[APX403]

With autopatch on Fiber, a PATCH whose PUT handler takes more than a second brings the server down:

app := fiber.New()
api := humafiber.New(app, huma.DefaultConfig("My API", "1.0.0"))

huma.Get(api, "/things/{id}", getThing)
huma.Put(api, "/things/{id}", func(ctx context.Context, input *struct {
	ID   string `path:"id"`
	Body Thing
}) (*ThingResponse, error) {
	time.Sleep(1200 * time.Millisecond) // a slow write
	return &ThingResponse{Body: &input.Body}, nil
})

autopatch.AutoPatch(api)
log.Fatal(app.Listen(":8888"))

The first PATCH ends the process with panic: i/o timeout (panic: test: timeout error 1000ms with humafiber.NewV2).

autopatch sends its internal GET and PUT through the adapter's ServeHTTP (autopatch.go#L317, #L449). Both Fiber adapters implement it with Fiber's Test, which gives up after one second by default, and panic on any error it returns (humafiber.go#L261-L269, humafiber_v2.go#L263-L271). Fiber doesn't recover panics by default, so the panic stops the server.

ServeHTTP now turns the timeout off (TestConfig{Timeout: 0} on v3, -1 on v2), so it waits for the handler like the net/http adapters do, and answers any other Test error with a 500 instead of panicking. The 500 has a generic body, since the error from Test can quote the raw request.

This is the timeout and the panic left out of #1140.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 94.72%. Comparing base (4035252) to head (556d0b8).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1152      +/-   ##
==========================================
- Coverage   94.73%   94.72%   -0.01%     
==========================================
  Files          25       25              
  Lines        5200     5193       -7     
==========================================
- Hits         4926     4919       -7     
  Misses        269      269              
  Partials        5        5              

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[robtme]

Thanks, this looks good. I checked the Fiber source: Timeout: 0 on v3 and -1 on v2 are the right ways to turn the timeout off (0 on v2 would time out immediately). With the adapter change reverted, the new tests panic with i/o timeout, so they cover the bug.