mirror of
https://github.com/vercel/workflow.git
synced 2026-09-14 19:59:43 +08:00
Alex Langenfeld
d864efb07b
## Summary & Motivation - Manifest coverage now declares an entry for every matrix app, uses real Vitest skips instead of silent early returns, and fails when a targeted app's manifest is missing, unknown, or unparseable. - Retries are scoped to deployment e2e runs (`DEPLOYMENT_URL` set), so a flaky unit or integration test can no longer be hidden by a second attempt. - The stop-workflow cookbook parks on a sleep between iterations, giving the hook an observable barrier to race instead of a fixed delay, and the AbortController hook test waits on queue state rather than a 10ms timer. - The world-postgres direct-storage fixture drives its run to a terminal state so the conformance worker doesn't recover and replay an unregistered workflow. - Generated e2e result sidecars are ignored and the committed copies removed; they're CI artifacts, not fixtures. ## Test Plan Existing coverage runs in CI. With retries disabled: the cookbook agent suite passed 8/8, the two stop-workflow tests passed 10/10, and the AbortController hook replay test passed 25/25 under `CI=1`. The manifest suite skips 52 apps explicitly when nothing is built, and fails on unknown or missing targeted apps. The Docker-backed Postgres spec could not run locally (Testcontainers found no container runtime); `@workflow/world-postgres` typechecks.
32 lines
1.8 KiB
TypeScript
32 lines
1.8 KiB
TypeScript
import { configDefaults, defineConfig } from 'vitest/config';
|
|
|
|
export default defineConfig({
|
|
test: {
|
|
testTimeout: 60_000,
|
|
// Deployment e2e suites can lose timing races to queue delays, cold
|
|
// starts, and watcher latency. They always set DEPLOYMENT_URL, so keep
|
|
// their one visible retry without masking deterministic unit/integration
|
|
// failures elsewhere in `turbo test`. The github-reporter annotates
|
|
// retried e2e tests and includes them in the PR summary. Harnesses where
|
|
// a failure is itself the signal (event-log-race-repro, benchmarks) pin
|
|
// `retry: 0` locally. Local runs also keep retry at 0 for reproduction.
|
|
retry: process.env.CI && process.env.DEPLOYMENT_URL ? 1 : 0,
|
|
// How many concurrent tests vitest runs from a `describe.concurrent`
|
|
// suite (vitest's own default is 5). Only the e2e conformance suite is
|
|
// concurrent, so this is effectively its dial. Tunable because the right
|
|
// value is a property of the runner and the deployment rather than of
|
|
// the tests: every CLI assertion spawns a `node` child, and a CI runner
|
|
// has few cores, so too high a value inflates per-test latency until
|
|
// tests exceed budgets written for an unloaded suite. Each lane logs
|
|
// what it observed (see `summarizeLoad` in the e2e utils).
|
|
maxConcurrency: Number(process.env.WORKFLOW_E2E_MAX_CONCURRENCY ?? 5),
|
|
// Positional file arguments are regex filters, not paths, so
|
|
// `vitest run packages/core/e2e/x.test.ts` also matches
|
|
// `.claude/worktrees/<name>/packages/core/e2e/x.test.ts` when agent
|
|
// worktrees live inside the repo (see .gitignore). Those copies belong to
|
|
// other branches: they would run their own version of the suite against the
|
|
// same backend and overwrite the same result files.
|
|
exclude: [...configDefaults.exclude, '**/.claude/**'],
|
|
},
|
|
});
|