Re: [Xen-devel] [PATCH] [RFC] Fix RegEx Issues with xendomains for both SXP and JSON outputs of lx list -l

[Ian Campbell <Ian.Campbell@xxxxxxxxxx>]

On Tue, 2013-06-25 at 16:39 +0100, Ian Murray wrote:
> 
> 
> 
> 
> 
> 
> >________________________________
> > From: David Sutton <kantras@xxxxxxxxx>
> >To: Ian Murray <murrayie@xxxxxxxxxxx> 
> >Cc: Ian Campbell <Ian.Campbell@xxxxxxxxxx>; George Dunlap 
> ><george.dunlap@xxxxxxxxxxxxx>; Joshua Tuttle <jtuttle@xxxxxxxxx>; Ian 
> >Jackson <Ian.Jackson@xxxxxxxxxxxxx>; "xen-devel@xxxxxxxxxxxxx" 
> ><xen-devel@xxxxxxxxxxxxx> 
> >Sent: Tuesday, 25 June 2013, 16:14
> >Subject: Re: [Xen-devel] [PATCH] [RFC] Fix RegEx Issues with xendomains for 
> >both SXP and JSON outputs of lx list -l
> > 
> >
> >
> >Reply below,
> >
> >
> >On Tue, Jun 25, 2013 at 9:55 AM, Ian Murray <murrayie@xxxxxxxxxxx> wrote:
> >
> >
> >>Date: Sat, 22 Jun 2013 13:38:11 +0100
> >>Message-Id: <1371904691-9842-1-git-send-email-murrayie@xxxxxxxxxxx>
> >>
> >>
> >>
> >>>>  As was suggested by David Sutton, there are probably more issues with
> >>>>  this script than has been covered so far, but this change makes
> >>>>  suspension and restoration work for me, at least. He suggests there
> >>>>  are problems with zombies, etc.
> >>>
> >>> You mean the handling of zombies by this script, rather than it somehow
> >>> causing zombies?
> >>
> >>Yes, handling of zombies.
> >>
> >>
> >>
> >Yes, it was handling of zombies - in particular the xendomains script
> was making use of the state information of the various domains, which
> isn't part of the information exported  by xl list -l in either format
> >
> 
> I am guessing, but I would say the script used to use xl/xm list
> without the -l and thiis is where $state was set originally.

That sounds like a good bet.

>  As you point this isn't available from xl list -l (should it be?). If
> I was approaching this as a re-write, the first thing I would do would
> be to devise some test cases for all the things this script is
> supposed to deal with. I don't know how to create zombies at will, for
> example. Perhaps the hypervisor gurus do. Until that can be done, this
> script is going to be hard to test properly. Just my opinion.

If you can create zombies then that is a bug I think, which we would fix
and break your test case ;-)

If I were going to write test cases for this script I would start by
collecting a library of "xl list -l" outputs (both real world and
handwritten) and devise a test mode which sucked this in instead of
running the command and operated in a dry-run mode.

This is effectively what we've done, to a greater or lesser extent
(mostly lesser...), for xl disk stanza parsing
(tools/libxl/check-xl-disk-parse) or pygrub (tools/pygrub/examples/).
It's nice because when someone reports a bug you can just add their
output to the set of things you test.

Ian.


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel