[xen-tools] Re: Outstanding bugs/requests for 3.0?

[Steve Kemp]

On Tue, Nov 28, 2006 at 06:28:27PM +0100, Henning Sprang wrote:

> I'd need to try using it to get a feeling, but it Sounds good so far,
> does what I wanted.

  Cool.

> O.K., in that case, --hostname would only be a requirement when
> --config is not set, and an error would thrown if after reading them
> all no hostname would be specified?

  Yes, as things currently stand :

    $CONFIG{'hostname'} is set if present global configuration file.

    $CONFIG{'hostname'} is set if --hostname=xx is used on the command line.

   Once the command line processing is updated to have an optional
  reading of a new file:

    $CONFIG{'hostname'} is set/updated if present in this new file.

  Only after all three steps are completed will we test to make sure
 we have a directory/lvm partition setup, a hostname, etc.

> Maybe --config is not a sensible enough name 'cause it's often used as
> a "replacement" for the main config file, not an addition/overloading
> config file.  I have no better idea right now

  I'm afraid I don't either.  "--template" would be the obvious one
 but since that is already in use ..

  The way things will work is that the global file is read, 
 optionally arguments are overwritten by the command line, and again
 optionally the new configuration file will overwrite any settings.

  After all that we'll have a set of options which will be used.

> Hmm, I don't understand exactly what you mean here. My idea with
> template scripting was only a thought if my wish for the "presets"
> solved above could somehow be done with this

  I misunderstood your mail.  I think I understood you to mean
 that you wanted to modify the xen-template files which are created in
 /etc/xen - but now I see that you meant template as somethign which
 soley referred to supplying arguments to xen-create-image.

> >  It seems redundant to me to specify --hostname=xxx rather than
> > just "xxx". Since we're only ever talking about hostnames/image names.
> >  I could overload the script to accept both if you liked. I have no
> > preference either way.
> 
> Uff, I am not usability expert enough to decide this. On the one hand,
> having both work with the same options seems logical - on the other
> hand, having xen-delete-image understand both might be a strange
> ambiguity.

  Yes.  I'll add both I think, since it is unlikely that many
 additional arguments will ever be added to the script I dont think
 being "forced" to maintain both approaches will come to cause problems
 in the future.

> Yeah, I know you aren't responsible for the xen kernel packages :)
> I have to write another report for them anyway, as I need nfs root and
> ip autoconfig in the xen domU kernel to use it with fai anyway...
> (only then xen-tools xen-kernel "autosensing" would be useful for
> me...)

  :)

  If you file a bug I'll post a simple implementation which the
 maintainers can approve/reject as they see fit.  I think it would
 be a nice addition but I can live without it.

  Expect the --config + --hostname changes to appear in CVS sometime
 this evening, and thanks for the feedback.

Steve
-- 
http://www.steve.org.uk/