On 09/02/2011 05:34 AM, Daniel Veillard wrote:
On Fri, Sep 02, 2011 at 07:23:38PM +0800, Osier Yang wrote:
> 于 2011年09月02日 18:21, Daniel Veillard 写道:
>> On Wed, Aug 31, 2011 at 04:26:13PM +0800, Osier Yang wrote:
>>> The modified function fallbacks to use virDomainBlockStats if
>>> virDomainBlockStatsFlags is not supported by the hypervisor driver.
>>> If the new API is supported, it will be invoked instead of the
>>> old API.
>> Again I would use a new command and keep the existing code
>> as is,
>>
>> Daniel
>>
>
> What does "again" here mean? as I couldn't find another place which
> introduce new code. :-)
Well I would not modify cmdDomblkstat and instead create a new command
and a new function for the new API
> And per the new API has some same fields with old API, and seems
> all of virsh commands try to fallback to old API if new API is introduced.
yes but the new API will provide informations in a different order
and is potentially more expensive, so I'm not sure I really want to use
the new API for the old command
Maybe something like "domblkfullstat"...
But if someone else disagrees with me I'm fine being in the minority :-)
Personally, I'd like to keep a single virsh command for both APIs, just
like 'virsh migrate' handles both virDomainMigrate and
virDomainMigrate2. I don't know if it's better to default to the old or
the new API, but it is easy enough to provide a flag that swaps the
default to call the alternate API.
--
Eric Blake eblake(a)redhat.com +1-801-349-2682
Libvirt virtualization library
http://libvirt.org