With the arrival of Pungi 4, the scheduler is no longer using
fedfind-provided BUILD and FLAVOR values, but ones derived from
Pungi properties. BUILD is now simply the Pungi compose_id.
FLAVOR is produced by joining the Pungi variant, type, and
format with '-' characters as the separators.
Pungi, unfortunately, does not treat 'Rawhide' as a release, it
synthesizes a release number for Rawhide composes and places
that in the compose ID. To cope with that, for now, the
scheduler will set RAWHIDE to '1' if the compose is a Rawhide
one. As we have to adapt all places where we parse the release
in any case, this commit consolidates them into a fedorabase
subroutine.
For the one place where we also used to parse the 'milestone'
from fedfind, there is a placeholder get_milestone subroutine
which currently returns an empty string, as I don't yet have a
good handle on how to draw the kinds of distinctions fedfind
mapped to 'milestone' from Pungi metadata.
Summary:
With the previous change to the server_(sata)_multi test, we
need to adjust the post-install test to use vdb not sdb, the
disks are virtIO now (not PATA as they were with 4.2)
Test Plan:
Check the server_multi test actually completes
properly now
Reviewers: garretraziel, jskladan
Reviewed By: jskladan
Subscribers: tflink
Differential Revision: https://phab.qadevel.cloud.fedoraproject.org/D732
Summary:
per T691, this has never actually tested SATA. With os-autoinst
4.2 it actually tests PATA, with os-autoinst 4.3 it breaks.
We can't find a way to make it do what we want with os-autoinst
4.3, so we're dropping the HDDMODEL bit entirely for now. We
will file a ticket upstream to see if this can be solved. We
keep the test itself because it's also the only test that hits
'guided_multi'; I'll post a matching diff for tools to change
the wiki reporting config to match.
Test Plan:
Schedule tests with os-autoinst 4.3 and see if this
one runs properly now.
Reviewers: jskladan, garretraziel
Reviewed By: garretraziel
Subscribers: tflink
Differential Revision: https://phab.qadevel.cloud.fedoraproject.org/D730
mostly cantarell changes. GTK+ changes are bugs, so put 'bug'
in the name to remind us of this. There's one non-typical
change, to main_hub_installation_source; I removed the match
on the text label of the spoke entirely instead of updating it,
in line with other similar needles, it's unnecessary and only
serves to make the needle more fragile.
Summary:
we have a KDE column for the 'base' tests, so we should run
them on the kde_live flavor.
Test Plan:
Schedule a full test run, ensure all three base
tests are scheduled and run correctly for KDE live.
Reviewers: jskladan, garretraziel
Reviewed By: garretraziel
Subscribers: tflink
Differential Revision: https://phab.qadevel.cloud.fedoraproject.org/D720
Summary:
D637 / ec6b3ff4 switched from using needle matches to using
validate_script_output when we want to run a console command
and check the result. validate_script_output is kinda over-
powered when all you want to do is check the command succeeded
(returned 0), though. testapi provides assert_script_run for
doing exactly that - it runs a script and fails if the script
fails (returns anything but 0). This gives us cleaner code and
is slightly more robust; validate_script_output uses the mini
web server on the worker, which I've occasionally seen crap
out, so it seems good to avoid using it when possible. assert_
script_run doesn't need it.
Test Plan: Check all (affected) tests still work properly.
Reviewers: jskladan, garretraziel
Reviewed By: jskladan, garretraziel
Subscribers: tflink
Differential Revision: https://phab.qadevel.cloud.fedoraproject.org/D714
Summary:
Not much to say, pretty much just implements the test case using
some commands I dug up that give us handy 0/1 exit statuses.
The assert_script_run function (from testapi) simply runs a
command/script and passes or fails based on the exit status;
we use a handy bash-ism when we *want* the exit status to be 1.
Test Plan: Run the test and check that it passes (properly).
Reviewers: jskladan, garretraziel
Reviewed By: garretraziel
Subscribers: tflink
Differential Revision: https://phab.qadevel.cloud.fedoraproject.org/D713
I messed up the last commit and mistakenly included this test,
which I've been working on and is not yet reviewed. The commit
was only supposed to add base_services_start. I'll send a new
diff for service_manipulation.
Summary:
pretty simple, just make sure no services failed to start. We
may run into the rngd issue here, not sure, let's land it and
see!
Test Plan:
I guess run the test and see what happens? I haven't
actually tested this myself yet, so, yeah.
Reviewers: garretraziel, jskladan
Reviewed By: garretraziel, jskladan
Subscribers: tflink
Differential Revision: https://phab.qadevel.cloud.fedoraproject.org/D710
Summary:
Along with the matching change to fedora-openqa-schedule to pass
these variables in when scheduling jobs, this avoids hardcoding
the release numbers for the upgrade tests (which means someone
has to remember to edit them every release). The new createhdds
similarly uses get_current_release() to decide what releases it
needs images for, so all this should hook up and work magically
without any human intervention required.
For clarity, the effect of the '_upgrade_' tests is "run an
upgrade from the 'current' Fedora release to whatever release
is being tested", and the effect of the '_upgrade_2_' tests is
"run an upgrade from the 'previous' Fedora release to whatever
release is being tested".
Test Plan:
Apply with D702, schedule upgrade tests, and make
sure the correct hard disk image filenames are used.
Reviewers: jskladan, garretraziel
Reviewed By: garretraziel
Subscribers: tflink
Differential Revision: https://phab.qadevel.cloud.fedoraproject.org/D703
This adds missing default_install tests to templates - we've missed
KDE live and Server DVD on UEFI for some reason and Workstation netinst was
missing. After this, we should have all non-optional tests from "Default
boot and install" section covered.
Differential Revision: https://phab.qadevel.cloud.fedoraproject.org/D693
Summary:
so here's our first attempt to use the 'carry on from a previous
test' stuff! This adds a base_selinux test that uses a disk
image from a previous default_install run, and adds jobtemplates
to run base_selinux for appropriate products: generic_boot
(for nightly tests), server_dvd, and workstation_live. Note that
you'll want to either update to the newest openQA package I just
built in COPR or create /var/lib/openqa/share/factory/tmp owned
by geekotest; openQA tries to use that directory as MOJO_TMPDIR
but in 4.2, if the directory doesn't exist, it doesn't create it,
and we wind up with the default MOJO_TMPDIR which is /tmp; when
the disk image is uploaded it creates a huge temp file in /tmp
and may well exhaust the available space as it's a tmpfs. I've
backported a recent upstream commit that tries to create the
directory if it doesn't exist, in 4.2-10.
It seems like openQA is smart enough to figure out the
dependencies correctly, so the 'base_selinux' test for each
product depends on the 'default_install' test for the same
product (not any of the other default_install runs) and will
use the hard disk image it produces.
Test Plan:
Do a full test run and make sure base_selinux tests
appear for appropriate products, depend on the correct default_
install test, the default_install test uploads the hard disk
image correctly, and the base_selinux test runs correctly. And
of course that nothing else broke in the process...
Reviewers: jskladan, garretraziel
Reviewed By: jskladan
Subscribers: tflink
Differential Revision: https://phab.qadevel.cloud.fedoraproject.org/D699
3.19.5 seems to change the colors of buttons slightly and also
the rendering of checkboxes/radio buttons, they're now placed
somewhat differently relative to their labels.
Summary:
this is following a couple of upstream changes I noticeed while
playing with dump_templates. Upstream has completely got rid of
the 'variables' keys from all dicts, and when you run dump_
templates the output does not contain any 'variables' keys, so
it seems reasonable to ditch these entirely (I think it was some
old thing that got subsumed into 'settings'). Also, when you
run dump_templates the 'prio' values come out in JobTemplates,
not in TestSuites; if you look at load_templates there's a bit
marked "we have to migrate the prio from the TestSuite to the
JobTemplate" which does exactly what it sounds like, so it
seems a good idea to move these instead of relying on that to
always be there.
While I had to re-do all the priorities anyway I tried to clean
them up a bit. The idea is roughly this:
10 - 19: most-critical sanity tests (i.e. default_install)
20 - 29: Alpha tests
30 - 39: Beta tests
40 - 49: Final tests
50+ : Optional tests
within each group I ordered 64-bit first, UEFI second, and
32-bit last (usually just as x0 / x1 / x2).
Test Plan:
Check that the file is valid and loads correctly,
and that everything works more or less the same except the
order of tests run is a bit different.
Reviewers: jskladan, garretraziel
Reviewed By: jskladan, garretraziel
Subscribers: tflink
Differential Revision: https://phab.qadevel.cloud.fedoraproject.org/D692
Summary:
Apparently it'll be something like anaconda.core.(PID). This
should result in that getting compressed and uploaded only if it
exists.
Test Plan:
Try and hit the Mystery Crasher and see if we get the
damn core file this time.
Reviewers: jskladan, garretraziel
Reviewed By: garretraziel
Subscribers: tflink
Differential Revision: https://phab.qadevel.cloud.fedoraproject.org/D688
Summary:
When anaconda manages to actually crash the python interpreter,
there should be a /tmp/anaconda.core containing the core dump
(per clumens, see calls to 'gcore' in isys.c). Let's upload it.
This might help us track down the mysterious occasional crashes
openQA seems to trigger (RHBZ #1289704)
Test Plan:
Try and trigger a python crash and see if the
file got uploaded. Of course, I did an entire freaking run on
staging and for *ONCE* not one test hit the mysterious crash,
thanks Murphy. I did at least check that this doesn't break
a 'normal' failure, if the file isn't there things don't
explode.
Reviewers: jskladan, garretraziel
Reviewed By: garretraziel
Subscribers: tflink
Differential Revision: https://phab.qadevel.cloud.fedoraproject.org/D686
garretraziel took these with F23, I think. Theoretically we
could just have these ones and ditch his, as we're not running
KDE install tests on F23, but we might want to discuss that kind
of question more generally, so for now let's just add the
variant.
We may need more variants for the post-install matches, but the
test doesn't actually get that far ATM because of dependency
issues in the KDE package set, so I'll deal with that once we
can actually do a successful KDE install.
Summary:
This is a bit icky, but it's the easiest way to solve a problem
I've seen a few times, the latest case being
https://openqa.stg.fedoraproject.org/tests/1664 . In that test,
_console_wait_login logs in to tty1 as user, then uefi_
postinstall wants to switch to tty3 and log in as root. When
it does that, sometimes the check_screen loop in console_login
gets hit before the display has actually switched from tty1 to
tty3, so everything gets out of sync.
An alternative would be to have root_console check that it's
either logged in or at the correct tty before handing off to
console_login, but that starts duplicating stuff, and it breaks
in the case the target tty is logged in as a user and the login
prompt is no longer visible...
Test Plan:
Check all tests run as normal, and maybe run UEFI
tests a few times to see that the bug no longer happens (but
it's hard to reliably trigger it anyway).
Reviewers: garretraziel, jskladan
Reviewed By: jskladan
Subscribers: tflink
Differential Revision: https://phab.qadevel.cloud.fedoraproject.org/D683
Summary:
As I figured out a while back, the seabios firmware font is
nearly the same as our console font (eurlatgr) but not quite,
significantly its @ character is different (shorter). We
already have a root_logged_in variant for the firmware font,
but we now need one for user_logged_in as well. The Atomic
installer image has had the Linux console fonts stripped from
its payload, so it's now using the firmware font, and user_
logged_in isn't matching - see
https://openqa.fedoraproject.org/tests/461 .
Test Plan:
Run the cloud_atomic_canned test (on an F23
post-release nightly, as it's not appearing for Rawhide
nightlies ATM) and confirm it now passes.
Reviewers: garretraziel, jskladan
Reviewed By: jskladan
Subscribers: tflink
Differential Revision: https://phab.qadevel.cloud.fedoraproject.org/D684
add KDE package set testcase - this only adds needles and
new testcase configuration, because with needles cleanup from D670,
it should work without change in code.
Differential Revision: https://phab.qadevel.cloud.fedoraproject.org/D673
Let's use the same principle with DESKTOP tag as we use with
LANGUAGE tag. Where there are any GNOME-only (or KDE-only) needles, tag
them with `DESKTOP-gnome` and delete then when different `DESKTOP` value
is specified. This will help with `QA:Testcase_Package_Sets_KDE_Package_Install`
testcase - we will be able to use almost the same code and check that KDE
got installed properly.
Differential Revision: https://phab.qadevel.cloud.fedoraproject.org/D670
damnit, that was still wrong. This should do the trick: we can
do 32-bit F23 upgrade tests now (virt-builder added the base
image we need), so re-add that, and put things in the best
order and fix the priorities.
Noticed some problems with the commit while testing it: we
weren't doing 32-bit minimal (to avoid having too many tests,
I think) so let's not do it for upgrade_2 for now, and also
the arch on one test was wrong.
Summary:
This bumps the existing upgrade tests to F23, and drops the
32-bit ones for now, as there is no 32-bit F23 base image for
virt-builder - RHBZ #1288733. It then adds new tests named
'upgrade_2_(etc)' and associates them with the F22 images. The
intent is that we should always have two sets of upgrade tests,
one for each of the currently-supported stable releases; when
we bump to testing F25, the 'upgrade' tests will be bumped to
F24 and the 'upgrade_2' tests to F23, and so on. There will be
a matching diff for openqa_fedora_tools.
Test Plan:
Execute a test run and make sure all the upgrade
tests run; of course you need to make sure you've built the
required disk images.
Reviewers: jskladan, garretraziel
Reviewed By: garretraziel
Subscribers: tflink
Differential Revision: https://phab.qadevel.cloud.fedoraproject.org/D679
Summary:
The 'reclaim_space_first_partition' needles aren't very good.
Using the area with the filesystem means we need to duplicate
the needle for NTFS, and the needles in fact also match the
*second* partition in most cases, sometimes causing the tests
to delete the second partition instead of the first and fail.
This design should work better: a single needle should match
all filesystems, and by including some of the highlighted top
row in the match, we should never match on the second row.
Including 'da' in the needle makes this really 'first disk,
first partition', but that's all the cases we currently use
it for. We don't really need this needle to work hard to make
sure we're on the right screen, as the subsequent needles
(which match buttons that only exist on this screen) do that
job fine.
We drop some needles that should no longer be necessary: F23
should now be using the cantarell17 version of the needle,
and F24 the freetype262 needle. We don't run installer tests
for any other release.
We could tweak second_partition needle to match, but nothing
is actually using it anyway.
Test Plan:
Do a full test and run and make sure the tests
which use this needle work, and always select the correct
partition.
Reviewers: jskladan, garretraziel
Reviewed By: garretraziel
Subscribers: tflink
Differential Revision: https://phab.qadevel.cloud.fedoraproject.org/D681
freetype 2.6.2 changes font rendering quite significantly.
Most text needles needed redoing. Committing without review
so maybe we finally get some damn useful test results...
Summary: seems like this got lighter for F24 too.
Test Plan:
Run the shrink tests and see if they get past this
screen.
Reviewers: jskladan, garretraziel
Reviewed By: garretraziel
Subscribers: tflink
Differential Revision: https://phab.qadevel.cloud.fedoraproject.org/D666
As we saw with F23 testing, qemu32 isn't really supported CPU. Also,
we cannot be sure that `std` is really supported graphics. This changes MACHINE
variables to use `host` CPU with 64bit machines (moreover, this is the case in
BOS now). It also deletes 32bit machine and schedules 32bit tests on 64bit
instead. It also changes graphics to `qxl`. Even though we aren't using SPICE, qxl has
better support (and higher priority) and it seems to work OK with VNC. Fixes T637.
Differential Revision: https://phab.qadevel.cloud.fedoraproject.org/D655
Summary:
this needle seems to be fragile not because of cantarell in
fact, but because the difference between the the left edge of
the language box and the 'F' is not consistent. So, let's just
not include it in the match.
Test Plan:
Check if French install test works (see test #8847
failure in BOS).
Reviewers: jskladan, garretraziel
Reviewed By: garretraziel
Subscribers: tflink
Differential Revision: https://phab.qadevel.cloud.fedoraproject.org/D661
This adds another Cantarrell 1.8 missing needle as well as
lighter variant of Done button (it's perhaps caused by displayed
warning about missing swap partition on no_swap test?). With this
in place, all tests on Rawhide passes except of KDE (but it seems
that KDE Live images are borked) and desktop upgrade tests.
Differential Revision: https://phab.qadevel.cloud.fedoraproject.org/D651
Summary:
Instead of sitting there waiting 6000 seconds twice, when DNF
explicitly tells us it failed, just die.
This is why we haven't been getting proper compose check reports
lately; the upgrade tests are failing, waiting 6000 seconds to
time out, then being cloned and tried again, waiting another 6000
seconds. This is just barely going beyond check-compose's 8 hour
wait limit, as it's some time before the upgrade tests even get
started (they're low in the priority list). We're still going to
have that problem if the tests fail any other way, but this at
least catches that case.
Test Plan:
Run the upgrade tests and see that they fail quicker
(assuming the dependency problems they're dying on aren't fixed).
Maybe also do a 22-23 upgrade test and check it still succeeds
properly.
Reviewers: jskladan, garretraziel
Reviewed By: garretraziel
Subscribers: tflink
Differential Revision: https://phab.qadevel.cloud.fedoraproject.org/D650
Summary:
When Rawhide's using a debug kernel, the install is pretty slow,
and often times out. This gives us a 33% longer timeout for the
install process when running on Rawhide. This will slow Rawhide
runs down a bit when there are genuine failures at this point,
but it seems kind of unavoidable.
Test Plan: Do a Rawhide run and see if we get fewer false fails.
Reviewers: jskladan, garretraziel
Reviewed By: garretraziel
Subscribers: tflink
Differential Revision: https://phab.qadevel.cloud.fedoraproject.org/D643
