v0.25.0 netacl and fixes

internal/sink/sink.go

@@ -36,9 +36,9 @@ type SinkStats struct {
 	Details           map[string]any
 }
 
-// FlushBound is how long a network sink's Stop gives its clients to take
-// what is queued: write_timeout_ms, at most core.SinkFlushTimeout, which a
-// reload also waits out for a client that stopped reading.
+// FlushBound is how long a network sink's Stop gives what is queued to reach
+// its clients or its downstream: its write or request timeout, at most
+// core.SinkFlushTimeout, which a reload also waits out for a stalled peer.
 func FlushBound(writeTimeout time.Duration) time.Duration {
 	if writeTimeout > 0 && writeTimeout < core.SinkFlushTimeout {
 		return writeTimeout
@@ -46,6 +46,24 @@ func FlushBound(writeTimeout time.Duration) time.Duration {
 	return core.SinkFlushTimeout
 }
 
+// FlushContext is ctx ending FlushBound(timeout) after done closes: a sink
+// that dials runs its loop under it, so Stop leaves its queue that long.
+func FlushContext(ctx context.Context, done <-chan struct{}, timeout time.Duration) (context.Context, context.CancelFunc) {
+	ctx, cancel := context.WithCancel(ctx)
+	go func() {
+		select {
+		case <-done:
+			select {
+			case <-time.After(FlushBound(timeout)):
+			case <-ctx.Done():
+			}
+		case <-ctx.Done():
+		}
+		cancel()
+	}()
+	return ctx, cancel
+}
+
 // Drain takes the payloads queued in input without waiting: a network sink's
 // last events, which each client then writes on its own.
 func Drain(input <-chan core.TransportEvent) [][]byte {