Processes are being cleaned up too quickly

[jamoore]

With the latest changes to get processor.py working on Windows, something like:

process = sr.acquireProcessor(job, 10)
process.wait()

can throw an Ice.ObjectNotExistException since by the time the wait has reached the remote python process, the subprocess is already complete, and everything cleaned up.

[jmoore]

It's not quite clear what the best method here is:

• teach users to catch the ONEE
• have a "process.keepAlive()"
• have the process stay alive "as long as possible"

The last point is quite tricky. The process could be kept alive forever since it's not bound to a session. However, obtaining the results for the process require that the session is alive, and therefore that might be a good time to pull the plug.

In the case of InteractiveProcessors, assuming a user doesn't call setDetach(true), then the death of the calling session will also close the subprocess session.

Note: importer is using setDetach(true)

[jmoore]

This only might be true. The ONEEs are being caused by Glacier2 filters (r4910). I've corrected this problem in r5294, and will re-test the refactoring I just performed to try to fix the alleged fast timeouts.

[jmoore]

Along with #1524 this seem to be fine. The glacier issues were certainly a much larger issue.