Killing an agent doesn't phase RLViz

[GoogleCodeExporter]

If you run an experiment with RLVizApp and an agent, and then you kill the
agent, RLVizApp remains running, and lets you click stuff.

Should find a way to detect the disconnect and die.  Maybe RL-Glue should
be dying if the agent dies.

We should fix this if easy and if not, bump to later milestone.

Original issue reported on code.google.com by brian.ta...@gmail.com on 24 Oct 2007 at 10:28

[GoogleCodeExporter]

Shouldnt calling any of the agent functions break this? ie: agent_step()... ill 
look into this

Original comment by rad...@gmail.com on 27 Oct 2007 at 1:18

• Changed state: Started

[GoogleCodeExporter]

Original comment by markp...@gmail.com on 29 Oct 2007 at 10:41

• Added labels: Milestone-Release1.0
• Removed labels: Milestone-TrainingRelease

[GoogleCodeExporter]

It seems to me that this shouldnt break it, unless we want to implement a 
timeout. From what i gather is 
happeneing, we make a call to agent, then sit there till we get a response... 
no agent = no response. and thus we 
hang. If we want to add some sort of wait time length, and then if it takes 
longer than this time, we assume the 
agent is dead and we close.

Original comment by rad...@gmail.com on 10 Jan 2008 at 10:54

• Changed state: Accepted