On 03/09/2012 11:59 PM, Chris Evich wrote: > On 03/08/2012 10:06 PM, Alex Jia wrote: >> On 03/08/2012 10:23 PM, Chris Evich wrote: >>> On 03/08/2012 01:33 AM, Alex Jia wrote: >>>> On 03/08/2012 11:52 AM, guyanhua wrote: >>>>> This patch adds three test cases for "virsh capabilities" command. >>>>> >>>>> Use three cases:(1) Call virsh capabilities >>>>> (2) Call virsh capabilities with an unexpected option >>>>> (3) Call virsh capabilities with libvirtd service stop >>>>> >>>>> Signed-off-by: Gu Yanhua<[email protected]> >>>>> --- >>>>> client/tests/libvirt/tests/virsh_capabilities.py | 51 >>>>> ++++++++++++++++++++++ >>>>> 1 files changed, 51 insertions(+), 0 deletions(-) >>>>> create mode 100644 client/tests/libvirt/tests/virsh_capabilities.py >>>> How to make sure output information of virsh capabilities are correct? >>>> for example,<cpu>,<power management>,<topology>,<guest> XML block >>>> information etc. >>> Good point, I agree some minimal level of checking should be done. My >>> suggestion would be to keep it very simple, maybe use xml.dom.minidom to >> Chris, agree with you, we should check some important points at least. >>> check for the top-level tags, and call that 'good enough'. I don't >>> think we want to be too pedantic about checking correctness because it >>> could make maintenance a pain in the future. Heck, maybe just a few >> But, I don't think so, it's our job to ensure that virsh commands >> correct and >> available, if we can't make sure it's completely correct, for example, >> the current >> hyperviosr is xen , however, virsh capabilities says domain type='kvm' >> and 'qemu'. >> so we still need to check some important points although they probably >> are too >> pedantic :-) > Yes, I think we're mostly on the same page, it's a balance. My concern > over "too much" correctness checking, is to avoid writing ourselves into > a maintenance headache if the output schema changes, libvirt is running > on a unknown host type (by today's standards), or new capabilities are > added to libivrt itself. > > I think "correct" can mean, the output XML parses w/o throwing > exceptions, and it contains some minimal expected capabilities info. > (and not likely to change much between versions or across hosts). For > example, check that Host has a non-empty UUID tag& data value, CPU arch > data != "Orangutan", and there's at least one CPU feature tag (w/o > caring about the value). > > Sound good? > Okay, It sounds good.
Thanks, Alex _______________________________________________ Autotest mailing list [email protected] http://test.kernel.org/cgi-bin/mailman/listinfo/autotest
