the first param to assert_and_click is *the button number*, not
the wait time. So as soon as I set that for anything it was not
gonna work. So going back to the original approach but fixing
that, hope this works.
Summary:
I really just want to add the desktop_terminal test, but I think
this refactor is in order now. It splits up loading of the
various test phases (much as SUSE do it) and allows us to run
the post-install tests without the install tests, for e.g. I
tweaked things to allow the upgrade tests to use the existing
_wait_login tests for final login and combine the two upgrade
postinstall tests into one simple one.
This comes with a bit of a behaviour change to make graphical
wait login behave the same as console wait login: it will log
in unless USER_LOGIN is set to 'false'. Previously it only
logged in if both USER_LOGIN and USER_PASSWORD were set, which
I don't think ever happened in a graphical test, so we never
actually did a graphical login. The intent here is we should do
a login on the default_install tests. That's going a bit beyond
the test case, but it seems like a reasonable thing to test. We
can set USER_LOGIN to false if we don't want to do it.
Test Plan:
Do a full test run, make sure the new tests work and
no old tests break.
Reviewers: jskladan, garretraziel
Reviewed By: garretraziel
Subscribers: tflink
Differential Revision: https://phab.qadevel.cloud.fedoraproject.org/D839
Summary:
These require openQA tap networking to allow the server and
client boxes to communicate, and require masquerading (NAT) so
the server at least can reach a repository (dnf/rolekit really,
really do not want to work without a repo connection).
They use the 'parallel' test support to have the server deploy
run first while the client enrol test waits at the grub menu
until the server is done before it goes ahead.
This is all deployed and working on stg. The really tricky bit
was getting all the openvswitch and firewall config right in
ansible.
We *could* do the server deploy test as a follow-on from the
default install test to save the install, but then we'd have to
teach it to change the hostname and set up static networking
post-install. I'm not sure if it's worth doing that.
This requires the corresponding openqa_fedora_tools commit that
adds the hard disks (containing the kickstarts - it's possible
to get them from remote during install, but we have to set up
name resolution or hard code the IP of the server).
Test Plan:
Deploy this and the openqa_fedora_tools commit,
generate the disks, configure the networking (good luck! See
the docs in openqa_fedora_tools) and see if you can run the
tests. If you're using Docker, uh...sorry. You somehow need to
set things up so the workers can use tap interfaces that can
talk to each other and are NATed to the outside world. Have fun.
I can talk you through it on IRC...
Reviewers: jskladan, garretraziel
Reviewed By: garretraziel
Subscribers: tflink
Differential Revision: https://phab.qadevel.cloud.fedoraproject.org/D831
Summary:
you can't use validate_script_output like this, it doesn't just
return a bool, it actually kills the test if the output doesn't
validate. So any time we didn't have anything in /var/tmp/abrt,
the post_fail_hook would die at that point and not upload the
contents of /var/log as we wanted it to.
Note that even script_output will kill the test if the return
code is >0. That's OK in this case, but always be careful with
these subroutines.
Test Plan:
Run a test that fails in post-install but doesn't
produce anything in /var/tmp/abrt (you may have to artificially
break something). Check that it now uploads /var/log. I noticed
this while working on FreeIPA tests...
Reviewers: jskladan, garretraziel
Reviewed By: garretraziel
Subscribers: tflink
Differential Revision: https://phab.qadevel.cloud.fedoraproject.org/D830
Summary:
There's a weird problem that we hit occasionally only on the
Server image, see e.g.:
https://openqa.fedoraproject.org/tests/14300
I *think* what's happening (thanks hergertme from #gtk+ IRC for
help with this) is that when the 'install done!' text and the
reboot button appears, there's a short transition for the button
so it doesn't look *exactly* like the screenshot right away. If
openQA's needle match loop happens to kick in and check the
match right after the text and button appear, while the
transition is happening, it decides that some random area of the
gradient down the left hand side of the screen matches *better*
than the button, so it 'clicks' that, and we don't get a reboot.
If the openQA needle match loop happens to kick in a millisecond
or whatever later, after the button is done transitioning, the
button is now a 100% match and the bug doesn't happen. This only
happens on Server because the left-hand side gradient differs in
each flavor, and only the Server one just happens to have a spot
which is nearly identical to the needle area.
So to avoid that, let's include a bit of the grey space next to
the button. This is a bit more fragile against GTK+ changes to
button design or padding or whatever, but it will at least avoid
the match issue.
The other option would be to try and hit the button with tab and
enter or something, I guess.
Test Plan:
bit hard since this is an intermittent bug, I guess
run install_default on a Server image a bunch of times and see if
it always works.
Reviewers: jskladan, garretraziel
Reviewed By: garretraziel
Subscribers: tflink
Differential Revision: https://phab.qadevel.cloud.fedoraproject.org/D824
according to rdieter the runner dialog could be white or grey
background, we should match on either. One is probably a bug
but it's best to just work with it. Removing the February
version without the 'search' text.
Summary:
Rawhide currently seems to have a bug in spell check dictionary
load, which causes the test to fail as it requires another Done
click. So add a workaround needle that handles this case.
Test Plan:
Apply the patch, run some tests, see if they work. I
did a test run on staging:
https://openqa.stg.fedoraproject.org/tests/13331
Reviewers: garretraziel, jskladan
Reviewed By: jskladan
Subscribers: tflink
Differential Revision: https://phab.qadevel.cloud.fedoraproject.org/D815
Summary:
BOOT_UPDATES_IMG_URL is a pretty misleading name - it used to
be the actual URL, but now it's simply a boolean that decides
whether we look for the effect of the openQA updates image or
not. TEST_UPDATES seems clearer.
GRUBADD does the same thing as GRUB, on top of it. The point of
this is so we can add an option to the scheduler CLI that lets
you say 'run the normal tests, but with this updates image' -
so we can easily (albeit manually triggered) check the impact
of some anaconda change that needs testing. It should never be
set in the templates or the tests, it's there strictly for the
scheduler (whether that's fedora_openqa_schedule or literally a
person calling `client isos post`) to use as a kind of override.
The tests that test updates image loading will probably fail
when doing this, but all other tests should work as intended,
including ones that specify GRUB, becase the extra params will
just get added on top. That's why I invented a new var instead
of just letting the scheduler override GRUB's value when POST
ing.
Test Plan:
Check the rename didn't break anything (updates tests
still work). Run tests with GRUBADD param, make sure value is
correctly appended to cmdline both when GRUB is also specified
and when it is not.
Reviewers: jskladan, garretraziel
Reviewed By: garretraziel
Subscribers: tflink
Differential Revision: https://phab.qadevel.cloud.fedoraproject.org/D801
Summary:
per details in T759, the 'unipony' updates image we use to test
the updates image features doesn't work with latest anaconda (f24
and Rawhide). I've built a new updates image which uses a neat
anaconda feature that allows you to override CSS with a file in
a special location; it sets the background for disk capacity
texts on the INSTALLATION DESTINATION spoke to be pink. This
lets us use a simple needle that just looks for a pink blob on
that spoke, on the basis that it's unlikely there'll ever be a
pink blob there for any other reason, so if there is one, the
updates image worked. There will be an accompanying tools diff
to change the updates disk image to use the new updates image.
Test Plan:
Do a test run and check the updates image tests pass
and no other tests are broken. You'll need to pull in the tools
diff and re-generate the updates disk image to check that test,
the scsi_updates_img test should work with just this diff.
Reviewers: jskladan, garretraziel
Reviewed By: garretraziel
Subscribers: tflink
Differential Revision: https://phab.qadevel.cloud.fedoraproject.org/D799
Some needle changes to account for latest Rawhide. There seems
to be a bug between anaconda and GTK+ 3.20 causing text in the
disk selection widget not to be styled correctly: this commit
adds needles for that which are tagged as 'workaround', meaning
openQA will complete the tests but mark them as requiring a
workaround to work ('soft fail'). The bug is #1322036. There is
also an intentional change to the volume list in custom part;
the mount point names are now in dark grey rather than black (in
fact this has always been intended, but it's been broken for a
long time, according to davidshea). I cleaned up the part select
needles at the same time; the new ones have the non-variant
names, as davidshea says the same change is coming to F24 and
the original needles haven't matched for months. The F23 Atomic
install test doesn't hit any of these needles, so we don't need
to keep the original needles around for that. The most recent
variants are kept, as they'll be needed for F24 tests until the
new anaconda build goes stable for F24. The 'freetype262'
variant of part_select_swap hasn't matched for two months and is
dropped.
Summary:
this should avoid unnecessary disk uploads and hopefully help
further reduce the incidence of weird failures in the chained
tests.
With this change we should only upload disk images for the cases
where we're actually going to run the chained tests: we won't
upload disk images for default_install runs on images we don't
run the chained tests for, or for the UEFI job for images we
*do* run the chained tests for.
We only actually need to run the current chained tests
for Server DVD, Workstation live and KDE live x86_64; there's
no need to run them for Everything boot, so we drop that.
Test Plan:
Do a full test run and make sure all tests run
properly and we now only upload disk images where we really need
to.
Reviewers: jskladan, garretraziel
Reviewed By: garretraziel
Subscribers: tflink
Differential Revision: https://phab.qadevel.cloud.fedoraproject.org/D794
Summary:
these together test QA:Testcase_kickstart_firewall from the
Server matrix. I'll have to come up with some kinda way to
handle reporting that, might be tricky.
Couple of tweaks to overall test flow: tests can now specify
a POSTINSTALL variable which will load a post-install test
following a naming convention, and tests can specify USER_LOGIN
as 'false' to disable the 'log in as a user' step entirely. We
could easily adjust the kickstarts to create a user so the test
could log in as one, but it seems like an unnecessary step and
I liked the idea of allowing the user login to be skipped.
Test Plan:
Schedule 'universal' tests, check the new tests run
and pass or fail as they should, check no other test is broken
by the logic flow changes.
Reviewers: jskladan, garretraziel
Reviewed By: garretraziel
Subscribers: tflink
Differential Revision: https://phab.qadevel.cloud.fedoraproject.org/D792
Summary:
We named a bunch of the tests 'server_foo' back when we were
starting out and didn't really know what we were doing. They're
all really installation tests, not 'server' tests. I actually
want to start adding some Server tests now, so it seems like a
good time to fix that mess. This standardizes on 'install_' as
a prefix for installation tests, converting all the 'server_'
tests and fixing up a couple of odd cases to use 'install_'.
Test Plan:
Apply along with the matching commit for tools, do
a full test run and report test, and make sure all tests run
with the new names and correct ResTups are generated for wiki
submission.
Reviewers: jskladan, garretraziel
Reviewed By: garretraziel
Subscribers: tflink
Differential Revision: https://phab.qadevel.cloud.fedoraproject.org/D790
Summary:
I believe the failures in the Server DVD chained Base tests are
happening because the VM is not cleanly shut down before the disk
image is uploaded. This adds a shutdown step to all tests that
upload a disk image (so, for now, just default_install). To keep
things simple it just runs 'shutdown' from a root console, rather
than using graphical desktop shutdown methods, as the aim is only
to make the disk state clean, not to test shutdown exactly.
I've tested this on staging; a Server DVD test run with this
change produced a full set of passed tests, as opposed to all
the Base tests failing because the system didn't boot properly.
Workstation and KDE tests seem to work fine also.
For the record, SUSE does much the same thing as this commit.
Test Plan:
Do a full test run and make sure everything that worked
before still does. Check that all default_install tests have a
_console_shutdown step added, and it works, and all chained tests
work (or fail for some unrelated reason, but make sure this
doesn't break them).
Reviewers: jskladan, garretraziel
Reviewed By: garretraziel
Subscribers: tflink
Differential Revision: https://phab.qadevel.cloud.fedoraproject.org/D787
Summary:
First off, this revises the anaconda crash handling needles a
bit. We ditch gtk3195 and update anaconda_error to reflect
current F24/Rawhide. We keep the old anaconda_error around for
now as anaconda_error-23, to handle crashes in the F23 two-week
Atomic nightlies. We also add an 'early' variant, which is for
when (I think) the installer crashes very early, before it's
loaded in GTK+ settings; when that happens, the dialog uses a
different font. The screenshot comes from a recent Rawhide test
that crashed.
We also restore the anaconda `post_fail_hook` code to click
the Report button when a crash happens. This was erroneously
removed in D637. Before the Report button is clicked, the
`anaconda-tb` file exists but the libreport stuff in `/var/tmp`
does not. By removing this, we lost the libreport bits from
the uploaded files, which makes it harder to report crashes. So
let's add it back.
Finally we fix the actual tarring and uploading of `/var/tmp`;
also in D637 this got broken because it was being tarred up in
whatever directory the commands happened to be running in, but
we were still trying to upload it from `/var/tmp`.
https://openqa.stg.fedoraproject.org/tests/8444 was run with
these changes, and has `/var/tmp` correctly uploaded.
Test Plan:
Run some test that crashes, make sure the crash
handling all works correctly.
Reviewers: jskladan, garretraziel
Reviewed By: garretraziel
Subscribers: tflink
Differential Revision: https://phab.qadevel.cloud.fedoraproject.org/D768
GTK+ 3.19.10(?) has changed things in current F24/Rawhide;
weirdly, done-freetype262 seems to match sometimes but not
always. Add a new needle from a current screenshot but keep
freetype262 around for now just in case. Drop some variants
that should never be needed any more, rename the original
needle to -23 to mark that we're only keeping it around for
the F23 two-week Atomic tests. For now keep french-cantarell20
as the 'official' French needle, we'll see if it hits any
failures when the tests are working a bit better.
with Pungi 4, the public repos are product-y, we need to add
/Everything/ to the path between the release and the arch.
Again pushing without review to get the tests working.
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.
