forward: 1006 reconnect never triggers (wrapped error vs IsCloseError), and exits with a usage dump

[willcowling]

Summary

gh webhook forward has retry logic for a 1006 abnormal closure, but it never runs. The error is wrapped before it reaches the check, and websocket.IsCloseError does not unwrap. The result is that the command exits on the first server-side disconnect — and, because the command sets neither SilenceUsage nor SilenceErrors, prints its entire help text on the way out.

In my case GitHub closes the forwarding socket roughly every five minutes, so a long-running forward dies about twelve times an hour and each death writes ~26 lines of usage output. 73% of my log is cobra help text.

The cause

webhook/forward.go, in runFwd:

for i := 0; i < 3; i++ {
    err := handleWebsocket(out, url, token, wsURL, activateHook)
    if err != nil {
        // If the error is a server disconnect (1006), retry connecting
        if websocket.IsCloseError(err, websocket.CloseAbnormalClosure) {
            time.Sleep(5 * time.Second)
            continue
        }
        ...
        return err
    }
}

handleWebsocket wraps the read error before returning it:

err := c.ReadJSON(&ev)
if err != nil {
    return fmt.Errorf("error receiving json event: %w", err)
}

and gorilla/websocket v1.5.0's IsCloseError is a bare type assertion, not errors.As:

func IsCloseError(err error, codes ...int) bool {
	if e, ok := err.(*CloseError); ok {

So the assertion fails on the wrapped error, the continue is unreachable, and runFwd returns immediately. The observed error confirms it — this is the wrapped text, not unable to connect to webhooks server, forwarding stopped, which is what exhausting the loop would produce:

Error: error receiving json event: websocket: close 1006 (abnormal closure): unexpected EOF

Suggested fix

var closeErr *websocket.CloseError
if errors.As(err, &closeErr) && closeErr.Code == websocket.CloseAbnormalClosure {
    time.Sleep(5 * time.Second)
    continue
}

and SilenceUsage: true on the command, since a runtime network error is not a usage error.

One more thing worth considering

Fixing the unwrap alone may not be enough for a long-running forwarder. i counts total disconnects across the process lifetime and is never reset after a successful session, so with a server that closes on a timer the command would still exit permanently after three closes — for me, about fifteen minutes. Resetting the counter after a session that lasted a meaningful length of time would make forward survivable as a long-running process.

Versions

• gh 2.98.0
• gh-webhook v0.2.0
• macOS 15 (arm64)