Skip to content

worker: handle detached `MessagePort` from a different context

Rodrigo Muino Tomonari requested to merge github/fork/juanarbol/juan/fix-worker-issue into main

When worker.moveMessagePortToContext is used, the async handle associated with the port, will be triggered more than needed (at least one more time) and with null data. That can be avoided by simply checking that the data is present and the port is not detached.

Fixes: https://github.com/nodejs/node/issues/49075

Note aside, this could be fixed as well by the next diff:

diff --git a/src/node_messaging.cc b/src/node_messaging.cc
index 00e1db3bab..b46ae022ce 100644
--- a/src/node_messaging.cc
+++ b/src/node_messaging.cc
@@ -741,9 +741,6 @@ MessagePort* MessagePort::New(
     // but it's better to be safe than sorry.)
     Mutex::ScopedLock lock(port->data_->mutex_);
     port->data_->owner_ = port;
-    // If the existing MessagePortData object had pending messages, this is
-    // the easiest way to run that queue.
-    port->TriggerAsync();
   } else if (sibling_group) {
     sibling_group->Entangle(port->data_.get());
   }

IMO that could change how the workers behave, then I prefer a "handle case" approach.

Merge request reports