It's really INSTALL_NO_USER, not USER_LOGIN='false'. Also, we
need to make root_console work with no root password, sigh.
Signed-off-by: Adam Williamson <awilliam@redhat.com>
Sigh, 'overloaded' product templates like that don't quite work.
So, let's try doing it a different way, in main.pm.
Signed-off-by: Adam Williamson <awilliam@redhat.com>
Workstation live installs for F28+ drop the user creation and
root password panes from anaconda, so we need to not try and
use them any more. But we still want the old behaviour for F27.
I'm hoping this approach will work, if not, we'll find out soon
enough. This removes the install_no_user test for F28+ as it
will no longer differ from the install_default test.
Signed-off-by: Adam Williamson <awilliam@redhat.com>
Atomic was renamed AtomicHost, and Workstation Ostree was renamed
AtomicWorkstation, for F28+. As we still have F26 and F27 images
which will have the Atomic name, we duplicate those templates,
but there should be no more 'Workstation Ostree' images, so we
just rename those templates to the new name.
Signed-off-by: Adam Williamson <awilliam@redhat.com>
In the most recent Rawhide compose, anaconda has changed how it
names volume groups (again), and now the VGs in openQA installs
have rather longer names. This winds up causing a problem for
this needle: the column where the 'Device Type' dropdown is
placed actually gets narrower (because the column to the right,
which has the volume group name in it, is wider), and so the
dropdown box is narrower, and the arrow protrudes into the area
which the needle expects not to have an arrow in it. So, let's
make the match area slightly narrower.
Signed-off-by: Adam Williamson <awilliam@redhat.com>
This adds a check that the default package set selection is
actually correct, where possible and appropriate, as part of
the `_software_selection` test. We do this by examining the
`packaging.log` log file and checking which environment group
was selected.
Signed-off-by: Adam Williamson <awilliam@redhat.com>
This is currently broken, but openQA doesn't notice; we really
should. We could also check the default in other cases, but I
think that's less clear-cut, as it's kind of an anaconda design
choice, it's not mandated in Fedora requirements anywhere.
Signed-off-by: Adam Williamson <awilliam@redhat.com>
There's a problem with using `--releasever=rawhide` for upgrade
tests ATM - see #1531356 . To avoid this, we'll try using the
real Rawhide release number (which I'm adapting the scheduler
code to discover and pass in as `RAWREL`).
Signed-off-by: Adam Williamson <awilliam@redhat.com>
Rawhide has a bit of a problem where its 'description' of an
iSCSI disk is so long that the other columns that should appear
in the CONFIGURE MOUNT POINT dialog don't. This means our
device_sda_selected needle doesn't match, because the column
where 'sda' should appear isn't visible.
So add a soft-fail needle to cover this case; we know what the
description for the disk that's 'sda' in this case looks like,
so match on that.
Signed-off-by: Adam Williamson <awilliam@redhat.com>
Time for an annual spring clean. Based on the admin UI's list
of needles that haven't been matched for a long time, but with
some manual tweaking (some are actually still needed).
Signed-off-by: Adam Williamson <awilliam@redhat.com>
virtio graphics still seem to suffer from RHBZ#1403365, and also
os-autoinst believes they don't support snapshotting. So let's
try qxl for x86_64 UEFI. *That* may still suffer from #1403343,
but oh well, seems like we have no good choices here.
It looks like ppc64 also suffers from the Plymouth bug that's
affecting x86_64 UEFI + 'std' graphics, so let's use virtio
there - qxl apparently isn't available on ppc64 VMs, at least it
doesn't work in our deployment.
Signed-off-by: Adam Williamson <awilliam@redhat.com>
This somehow snuck in (in the wrong directory) with the ppc64
merges, but it's not needed. I've verified with the admin tab
for checking on needle use; this needle has never been matched,
and all the ppc64 text install tests match the non-variant
needle just fine. So I'm removing this unneeded variant.
The way GNOME Software displays an error after running into RHBZ
1314991 has changed, it seems, so we need a variant needle to
cope with that. Also, when an upgrade notification is visible,
the 'Restart & Update' button for doing a regular update is
shown in grey (not blue), so we need a variant needle to handle
that too.
While Rawhide suffers from #1518464 , virtio is a better choice
for UEFI tests, even if we get the problem with consoles using
the wrong color scheme again...
in templates-updates file
Only support "updates-server" flavor
do not support "updates-workstation" neither "updates-kde"
Signed-off-by: Michel Normand <normand@linux.vnet.ibm.com>
There's a bug causing the 'getting started' screen to crash.
This doesn't really make the system unusable, so treating it
as a soft failure seems appropriate, especially as this will
unblock all the post-install tests on Workstation.
It takes an unusally long time for Modular images to get from
language selection to the 'timbuktu screen', so give 'em a bit
more time. See bug report for more info.
Modular composes don't include these packages, but we need them
to run the web UI tests for FreeIPA and Cockpit. This is the
most reasonable hack I can come up with for now: just use a
non-modular fedora repo to source these packages when doing
Modular compose testing.
If we ever reach an all-Modular future, these packages should
be available in Modular composes I guess, but for now they are
not.
Current Modular composes don't include any graphical desktop
bits, so they're never going to want to do this. (This makes
sense for e.g. non-English install tests; we still want to
make sure Russian install works on Modular composes, we just
have to skip the post-install graphical input testing).
Since April there's been some kind of issue in the F26 base
image which means gnome-initial-setup doesn't run on the first
user login (as it should). The F25 base image is fine. I've
not yet had the time to look into this.
I put a workaround in place to prevent this problem causing
false fails of update tests that boot from the F26 base image,
but didn't apply the same workaround to upgrade tests, which
is why upgrade tests from F26 Workstation always fail - they
expect g-i-s to run on first login (which happens after the
upgrade, in upgrade tests) and it doesn't. So let's just extend
the workaround to apply to upgrade tests too, for now, until we
can figure out why this happens.
The default action on the reboot confirmation dialog changed
from Reboot to Cancel, so when we hit enter, we just cancel the
reboot. Tweak this to hit tab on F27+ (but not <F26, so update
tests continue to work too).