Community
Participate
Working Groups
When doing two provisioning operations in a row, I get the following exception (on the second operation): org.eclipse.equinox.internal.p2.touchpoint.natives.ClosedBackupStoreException: Can not perform restore on closed backup store at org.eclipse.equinox.internal.p2.touchpoint.natives.BackupStore.restore(BackupStore.java:406) at org.eclipse.equinox.internal.p2.touchpoint.natives.LazyBackupStore.restore(LazyBackupStore.java:53) at org.eclipse.equinox.internal.p2.touchpoint.natives.NativeTouchpoint.rollback(NativeTouchpoint.java:108) at org.eclipse.equinox.internal.p2.engine.EngineSession.rollback(EngineSession.java:194) at org.eclipse.equinox.internal.p2.engine.Engine.perform(Engine.java:82) at org.eclipse.equinox.internal.p2.engine.Engine.perform(Engine.java:44) at org.eclipse.equinox.internal.p2.engine.Engine.perform(Engine.java:48) After some digging, I know what's going on. 1. The first operation creates a backup store and then eventually discards it (in the commit) 2. However, the backup store is cached in a weak hash map (static) in NativeTouchpoint 3. The next operation needs to use the store, and it finds the cached one 4. It grabs it, and tries to use the store 5. Boom... already closed. I think we need to clear the backup store cache during the commit. I'll put a patch together.
Created attachment 173862 [details] fix v1 This clears the profile state stored in the native touchpoint during the commit.
I've released this to head. What do you think of this for 3.6.1? It's pretty safe since all we're doing is proactively clearing a weak hash map when we're done with it.
