Bug #2169
Updated by Henning Blohm 2 months ago
The following sequence of events may lead to untracking of a still valid resource: 1. In NamespaceImpl _getResourceHandleImplNonNull we find a vh!=null && rh==null and (in another loop) put the vh directly into the invQ. 2. A subsequent tick() removes the vh.getKey() from the resource map and we can bind a new entry Later: 3. In another tick we ingest the gc queue from the resource map into invQ. Because only now the GC noted the collected value, we get the same vh from above again in the invQ and the tick removes the key again from the resource map. h2. Solution * In tick make sure to check for value holders that are still bound in the map * Only remove from the map, if it is still bound.