Files
yetone 2da8fbe488 tests: every test that runs plugins on the bun on PATH runs it once first, through testenv.Bun, so a bun that has never run here doesn't time out internal/provider's and internal/plugin's plugin tests (#1306)
65dcc573 (ggbdpq's patch 1) warmed the bun on PATH before the root
package's one-minute plugin contexts. The same lookup of the bun on PATH
is in 40 more places in internal/provider (28), internal/gateway (8),
internal/plugin (3) and internal/gui (1), and in the root package's
TestPluginListSaysClash. Many of them then open a one- or two-minute
context (internal/provider's sign-in, fetch and move tests,
internal/plugin's sandbox), and there macOS's first-exec check of a
just-installed or upgraded bun landed inside the deadline. The rest
are covered the same way, so no test has to remember which is which.

testenv.Bun now does what the root package's pathBun did: it finds the
bun on PATH (skipping the test when there is none) and runs
`bun --version` before the test opens its context, once per bun per test
binary. pathBun is gone; every `exec.LookPath("bun")` + skip block in
the tests uses testenv.Bun instead. Three tests reach for the bun some
other way and are left as they were: TestHostRunsAsMagpieBun (7ddea2c2
covers its copy), the plugin_git test's realBun and netproxy's
TestBunThroughBridge, none of which runs the plugin host under a
deadline on the PATH bun. migrate_test.go needed two `err` declarations
that the removed block used to provide.

What was run (macOS arm64, bun 1.4.2), with a wrapper bun first on PATH
whose first exec sleeps 61s (marker-gated, then execs the real bun):
- Red, old code:
  - internal/provider TestPluginSignInTakesAPastedCallback: FAIL at
    61.1s, "pluginsignin_paste_test.go:36: starting plugins: context
    deadline exceeded".
  - internal/plugin TestFakePlugin: FAIL at 62.0s, "plugin_test.go:66:
    Plugins = [], starting plugins: context deadline exceeded".
  - internal/gateway TestPluginFailover passed on the old code too (its
    calls have no deadline that short), so the gateway is not shown red.
- Green, this change, same cold wrapper: the three tests passed (61.1s,
  62.9s, 61.2s).
- go test -tags nogui -count=1 for the root package and for
  internal/testenv, gateway, gui, plugin and provider, under a temp HOME
  with GOMODCACHE, GOCACHE and GOPATH pinned: all ok.
- gofmt -l on the changed files: clean. go vet of those packages for
  darwin (-tags nogui), GOOS=linux and GOOS=windows (-tags nogui), and
  GOOS=windows without the tag: clean.

Not run: a really cold bun on a loaded machine; the sleep stands in for
the check. Not run on Linux or Windows, where the warm-up is one extra
`bun --version` per test binary.
2026-10-10 10:22:59 +08:00
..
gateway, provider tests: the tests stay off the network (testenv.Offline in their TestMains) — after 555f5706 (#1016) routing reads a plan key's quota windows, and a test whose provider has a vendor's real base URL and a made-up key would ask the vendor in the background. Routing's own read (keyAllowance, allowances) was already stubbed in routing_test.go's init, and a survey with a dial guard found no plan-key dial, but it found others: every plugin host's first start fetched models.dev's catalog (65 dials in the gateway's plugin tests), and provider tests asked api.github.com (TestSubscriptionUsageEachAccount, TestSubscriptionUsageServesStale), chatgpt.com (TestCopilotQuotaWithoutEditorsSignIn), server.codeium.com (TestDevinSignedInOnlyInMagpie) and decide.test (TestModelTest) with made-up accounts; each passed whether the vendor answered or not. internal/testenv's Offline wraps http.DefaultTransport's DialContext (so http.DefaultClient and every transport cloned from it later): localhost, *.localhost, loopback and unspecified addresses are dialled, anything else is refused before it is looked up, said once per address on stderr ("testenv: a test asked models.dev:443, refused: tests stay off the network"). internal/gateway's TestMain and internal/provider's (kiro_main_test.go, in RunIn's setup) call it before the tests run. TestLocal (the address table) and TestOffline (a loopback httptest server answers through DefaultClient; models.dev is refused through it and through a Clone made after the call) pass -race -count=3. go test -race -tags nogui -timeout 30m ./internal/testenv ./internal/provider ./internal/gateway passes under a temp HOME (gateway 516 s). The gateway and provider test binaries were then run under sandbox-exec with every non-loopback network-outbound denied, with the kernel's Sandbox reports streamed: both PASS, Offline refused only the six addresses above, and the kernel reported no denial, so no transport outside DefaultTransport reaches the network either. go vet, gofmt -l, linux and windows (nogui) builds pass. Tests only.
2026-10-06 21:08:57 +08:00
gateway, provider tests: the tests stay off the network (testenv.Offline in their TestMains) — after 555f5706 (#1016) routing reads a plan key's quota windows, and a test whose provider has a vendor's real base URL and a made-up key would ask the vendor in the background. Routing's own read (keyAllowance, allowances) was already stubbed in routing_test.go's init, and a survey with a dial guard found no plan-key dial, but it found others: every plugin host's first start fetched models.dev's catalog (65 dials in the gateway's plugin tests), and provider tests asked api.github.com (TestSubscriptionUsageEachAccount, TestSubscriptionUsageServesStale), chatgpt.com (TestCopilotQuotaWithoutEditorsSignIn), server.codeium.com (TestDevinSignedInOnlyInMagpie) and decide.test (TestModelTest) with made-up accounts; each passed whether the vendor answered or not. internal/testenv's Offline wraps http.DefaultTransport's DialContext (so http.DefaultClient and every transport cloned from it later): localhost, *.localhost, loopback and unspecified addresses are dialled, anything else is refused before it is looked up, said once per address on stderr ("testenv: a test asked models.dev:443, refused: tests stay off the network"). internal/gateway's TestMain and internal/provider's (kiro_main_test.go, in RunIn's setup) call it before the tests run. TestLocal (the address table) and TestOffline (a loopback httptest server answers through DefaultClient; models.dev is refused through it and through a Clone made after the call) pass -race -count=3. go test -race -tags nogui -timeout 30m ./internal/testenv ./internal/provider ./internal/gateway passes under a temp HOME (gateway 516 s). The gateway and provider test binaries were then run under sandbox-exec with every non-loopback network-outbound denied, with the kernel's Sandbox reports streamed: both PASS, Offline refused only the six addresses above, and the kernel reported no denial, so no transport outside DefaultTransport reaches the network either. go vet, gofmt -l, linux and windows (nogui) builds pass. Tests only.
2026-10-06 21:08:57 +08:00