[libvirt] [PATCH v2]

Stefan Berger stefanb at us.ibm.com
Thu Apr 8 11:16:00 UTC 2010


Daniel Veillard <veillard at redhat.com> wrote on 04/08/2010 03:54:07 AM:


> 
> Please respond to veillard
> 
> On Wed, Apr 07, 2010 at 05:34:50PM -0400, Stefan Berger wrote:
> > This patch adds new test cases to the existing nwfilter test program 
and
> > adds a test script that must be run while a VM is running. This test
> > script verifies that input network filter XML creates expected
> > user-defined tables and rules on ebtables, iptables & ip6tables layer
> > and verifies their content against expected content. The idea is that
> > these tools always return exactly the same output when displaying the
> > content of a user-defined table so that the diff tool can be used for
> > simple text comparison. All supported protocols have at least one test
> > case. 
> > This test program is not run automatically since it requires a running
> > VM.
> 
>   I'm wondering a bit about this.
> 
> So far all our embedded tests from "make tests" are supposed to be
> agnostic on the machine state, and I wonder if it ain't preferable to
> kepp them that way. For example we have no real save/restore or
> migration tests here because like the nwfilter rukes testing that would
> require actual VM runnings.
> To me it seems this test should rather be targetted to the TCK, the
> separate test suite for libvirt:
> 
>   http://libvirt.org/git/?p=libvirt-tck.git
> 
> that's really the place where tests relying on actual system behaviour
> should be implemented, and IMHO the more people start to look at it the
> better.
> Dan sent an introduction when he created the project
>   http://www.mail-archive.com/libvir-list@redhat.com/msg12703.html
> 
> and an update recently:
>   http://www.mail-archive.com/libvir-list@redhat.com/msg21503.html
> 
> so some of the hairy shell scripting for this test could be replaced
> with (less hairy ?) Perl testing, but in a framework really intended
> to have actual running domains.
> 
>   I understand that the current version of the test is not supposed to
> run if there is no domain target, but really it break the rule that
> libvirt "make tests" should not be dependant on the machine state,

Ok, I'll adapt it for the TCK project.

   Stefan

> 
> Daniel
> 
> -- 
> Daniel Veillard      | libxml Gnome XML XSLT toolkit  
http://xmlsoft.org/
> daniel at veillard.com  | Rpmfind RPM search engine http://rpmfind.net/
> http://veillard.com/ | virtualization library  http://libvirt.org/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/libvir-list/attachments/20100408/02c7f398/attachment-0001.htm>


More information about the libvir-list mailing list