experts said:
code-reviewer: approved — no blockers; stale comments and schema fixture hardening addressed in review→fix cycle
ad-tech-protocol-expert: approved — X_ prefix aligns with spec's extension model; AUTH_INVALID allowlist is protocol-correct given the spec's explicit forward declaration of the AUTH_REQUIRED split
dx-expert: AUTH_INVALID is a named future spec code — do not use X_AUTH_INVALID; allowlist approach preserves the semantic distinction the spec plans to formalize
My take: Issue is clear-cut and all criteria for Execute are met. One design question (rename AUTH_INVALID → AUTH_REQUIRED vs. allowlist it as a provisional spec code) is surfaced in the PR body for reviewer decision; both approaches are valid and the PR uses the lower-scope one.
Activity
bokelley commentedon May 3, 2026
Triage
Classification: Feature request
Bucket(s): client, handlers
Status: drafting-pr
Milestone: (omitted — no target version specified)
What the experts said:
X_prefix aligns with spec's extension model;AUTH_INVALIDallowlist is protocol-correct given the spec's explicit forward declaration of the AUTH_REQUIRED splitAUTH_INVALIDis a named future spec code — do not useX_AUTH_INVALID; allowlist approach preserves the semantic distinction the spec plans to formalizeMy take: Issue is clear-cut and all criteria for Execute are met. One design question (rename
AUTH_INVALID→AUTH_REQUIREDvs. allowlist it as a provisional spec code) is surfaced in the PR body for reviewer decision; both approaches are valid and the PR uses the lower-scope one.PR: #436 — AST conformance test + rename
INTERNAL_ERROR→X_INTERNAL_ERROR,BILLING_NOT_PERMITTED_FOR_AGENT→X_BILLING_NOT_PERMITTED_FOR_AGENT. 231 tests pass.Triaged by Claude Code. Session: https://claude.ai/code/session_01SpT9n4csqsXxbogX4q6GHo
Generated by Claude Code
bokelley commentedon May 14, 2026
Shipped via #429 (merged). Closing as completed — the triage-bot draft was superseded by the merged PR but the issue was never closed.