Re: [libvirt] [PATCH] Introduce virt-console

[John Levon <john levon sun com>]

On Thu, Jan 15, 2009 at 02:04:56PM +0000, Daniel P. Berrange wrote:

> > > You explicitly break virt-install by doing this.
> > 
> > Break how?
> 
> It relies on virsh console exiting when the domain shuts down.

I'm confused, we've been using virt-install without apparent problems
for quite some time like this?

> > > Have virt-console provide the more sensible default auto-reconnect
> > > semantics, and make 'virsh console' call it with a flag to turn
> > > this off to preserve existing semantics & not break users like
> > > virt-install.
> > 
> > This is horrible IMHO - the user has to run some strange command instead
> > of virsh like they use for everything else? I'd rather not
> > auto-reconnect than this.
> 
> Its not so strange in the context of all the other virt commands we
> have, in particular in relation to virt-viewer
> 
>   virt-viewer  - graphical console

The existence of the other tools doesn't make it sensible for obscure
behaviour changes between two ways of doing the exact same thing.

The /only/ reason I'm introducing virt-console is to correctly implement
the least priv support.

virt-manager aside (which is more or less a graphical
virsh/virt-install), all of these should have indeed been virsh
commands. It's horrible that I have to use a separate tool with a
different interface to install my guest, then virsh to manage it.

regards,
john