test(tpagectl): add comprehensive test suite #612

Merged

fuzzy merged 1 commit from test/tpagectl-test-suite into main

2026-08-08 16:07:56 +00:00

automation commented

2026-08-08 15:59:49 +00:00

Owner

What

Implements Phase 8 test coverage (roadmap #288).

#284 — client unit tests with mocks — a roundTripFunc mock http.RoundTripper covering the previously-untested do() branches (JSON encode-request error, decode-response error on a 2xx with invalid JSON) plus mock-driven bearer-token attachment. Client coverage → 90.9%.
#285 — integration tests against a test instance — new cmd/tpagectl/integration_test.go with a stateful fake daemon (httptest) implementing the full /api/v1 surface + /healthz//readyz, running the real root command end-to-end (config → client → command → output → exit code): site lifecycle (create/list/get/update/delete), deployment list/get/promote/rollback, preview deploy/list/get/delete, status, --format json/yaml, and exit-code classification (not-found 4, validation 5). Running the actual daemon requires Forgejo+K8s, noted as out of scope.
#286 — golden tests — internal/tpagectl/testutil.Golden with a -update flag; .golden files for site list (table/json), site get, deployment list, preview list, status, and root --help.
#287 — coverage >80% — aggregate raised to 86.4% (weighted, tpagectl packages): credentials 3.8% → 88.5% via an injectable openKeyring opener tested with keyring.NewArrayKeyring; gap tests for config (→81.9%), docs (→89.3%), auth (→87.3%), interactive (→72.8%; the TTY-gated picker path is documented as not unit-testable). Added a Makefile coverage target gating weighted coverage of ./internal/tpagectl/... ./cmd/tpagectl/... at ≥80%.

Why

Roadmap task #288.

Testing

go test -race ./... passes (30 packages)
make coverage reports 86.4% and passes the ≥80% gate
golangci-lint run clean
pre-commit hooks pass (gofmt, go mod tidy, go test)

Breaking Changes

None. credentials.open() now uses an injectable openKeyring var (default keyring.Open); behavior unchanged.

Notes

The interactive ResolveSite TTY path (raw-mode fuzzy picker) cannot be unit-tested without a pty; its selectInteractive core is covered via synthetic key bytes. CI enforcement of the coverage gate is deferred to task #293 (CI pipeline).

Closes #284
Closes #285
Closes #286
Closes #287

## What Implements Phase 8 test coverage (roadmap #288). - **#284 — client unit tests with mocks** — a `roundTripFunc` mock `http.RoundTripper` covering the previously-untested `do()` branches (JSON encode-request error, decode-response error on a 2xx with invalid JSON) plus mock-driven bearer-token attachment. Client coverage → 90.9%. - **#285 — integration tests against a test instance** — new `cmd/tpagectl/integration_test.go` with a stateful **fake daemon** (httptest) implementing the full `/api/v1` surface + `/healthz`/`/readyz`, running the real root command end-to-end (config → client → command → output → exit code): site lifecycle (create/list/get/update/delete), deployment list/get/promote/rollback, preview deploy/list/get/delete, status, `--format json/yaml`, and exit-code classification (not-found 4, validation 5). Running the actual daemon requires Forgejo+K8s, noted as out of scope. - **#286 — golden tests** — `internal/tpagectl/testutil.Golden` with a `-update` flag; `.golden` files for site list (table/json), site get, deployment list, preview list, status, and root `--help`. - **#287 — coverage >80%** — aggregate raised to **86.4%** (weighted, tpagectl packages): `credentials` 3.8% → 88.5% via an injectable `openKeyring` opener tested with `keyring.NewArrayKeyring`; gap tests for `config` (→81.9%), `docs` (→89.3%), `auth` (→87.3%), `interactive` (→72.8%; the TTY-gated picker path is documented as not unit-testable). Added a Makefile `coverage` target gating weighted coverage of `./internal/tpagectl/... ./cmd/tpagectl/...` at ≥80%. ## Why Roadmap task #288. ## Testing - [x] `go test -race ./...` passes (30 packages) - [x] `make coverage` reports 86.4% and passes the ≥80% gate - [x] `golangci-lint run` clean - [x] pre-commit hooks pass (gofmt, go mod tidy, go test) ## Breaking Changes None. `credentials.open()` now uses an injectable `openKeyring` var (default `keyring.Open`); behavior unchanged. ## Notes The interactive `ResolveSite` TTY path (raw-mode fuzzy picker) cannot be unit-tested without a pty; its `selectInteractive` core is covered via synthetic key bytes. CI enforcement of the coverage gate is deferred to task #293 (CI pipeline). Closes #284 Closes #285 Closes #286 Closes #287

automation added 1 commit

2026-08-08 15:59:49 +00:00

test(tpagectl): add comprehensive test suite

Test and Release / lint (pull_request) Successful in 7m46s

Details

Test and Release / test (pull_request) Successful in 8m4s

Details

d635624e02

- #284: client unit tests with a mock RoundTripper covering encode and
  decode failure paths plus bearer-token attachment.
- #285: integration tests against a stateful fake daemon (httptest)
  running the real root command end-to-end: site lifecycle, deployments,
  previews, status, output formats, and exit-code classification.
- #286: golden tests for stable outputs (site list table/json, site get,
  deployment list, preview list, status, root help) with a -update flag.
- #287: coverage raised to 86.4% aggregate (credentials 3.8% -> 88.5% via
  an injectable keyring opener and ArrayKeyring tests; config/docs/auth/
  interactive gap tests) and a Makefile coverage target gating the
  tpagectl packages at >= 80%.

Closes #284
Closes #285
Closes #286
Closes #287

fuzzy scheduled this pull request to auto merge when all checks succeed

2026-08-08 16:06:15 +00:00

the.auditor approved these changes

2026-08-08 16:07:49 +00:00

the.auditor left a comment

Excellent test suite. The stateful fake daemon covers the full /api/v1 surface end-to-end (config → client → command → output → exit codes) with deterministic ordering (slice-based payloads, substring assertions on the map-order list), the credentials tests close the #593 coverage gap cleanly via the injectable openKeyring (with a withOpener restore-helper, covering unavailable/genuine-failure/notfound paths), golden tests use fixed-body servers for determinism, and the Makefile coverage gate math is correct. Approving; no new issues — the CI enforcement gap is already tracked as #293.

Notes (non-blocking)

cmd/tpagectl/integration_test.go:493 – var _ = fmt.Sprintf is a dummy keeping an unused fmt import; trivial cleanup.
credentials.openKeyring is a mutable package-global for test injection — safe today (no t.Parallel), but a latent footgun if parallel tests are added later.
Golden root-help.golden is brittle by design (help output changes with any flag/command addition); standard golden-test tradeoff.

Praise

Client roundTripFunc mock covers the previously-untested do() encode/decode-error branches and bearer-token attachment.
Integration tests exercise the real root command with --token/--server, keeping the keychain out of the path.
Coverage improvements are targeted (gap tests per package) rather than padding, and the gate is measurable and reproducible.

Excellent test suite. The stateful fake daemon covers the full `/api/v1` surface end-to-end (config → client → command → output → exit codes) with deterministic ordering (slice-based payloads, substring assertions on the map-order list), the `credentials` tests close the #593 coverage gap cleanly via the injectable `openKeyring` (with a `withOpener` restore-helper, covering unavailable/genuine-failure/notfound paths), golden tests use fixed-body servers for determinism, and the Makefile coverage gate math is correct. Approving; no new issues — the CI enforcement gap is already tracked as #293. ## Notes (non-blocking) - `cmd/tpagectl/integration_test.go:493` – `var _ = fmt.Sprintf` is a dummy keeping an unused `fmt` import; trivial cleanup. - `credentials.openKeyring` is a mutable package-global for test injection — safe today (no `t.Parallel`), but a latent footgun if parallel tests are added later. - Golden `root-help.golden` is brittle by design (help output changes with any flag/command addition); standard golden-test tradeoff. ## Praise - Client `roundTripFunc` mock covers the previously-untested `do()` encode/decode-error branches and bearer-token attachment. - Integration tests exercise the real root command with `--token`/`--server`, keeping the keychain out of the path. - Coverage improvements are targeted (gap tests per package) rather than padding, and the gate is measurable and reproducible.

fuzzy merged commit d635624e02 into main

2026-08-08 16:07:56 +00:00

fuzzy deleted branch test/tpagectl-test-suite

2026-08-08 16:07:56 +00:00