[sarankumar1325]

RunPushGatewayGatherer panics with a nil pointer dereference whenever the push gateway can't be reached (connection refused, DNS failure, timeout) and the ErrorHandler returns nil, which is what the default ErrorHandler does:

res, err := client.Do(req)
if err != nil {
	if hErr := config.ErrorHandler(...); hErr != nil {
		return hErr
	}
}
if res.StatusCode != http.StatusOK { // res is nil here

So with the default config, a push gateway that goes down for a moment crashes the application's push goroutine (and the process). The response body was also never closed, which leaks a connection on every push.

Changes:

• continue to the next tick after a client.Do error has been passed to the ErrorHandler.
• Drain and close the response body after each push.
• New test TestRunPushGatewayGatherer_unreachableGatewayDoesNotPanic: it points at a closed httptest server with an ErrorHandler that returns nil. On master it panics at prometheus.go:448. With the fix it returns context.DeadlineExceeded, and the handler sees "error sending to push gateway".

Tests: go test ./echoprometheus/, go test -race ./echoprometheus/, go vet, gofmt and staticcheck -tests=false ./echoprometheus/ are clean.

The same code was copied into labstack/echo-prometheus, where labstack/echo-prometheus#4 proposes the same fix.

This change was prepared with AI assistance and reviewed by me.