subprocess masks the call stacks of children

[buhman]

I should be able to interrupt hubble and get a stack trace all the way down to what each "child" is doing.

I think there are at least two ways to solve this:

• we could add a new standardized (setuptools entrypoint) loading mechanism, where, instead of starting a new unrelated python interpreter, the "command" is run in some introspectable thread/future.
• hubble could explictly handle signals, and pass them to child processes

[thrawn01]

we could add a new standardized (setuptools entrypoint) loading mechanism, where, instead of starting a new unrelated python interpreter, the "command" is run in some introspectable thread/future.

This assumes hubble is only going to be used with only python application, I don't think that is the case. Don't really like this idea.

hubble could explictly handle signals, and pass them to child processes

This is something we should support 1.2