mirror of
https://github.com/hanzoai/iam.git
synced 2026-11-19 18:37:16 +00:00
41 lines
1.7 KiB
Go
41 lines
1.7 KiB
Go
// Copyright 2026 Hanzo AI, Inc.
|
|
// SPDX-License-Identifier: MIT OR Apache-2.0
|
|
|
|
// Package testhttp is the ONE way a test drives the registered router.
|
|
//
|
|
// It exists for a single value: the timeout. fiber's App.Test defaults to
|
|
// TestConfig{Timeout: time.Second}, and this repo's hot paths are argon2id —
|
|
// deliberately expensive, and under -race (the detector costs roughly an order of
|
|
// magnitude) on a box running `go test ./...` with packages in parallel, a signup,
|
|
// an onboard, a registry token or a SCIM create routinely exceeds one second. The
|
|
// suite then fails with `i/o timeout` on whichever package lost the CPU race, in a
|
|
// different place each run.
|
|
//
|
|
// That is the worst failure a gate can have: it is red without a defect, so it
|
|
// teaches everyone to re-run until green, and a real regression hiding among the
|
|
// noise gets re-run away with it. The KDF cost is deliberate and must not be
|
|
// lowered for tests; the one-second ceiling was simply never chosen.
|
|
//
|
|
// So the ceiling is stated once, here, generously — long enough that only a
|
|
// genuine hang trips it, short enough that a deadlock still fails rather than
|
|
// blocking the run forever. Every suite calls Do; none carries its own timeout,
|
|
// so none can drift.
|
|
package testhttp
|
|
|
|
import (
|
|
"net/http"
|
|
"time"
|
|
|
|
"github.com/zap-proto/zip"
|
|
)
|
|
|
|
// timeout bounds one in-process test request. It is not a latency budget — no
|
|
// assertion depends on it — only a deadlock guard.
|
|
const timeout = 2 * time.Minute
|
|
|
|
// Do issues req against app's registered router in process and returns the
|
|
// response. The caller closes the body.
|
|
func Do(app *zip.App, req *http.Request) (*http.Response, error) {
|
|
return app.Test(req, zip.TestConfig{Timeout: timeout, FailOnTimeout: true})
|
|
}
|