@konstruktoid I don't really understand how the test suite works, but I made an attempt.

This works on Debian with libvirt, because vagrant up destroys the VM on error. With VirtualBox, Vagrant does not seem to destroy it when there is an error in the provision script. So I don't know how to include that info in the test suite.

This works on Debian with libvirt, because vagrant up destroys the VM on error. With VirtualBox, Vagrant does not seem to destroy it when there is an error in the provision script. So I don't know how to include that info in the test suite.

oh... hashicorp/vagrant#11692 (comment)
Indeed, this will make things more complicated to test.

Any feedback on how to test this? Is the test still required? This is a simple change to support troubleshooting.

@eighthave Given what I said in my comment, I'm not convinced that it can be tested at this moment since the CI is testing only vagrant+vbox. The solution would be to run the test only if testing against libvirt but I'm not sure it's easy and I don't know if other devs except me are testing with vagrant-libvirt.

tbh, I was hoping to get some feedback from others. @konstruktoid ? @ssbarnea ? Should the test be adapted to run only on vagrant-libvirt or any idea on how to test this or the test should be removed ?

I'm so sorry @eighthave, but I've completely missed the updates on this PR.
If the test is easily adapted to run only on vagrant-libvirt I'm all for it, otherwise I'd say we remove the test and push the issue to our future selfs if there's issues later on.

FWIW I'm using mostly vagrant-libvirt, in testing and production. I don't know the test harness here at all, I'm happy to add something to make it libvirt only, but I don't know what that is.

FWIW I'm using mostly vagrant-libvirt, in testing and production. I don't know the test harness here at all, I'm happy to add something to make it libvirt only, but I don't know what that is.

The test suite currently relies on the TEST_PROVIDER variable to know which provider is used so the idea would be to mark the test failure is fine on providers not libvirt with xfail not tested:

@pytest.mark.xfail(TEST_PROVIDER != "libvirt", reason="libvirt specific behaviour")

Skipping the test may also be an option but I guess that xfail would allow to keep the same value of PYTEST_REQPASS for all providers

@eighthave Any chance to rebase and fix this PR, so we can merge it? I think it would be ok to skip the test when the backend is not the ones for which it was designed.

I am closing this PR because it was not fixed in a long period of time. Feel free to reopen if still needed.