My goal is to use the launch framework for configurations that attach to agents without launching a process. I started looking at The Agent Discoverer as a sufficiently close example. Unfortunately I got confused by the logic that is executed for this case (I use the "Java Application" launch configuration):
The confusing thing is that, even though the Agent Discoverer is really an agent object, in this particular case the agent returned to the primary launch delegate must be null. Then, even though no process is associated with it, a dummy process of class CustomProcess must be created to circumvent some problem with the RAC. Then later a dummy agent of class DummyAgent is created to simulate an active event, in a rather obscure place/logic path. What I would expect in this case is just the opposite, namely that a non-null agent is returned to the primary launch delegate, and that the logic of how to launch a process-less data collector is implemented in the primary launch delegate itself, without a need for a dummy agent and a dummy process. I tried modifying the CustomDelegate example documented under "Registering a Custom Data Collection Mechanism with TPTP's Launch Configuration Framework " by implementing IDataCollectorAgentLauncher instead of subclassing off AbstractProcessLauncher but then no tree node representing an agent shows up in the profile monitor. Thanks in advance for any advice, Henk Aling Wind River _______________________________________________ tptp-tracing-profiling-tools-dev mailing list tptp-tracing-profiling-tools-dev@eclipse.org https://dev.eclipse.org/mailman/listinfo/tptp-tracing-profiling-tools-dev
