Present: Rob, Frederic, Conor, Tom and Kanna

•  Conor discussed the state of the VCS/AMC integration while working with Rich last week:
  • the major bug tackled was that VCS was not flushing messages in the queue when the VCS is restarted. The consequence was that previous messages in the dispatch stack of the layered control were still being sent off for execution.
  • Two ways to mitigate:
    1. when the AMC fires up, it will await an explicit initialization coming from the VCS
    2. In addition Rich will implement a mechanism to ensure that on a replan action, the control stack is reinitialize
  • A number of short runs with a lot of inserts/deletes done to test the AMC/VCS handshaking and messages exchange. All sans above, seems to be well.
• Tom raised the question of a test plan for VCS with all the changes in layered control. His suggestion was to run, at the very least, a leak checker.
  • Frederic has some simple graphing tool which shows memory usage over time and can provide a simple way to see if memory is being hogged without Free's taking place. He has used this on QNX to ensure his clustering code is kosher.

• • Rob suggested checking with Hans for 'spy' which appears to be equivalent to 'top'. We need to check with Hans.
• Socket related issues from last week were fixed; it appeared to be a timing issue with the read buffer getting read at a faster rate.
• We will use JIRA for any issue tracking (see http://oceana:8082/browse/AUT)
• We all had a lively discussion with Rob about a related if not directly connected to this project, of how the Gulper behavior will work. Rob's block diagram of the process is as attached to this note.

Gulper-behavior.jpg (image/jpeg)