[gql_websocket_link] Crash events on host lookup

[bverhagen]

I switched to gql_websocket_link to connect to my GraphQL backend (before I used hasura_connect). Since this switch, Firebase Crashlytics reports:

WebSocketChannelException: WebSocketChannelException: SocketException: Failed host lookup: '<myhost>' (OS Error: No address associated with hostname, errno = 7)

I think this is due to the connection being interrupted by me switching off mobile and wifi connectivity. So, obviously, the actual error makes sense. However, I don't think the app should crash nor be reported on Crashlytics when I do.

Currently, the connection is implemented as:

final link = WebSocketLink(
      "wss://<myhost>",
      autoReconnect: true,
      inactivityTimeout: const Duration(seconds: 5),
      initialPayload: () async {
        final authHeaders = await getAuthHeaders();
        return {
          'headers': {
            'x-hasura-role': role.id,
            ...authHeaders,
          }
        };
      },
    );

I thought switching to the channelGenerator argument (with a null url), I would be able to wrap the WebSocketChannelException in a try/catch clause to avoid this. However, somehow, the SocketException keeps on escaping my try/catch and hence keeps on crashing the app:

final link = WebSocketLink(
      null,
      channelGenerator: () {
        while(true) {
          try {
            return WebSocketChannel.connect(Uri.parse("wss://<myhost>"));
          } catch(_) {

          }
        }
      },
      autoReconnect: true,
      inactivityTimeout: const Duration(seconds: 5),
      initialPayload: () async {
        final authHeaders = await getAuthHeaders();
        return {
          'headers': {
            'x-hasura-role': role.id,
            ...authHeaders,
          }
        };
      },
    );

My main/AndroidManifest.xml contains the android.permission.INTERNET permission:

<manifest xmlns:android="http://schemas.android.com/apk/res/android"
    package="eco.futureproof.app">
    <uses-permission android:name="android.permission.FOREGROUND_SERVICE" />
    <uses-permission android:name="android.permission.INTERNET" />
    <application ...>
      <my application stuff>
    </application>
 </manifest>

Even though I really doubt I am the first one to see this issue, I do not find any related issues. Can you point me to an example or the typical solution to fix this?

[bverhagen]

I just realized that, around the same time, I switched on fatal and async error recording for Firebase Crashlytics (which may explain why I did not see this happen with hasura_connect):

FlutterError.onError = FirebaseCrashlytics.instance.recordFlutterFatalError;

PlatformDispatcher.instance.onError = (error, stack) {
  FirebaseCrashlytics.instance.recordError(error, stack, fatal: true);
  return true;
};

I assume it is the async error recording that is the culprit here. Still, is there a way to catch it?

[bverhagen]

In the meantime I found this: https://github.com/dart-lang/web_socket_channel/issues/38.

I am going to test whether the suggested solution of awaiting socket.ready will allow me to catch the error in the channelGenerator callback.

[bverhagen]

Preliminary testing seems to indicate this fixes it for initial connection errors:

channelGenerator: () async {
        while (true) {
          final socket = WebSocketChannel.connect(
              Uri.parse("wss://$apiDomain/v1/graphql"));
          try {
            await socket.ready;
            developer
                .log("Successfully connected with websocket to '$apiDomain'!");
            return socket;
          } catch (e, stacktrace) {
            developer.log("Error connecting to websocket: $e");
            await Future.delayed(const Duration(seconds: 5));
          }
        }
      }

Shouldn't the default channelGenerator be robust against this to?

Additionally, returning the errored socket, also results in an uncaught async error later on in its usage.