Skip to content

TESTING.md

Latest commit

 

History

History
76 lines (54 loc) · 3.93 KB

TESTING.md

File metadata and controls

76 lines (54 loc) · 3.93 KB
Raw

Testing Guide

Validation surface for oc-codex-multi-auth. For suite layout and conventions (naming, directories, anti-patterns), see test/AGENTS.md; this file covers the commands and the checks that matter when changing things.

Release-Gate Commands

Run all of these before opening a PR — the same six commands run in CI (.github/workflows/ci.yml):

npm ci
npm run typecheck
npm run lint
npm run build
npm test
npm run audit:ci
Command What it checks
npm run typecheck tsc --noEmit
npm run lint ESLint flat config (no-explicit-any; tests linted under relaxed rules)
npm test Vitest suite — auth, config, transforms, storage, rotation, tools, TUI, recovery, docs parity
npm run test:coverage Coverage thresholds from vitest.config.ts
npm run build clean dist/, compile, copy OAuth success page
npm run audit:ci production audit + dev-advisory allowlist

Suite Layout

find test -name '*.test.ts' is the inventory — do not hardcode counts or file lists anywhere. Roughly: top-level test/*.test.ts unit/integration suites (named after the module under test, e.g. lib/storage/keychain.ts → storage-keychain.test.ts), plus test/chaos/ fault injection, test/property/ fast-check tests (FC_SEED=<n> replays a run), test/contracts/ upstream wire shapes, and test/tools-codex-*.test.ts per-tool regressions.

Conventions: vitest globals on; vitest.config.ts redirects HOME to a per-run throwaway dir so suites never touch real credentials (OC_CODEX_TEST_HOME to opt into your own); OAuth tests bind the real port 1455; fake timers over wall-clock assertions; source imports only, never dist/.

Coverage Floors

vitest.config.ts enforces perFile thresholds per directory glob — each floor sits under the weakest file it matches today, so a real drop trips the gate without blocking routine drift. Deliberately low: lib/ui/** (interactive widgets can't run under the harness), tui.ts, and the re-export barrels. Keeping them instrumented keeps the gap visible.

Documentation-Adjacent Checks

When docs touch setup/config/tooling claims, verify against the live surface:

  1. Confirm commands exist in lib/tools/index.ts and each has a matching lib/tools/codex-*.ts module.
  2. Confirm config examples match config/opencode-modern.json, config/opencode-legacy.json, config/minimal-opencode.json.
  3. Confirm install/update guidance matches scripts/install-oc-codex-multi-auth.js.
  4. Confirm repo scripts quoted in docs still exist in package.json.

npm test -- test/doc-parity.test.ts automates most of this — catalog counts, tool counts, path references, npm scripts, auth labels, version strings.

Manual Smoke Checks

npx -y oc-codex-multi-auth@latest --dry-run      # install + config merge
opencode debug config                          # merged provider.openai + plugin entry
ENABLE_PLUGIN_REQUEST_LOGGING=1 opencode run "ping" --model=openai/gpt-5.5 --variant=medium

Verify the last one writes logs under ~/.opencode/logs/codex-plugin/ and keeps store: false + reasoning.encrypted_content. For payload-level debugging add DEBUG_CODEX_PLUGIN=1 CODEX_PLUGIN_LOG_BODIES=1 — it can log sensitive request/response bodies, so use it only when needed.

Interactive commands worth a manual pass when the auth menu or account flows change: codex-setup, codex-doctor, codex-next, codex-list, codex-dashboard.

Failure Triage

Surface Command
lint/style npm run lint
type drift npm run typecheck
transform/request behavior npm test -- request-transformer
storage/migration npm test -- storage
tool output npm test -- tools-codex-<name> or npm test -- index
docs/metadata drift npm test -- doc-parity

See Also