On Thu, Oct 25, 2012 at 07:16:35AM -0600, Eric Blake wrote:
On 10/25/2012 03:03 AM, Michal Privoznik wrote:
>
> While I agree that this design is broken I don't think we can do this.
> Okay, for now we only support 0; but what if in the future we invent a
> new flag? With current virsh one is able to use it however with your
> patch he isn't.
But you could apply that argument to any number of other interfaces that
take a flags argument. virsh simply does not know how to export
arbitrary flags that were only added to newer servers - you HAVE to
upgrade your virsh to match.
>
> Therefore I'd rather see slightly different approach. Like we do for
> other broken options/arguments in virsh - hide it, don't mention it
> anywhere but keep the code.
I disagree - there's no point in keeping a hidden argument. It is a
disservice to users to make them have to pass a numeric flags value - if
they know they are talking to a new enough server that supports a new
flag, then they should be able to upgrade to a new enough virsh that
exposes that new flag as a human-readable option name, or directly code
their task using C or python bindings instead of bothering with virsh.
I think the patch is fine as-is.
+1
Dave
--
Eric Blake eblake(a)redhat.com +1-919-301-3266
Libvirt virtualization library
http://libvirt.org
--
libvir-list mailing list
libvir-list(a)redhat.com
https://www.redhat.com/mailman/listinfo/libvir-list