rfq: store outgoing request before sending message

We store the outgoing request in the request map, so we can interpret
any response and map it to the original request. We do this before
sending the message, because otherwise we could potentially receive
the response while we're waiting for the atomic lock to be released,
which would lead to us not knowing about the request in the first
place, and we couldn't map the response.

Author: guggero

rfq/stream.go

@@ -154,13 +154,12 @@ func (h *StreamHandler) HandleOutgoingMessage(
 	ctx, cancel := h.WithCtxQuitNoTimeout()
 	defer cancel()
 
-	err = h.cfg.PeerMessenger.SendCustomMessage(ctx, lndClientCustomMsg)
-	if err != nil {
-		return fmt.Errorf("unable to send message to peer: %w",
-			err)
-	}
-
-	// Store outgoing requests.
+	// We store the outgoing request in the request map, so we can interpret
+	// any response and map it to the original request. We do this before
+	// sending the message, because otherwise we could potentially receive
+	// the response while we're waiting for the atomic lock to be released,
+	// which would lead to us not knowing about the request in the first
+	// place, and we couldn't map the response.
 	switch msg := outgoingMsg.(type) {
 	case *rfqmsg.BuyRequest:
 		h.outgoingRequests.Store(msg.ID, msg)
@@ -173,6 +172,12 @@ func (h *StreamHandler) HandleOutgoingMessage(
 			msg.ID.String())
 	}
 
+	err = h.cfg.PeerMessenger.SendCustomMessage(ctx, lndClientCustomMsg)
+	if err != nil {
+		return fmt.Errorf("unable to send message to peer: %w",
+			err)
+	}
+
 	return nil
 }