feat(harness): agent-drivable dev harness and CI that gates
Build & publish Arch package / arch-package (push) Successful in 2m8s
CI / check (push) Failing after 1m56s
CI / e2e (push) Skipped
Search index maintenance / maintain-index (push) Successful in 13s

A coding agent could develop this repo's Go packages and could not
develop the application: every path to running YellowJacket ended in a
blocking GTK window, so 265 bound methods, 46 events, 33 component
directories and 13 stores had exactly one form of verification
available — `tsc --noEmit`.

The unlock is that `wails dev`'s dev server on :34115 serves the real
frontend with the real generated bindings against the same Go backend a
desktop window attaches to, so a plain Chromium under Xvfb gets a fully
functional app. Four test tiers now exist, cheapest first:

- `make ui-test` — 313 Vitest tests in a real browser in ~2 s, no app,
  no backend, no display. Works because `frontend/wailsjs/` is a pure
  passthrough to `window.go`/`window.runtime`, so faking just those two
  globals runs the real bindings and the real store code.
- `make test` — services in-process, asserting on the payload the
  frontend would receive, via a new `events.Emit` wrapper.
- `make dev-headless` + `playwright-cli` — the real app, driven
  interactively, with an event bridge on `window.__yjEvents` and a
  dev-only control surface at `/__test/`.
- `make e2e` — 19 of those flows frozen as Playwright specs.

`events.Emit(ctx, …)` replaces all 35 direct `runtime.EventsEmit` call
sites: wails' `getEvents` `log.Fatalf`s on any context without its
runtime, so those paths could not run under test and a background
worker could take the app down. Four packages had each hand-rolled the
same guard; nine more guarded on `ctx != nil`, which does not help.
`TestNoDirectRuntimeEmits` fails the build on a new one.

Fixtures are generated, not committed (`make testdata`), and seeds are
built by *running the app* — never by hand-writing config and DB rows,
which would be a second description of a valid YJ_HOME.

`.gitea/workflows/ci.yml` is the first workflow here that tests
anything; the other three only package, so `gitea_ci` reported only
packaging jobs and misled anyone asking whether a push was healthy.
Both jobs were prototyped to green in a bare ubuntu:24.04 container
before the YAML was written, which immediately caught `make lint`
linting three configurations that nothing builds: all three passes
omitted `webkit2_41`, so wails resolved webkit2gtk-4.0 — which Arch
still ships and Ubuntu 24.04 dropped.

Operational instructions live in `.pi/skills/yellowjacket-dev/`,
measured discoveries in `.planning/NOTES.md`, and architecture in
`CLAUDE.md` — split by tense, not by topic, because a topical split
gives every new fact two plausible homes. `make skill-check` fails a
commit if the skill cites a make target that does not exist.
This commit is contained in:
2026-08-10 23:20:42 -04:00
parent 65333857e2
commit 5ca6cad45a
117 changed files with 14585 additions and 262 deletions
+127 -2
View File
@@ -22,11 +22,23 @@ Numbering is sequential and stable across status moves (a plan keeps its `NNN-`
```bash
make dev # Hot-reload development (installs deps, generates code, cleans frontend)
make dev-debug # Same as dev but with YJ_LOG_LEVEL=debug
make dev-headless # Start headless in the background and return (SEED=<name> to seed)
make dev-stop # Stop it (SIGTERM, so shutdown hooks run)
make dev-logs # Tail .dev/app.log
make testdata # Generate the deterministic fixture music library
make sandbox-seed NAME=<n> # Build a seeded YJ_HOME by *running* the app
make build-dev # Debug build with symbols
make build-prod # Production build (stripped, UPX-compressed)
make generate # Run code generators (sqlc + templ via go generate)
make lint # golangci-lint v2 (strict), both build configurations
make test # All tests with race detector, both build configurations
make e2e # Playwright smoke suite against a running dev-headless app
make e2e-setup # Install the e2e runner + its browser (once)
make ui-test # Vitest component/store suite in a real browser (no app)
make ui-visual # Same, including toMatchScreenshot comparisons
make ui-setup # Install the Vitest provider's own Chromium (once)
make bindings-check # Fail if frontend/wailsjs is stale vs the Go bindings
make skill-check # Fail if .pi/ documents a make target that doesn't exist
make lint # golangci-lint v2 (strict), all three build configurations
make test # All tests with race detector, all three build configurations
make vulncheck # govulncheck for CVEs
make setup # Install go tools, frontend deps, git hooks (lefthook)
```
@@ -49,8 +61,75 @@ needs it spelled out:
go test -tags "webkit2_41 indexbuild" ./backend/explore/... ./cmd/...
```
`backend/testctl` is behind a third tag and needs its own pass too
(`make test` runs all three):
```bash
go test -tags "webkit2_41 dev" ./backend/testctl/...
```
Audio playback integration tests require `YELLOWJACKET_INTEGRATION=1`.
### Fixtures and the headless harness
`test_data/music_library_test/` is **generated, not committed**: run
`make testdata` (~1 s) before anything that needs audio. Tests reach it
through `internal/testfixtures`, selecting files by *case*
(`CaseCoverDedup`, `CaseUnicode`, `CaseDuplicates`, …) rather than by
path, and skip themselves when it has not been generated.
The app itself can be run without a blocking window — `make
dev-headless` — and driven with `playwright-cli` against the dev server
on `:34115`, which is the real app with real bindings on `window.go`,
bridged to the same Go backend a desktop window would use.
**The operational half of all this lives in the
`yellowjacket-dev` skill** (`.pi/skills/yellowjacket-dev/`): which tier
to reach for, the exact command sequences, seed lifecycle, and the
failure modes worth knowing before you meet them. It is deliberately
not repeated here — this section describes what exists, the skill says
what to run.
Two things ride on top of the headless launch, both from plan 005
phase 3:
- **The event bridge.** `.playwright/cli.config.json` loads
`.playwright/init-events.js` as an `initScript`, which records every
backend event on `window.__yjEvents`. Half this app is push-driven,
so assertions **await an event, not a timeout**:
`await window.__yjEvents.wait('LibraryScanComplete', {timeoutMs: 60000})`.
It also provides `ready()` and `call('queue.Queue.GetState', [])`,
which times out instead of hanging.
- **The dev-only control surface**, `backend/testctl`, mounted at
`/__test/` on the same port: `health`, `db/snapshot`, `db/restore`,
`emit` (force any backend event, which renders push-driven views
without staging the work that would produce them) and `sql`. It is
compiled out of non-dev builds and additionally requires
`YJ_TESTCTL=1`, which `dev-headless.sh` sets and `make dev` does not.
Frozen regression specs live in `e2e/` (its own npm package, so the
Vitest browser mode does not share a package with the Playwright
runner): `make e2e` against an already-running app.
**The cheapest tier needs none of that.** `make ui-test` runs 313
Vitest tests in a real Chromium in ~2 s with no Wails, no backend, no
seeded library and no virtual display, because `frontend/wailsjs/` is a
pure passthrough to `window.go` / `window.runtime` and
`frontend/test/support/wails-fake.ts` replaces just those two globals —
so the tests exercise the real generated bindings and the real store
code.
**`frontend/wailsjs/` is generated by `wails`, not `go generate`**, so
the pre-commit codegen check does not cover it. `make bindings-check`
(~1.5 s, also a pre-commit hook) regenerates it and fails on a dirty
tree; `make bindings` regenerates it for real.
**Seeds are produced by running the app**, never by hand-writing a
`config.toml` and DB rows — the same discipline `sql/schemas/` gets,
for the same reason.
See `.planning/plans/active/005-agent-development-harness.md`.
## Architecture
**Wails app lifecycle** (`main.go``backend/app.go`): `YellowJacketApp` is the root struct bound to Wails. Its methods are callable from the frontend. Lifecycle hooks: `OnStartup` (init audio), `OnDomReady` (start library scan), `OnBeforeClose` (save window state), `OnShutdown` (persist player/queue state).
@@ -140,6 +219,19 @@ work happens **once, centrally**, and users download the result:
**Event-driven communication**: Backend emits events via Wails runtime; frontend stores subscribe to them. Event names are constants in `backend/events/`.
Emit through **`events.Emit(ctx, name, data...)`**, never
`runtime.EventsEmit` — wails `log.Fatalf`s (unrecoverably) on any
context that does not carry its runtime, which includes every
`context.Background()`, so a direct call cannot run under test and can
kill the app from a background worker. `TestNoDirectRuntimeEmits` fails
the build on a direct call anywhere outside `backend/events`.
That wrapper is what makes services testable in-process: install a
recorder with `events.WithSink(ctx, rec)` and assert on the payload the
frontend would receive (`backend/queue/emit_test.go` is the model).
`events.Deliver` is the same call returning an error instead of
dropping, and has one legitimate caller — `/__test/emit`.
## Code Generation
Two generators run via `go generate ./...` (or `make generate`):
@@ -162,3 +254,36 @@ Tests use `database.NewTestDB(t)` for in-memory SQLite, built by the same
## Git Workflow
Feature branches and PRs are the norm, but direct pushes to `main` are allowed. Pre-commit runs vet, lint, codegen check, and frontend typecheck in parallel. Pre-push runs the full test suite.
## CI
Four workflows in `.gitea/workflows/`. Three of them package and
publish (`arch-package`, `homebrew-formula`, `index-artifact`); only
`ci.yml` gates, and it is the one to look at when deciding whether a
push was healthy.
Two jobs, both in an `ubuntu:24.04` container:
- **`check`** — no display: `make lint` and `make test` (three build
configurations each), `tsc --noEmit`, `make ui-test`,
`make bindings-check`, `make skill-check`.
- **`e2e`** — under Xvfb and a private D-Bus: fixtures, a seed built by
running the app, `make dev-headless`, then the Playwright suite
against **both** Chromium and WebKit. Playwright's Linux WebKit links
Ubuntu 24.04 libraries that Arch does not provide, so CI is the only
place it can run, and it is the closest available approximation of
the WebKit2GTK renderer that ships.
Two things the container needs that a developer machine does not. It
has no PulseAudio socket, so `/etc/asound.conf` makes ALSA's `null`
plugin the default device — that plugin advances its pointer on a
timer, so playback is consumed at real-time rate and the elapsed clock
moves, which `e2e/specs/playback.spec.ts` asserts. And
`YJ_CORE_INDEX_URL` points at a dead address so no run fetches the real
explore artifact, matching what `scripts/seed-sandbox.sh` already does.
**`make lint`'s tag sets must stay identical to `make test`'s.**
Without `webkit2_41` wails resolves `webkit2gtk-4.0`, which Arch still
ships and Ubuntu 24.04 does not — so a mismatch lints a configuration
that only builds on one developer's distro, and says nothing about what
ships.