Some Eclipse Foundation services are deprecated, or will be soon. Please ensure you've read this important communication.

Bug 333267

Summary:	[Workbench] UILockListener overwritten in LockManager#setLockListener
Product:	[RT] RAP	Reporter:	Ralf Sternberg <rsternberg>
Component:	Workbench	Assignee:	Project Inbox <rap-inbox>
Status:	NEW ---	QA Contact:
Severity:	normal
Priority:	P3	CC:	ivan, wolfgang.pedot
Version:	1.4
Target Milestone:	---
Hardware:	PC
OS:	Linux
Whiteboard:

Description Ralf Sternberg

2010-12-28 11:09:50 EST

Since the org.eclipse.core.internal.jobs.LockManager can only keep one single lock listener, different sessions in RAP overwrite each other's listeners. As a result, only the one Workbench instance that has been started latest, will receive aboutToWait() and aboutToRelease() notifications from the JobManager. See Workbench#runUI, line 2545.

Comment 1 Wolfgang Pedot

2016-08-29 11:38:52 EDT

This also keeps the Display instance from being GCed until the locklistener is overwritten by a new one.

Is using "allowDeadLock" a feasible workaround for this? The name of that parameter sure is scary...

Comment 2 Ivan Furnadjiev

2016-08-30 03:17:44 EDT

(In reply to Wolfgang Pedot from comment #1)
> This also keeps the Display instance from being GCed until the locklistener
> is overwritten by a new one.
> 
> Is using "allowDeadLock" a feasible workaround for this? The name of that
> parameter sure is scary...

Could you provide a fix through Gerrit [1]?

[1] https://eclipse.org/rap/source/ 
[2] https://wiki.eclipse.org/RAP/Git

Comment 3 Wolfgang Pedot

2016-08-30 03:57:00 EDT

Maybe I was not perfectly clear: "allowDeadLock" is a commandline-parameter for the framework so there is no code-change. If it is set the code-block that calls setLockListener is skipped (Workbench.runUI) but I dont know enough internals to decide if its a good idea to set that parameter.