Re: [Autotest] Adding kvm_subprocess

[Lucas Meneghel Rodrigues <lmr@redhat.com>]

Hi Michael,

On Thu, 2009-06-18 at 11:27 -0400, Michael Goldish wrote:
> kvm_subprocess is a little weird in that it does two different things --
> handling of both non-interactive subprocesses and SSH sessions.
> With this approach I don't think we need to write an SSH subclass of
> kvm_spawn because it already does a lot of SSH stuff.

Fair enough. I just happen to like the approach took by pexpect folks to
separate SSH implementation on a subclass, because the generic problem
itself is 'interacting with interactive programs on a programatic way',
and handling SSH connections is a subset of this problem.

> If we do any subclassing at all -- I suggest that we remove all the SSH
> stuff from kvm_spawn and put it in a subclass somehow, so that kvm_spawn
> natively only handles non-interactive subprocess (with _tail(),
> get_output(), get_status() etc), and the subclass does everything else
> (read_up_to_prompt(), get_command_status_output() etc).  ssh_login()
> can remain an external function that creates and returns a kvm_spawn
> object.
> This will have to be done carefully because each 'user' of the kvm_spawn
> server needs a named pipe of its own, which will have to be handled by
> the constructor.
> If you think this is a good idea I'd rather make the necessary changes
> to kvm_subprocess myself.

Sounds good to me, I am totally OK with it.

> Does this make sense to you, or did I misunderstand what you meant by
> writing an SSH subclass?

No, you've nailed it. Sorry for the delay in answering.