[libvirt] libvirt-1.0 fails virdrivermoduletest

[Daniel P. Berrange]

On Wed, Nov 21, 2012 at 11:56:54AM +0100, Ján Tomko wrote:
> On 11/21/12 10:06, Daniel P. Berrange wrote:
> > On Wed, Nov 21, 2012 at 09:22:25AM +0100, Ján Tomko wrote:
> >> On 11/20/12 20:36, Toralf Förster wrote:
> >>>
> >>> n22 /var/tmp/portage/app-emulation/libvirt-1.0.0/work/libvirt-1.0.0/tests # VIR_TEST_DEBUG=1 ./virdrivermoduletest
> >>> TEST: virdrivermoduletest
> >>>  1) Test driver "storage"                                             ... OK
> >>>  2) Test driver "nodedev"                                             ... OK
> >>>  3) Test driver "secret"                                              ... OK
> >>>  4) Test driver "nwfilter"                                            ... OK
> >>>  5) Test driver "interface"                                           ... OK
> >>>  6) Test driver "qemu"                                                ... FAILED
> >>>  7) Test driver "lxc"                                                 ... FAILED
> >>>  8) Test driver "uml"                                                 ... OK
> >>
> >> The "qemu" and "lxc" driver tests depend on the "network" driver, which
> >> is missing. Enabling the virt-network USE flag should fix the tests and
> >> allow you to configure the network with libvirt.
> > 
> > Hmm, if we allow the possibility of biulding without the network
> > driver, then we should make sure the tests pass. I'm curious as
> > to what the actual failure is there.
> > 
> > Daniel
> 
> The failure is virdrivermoduletest not being able to load the "network"
> module, which it has listed as a dependency for "qemu" and "lxc".
> Removing this dependency fixes the test for me. [1]
> 
> virnetsockettest passes in make check but fails when run as
> ./virnetsockettest because it uses the dummy ssh in tests/ in the first
> case and the system-wide ssh in the other one. It should work as:
> VIR_TESTS_DEBUG=1 PATH=.:$PATH ./virnetsockettest

Yep, that's a false failure.

> We could also suppress the "statement with no effect" warning for no-op
> functions from network/bridge_driver.h to make build with -Werror possible.

Yep, that should be done. We should be warning-free in all configurations

> diff --git a/tests/virdrivermoduletest.c b/tests/virdrivermoduletest.c
> index 9c14e79..447e18e 100644
> --- a/tests/virdrivermoduletest.c
> +++ b/tests/virdrivermoduletest.c
> @@ -83,10 +83,10 @@ mymain(void)
>      TEST("interface", NULL);
>  #endif
>  #ifdef WITH_QEMU
> -    TEST("qemu", "network");
> +    TEST("qemu", NULL);
>  #endif
>  #ifdef WITH_LXC
> -    TEST("lxc", "network");
> +    TEST("lxc", NULL);
>  #endif
>  #ifdef WITH_UML
>      TEST("uml", NULL);

If you make that change conditional on #ifdef WITH_NETWORK, then we
could apply it

Daniel
-- 
|: http://berrange.com      -o-    http://www.flickr.com/photos/dberrange/ :|
|: http://libvirt.org              -o-             http://virt-manager.org :|
|: http://autobuild.org       -o-         http://search.cpan.org/~danberr/ :|
|: http://entangle-photo.org       -o-       http://live.gnome.org/gtk-vnc :|