Enable testing with ubuntu-bionic (#69)

* Enable testing with ubuntu-bionic

* tools/test-setup.sh: Fix libvirt access

Most distributions relies on unix rights to configure
libvirt socket access, so the user has to be added to the 'libvirt'
group to use it. The challenge is that it implies login/logout.

The workaround to use newgrp is failing since it's spawning a new
shell, so the group change is not applied to current shell.

As a solution, use policykit to handle the access rights to the
unix socket. It's the same thing as for unix rights, except that
policykit doesn't rely on the current shell, so no need to login/logout
to get the group change applied.

Signed-off-by: Arnaud Patard <[email protected]>

Co-authored-by: Arnaud Patard <[email protected]>

Author: ssbarnea

README.rst

@@ -37,6 +37,7 @@ multiple platforms but some are disabled due to known bugs.
 
 * ✅ MacOS with VirtualBox - GitHub Actions
 * ✅ Fedora 32 with libvirt - Zuul
+* ✅ Ubuntu Bionic (18.04) with libvirt - Zuul
 * ❌ CentOS 8 with libvirt - Zuul DISABLED due to 1127_ and 11020_
 
 Please **do not file bugs for unsupported platforms**. You are welcomed to

tools/test-setup.sh

@@ -22,7 +22,20 @@ sudo $PYTHON -m pip install -U tox
 
 # === LIBVIRT SETUP ===
 sudo systemctl enable --now libvirtd
+sudo sed \
+    -e 's!^[# ]*unix_sock_rw_perms = .*$!unix_sock_rw_perms = "0777"!g' \
+    -e 's!^[# ]*auth_unix_rw = .*$!auth_unix_rw = "polkit"!g' -i /etc/libvirt/libvirtd.conf
+# on fedora, looks like virbr0 iface stays there after a restart of libvirt but the network is not
+# started back, leading to network failure when the script tries to start the network
+if sudo virsh net-list --name | grep -qw default;  then
+    sudo virsh net-destroy default
+fi
+sudo systemctl restart libvirtd
 sudo usermod --append --groups libvirt "$(whoami)"
+# on some dists, it's auto-started, on some others, it's not
+if virsh -c qemu:///system net-list --name --inactive | grep -qw default;  then
+    virsh -c qemu:///system net-start default
+fi
 
 # only info about the virtualisation is wanted, so no error please.
 sudo virt-host-validate qemu || true
@@ -128,8 +141,8 @@ vagrant plugin list | tee >(grep -q "No plugins installed." && {
 cd $DIR
 
 # sudo su: dirty hack to make sure that usermod change has been taken into account
-sudo su -l "$(whoami)" -c "cd $(pwd) && timeout 300 vagrant up --no-tty --debug"
-sudo su -l "$(whoami)" -c "cd $(pwd) && timeout 300 vagrant halt"
-sudo su -l "$(whoami)" -c "cd $(pwd) && timeout 300 vagrant package --output testbox.box"
-sudo su -l "$(whoami)" -c "cd $(pwd) && vagrant box add testbox.box --name testbox"
-sudo su -l "$(whoami)" -c "cd $(pwd) && vagrant destroy -f"
+timeout 300 vagrant up --no-tty --debug
+timeout 300 vagrant halt
+timeout 300 vagrant package --output testbox.box
+vagrant box add testbox.box --name testbox
+vagrant destroy -f

zuul.d/layout.yaml

@@ -33,6 +33,16 @@
 #       tox_envlist: devel
 #     timeout: 3600
 
+- job:
+    name: molecule-vagrant-ubuntu-bionic
+    description: Run py36 tox environment
+    parent: ansible-tox-py36
+    nodeset: ubuntu-bionic-1vcpu
+    attempts: 2
+    vars:
+      tox_envlist: py36
+    timeout: 3600
+
 - project:
     check:
       jobs: &defaults
@@ -43,5 +53,6 @@
         # - molecule-vagrant-centos
         - molecule-vagrant-fedora
         # - molecule-vagrant-devel-centos
+        - molecule-vagrant-ubuntu-bionic
     gate:
       jobs: *defaults