action #45020
closed[functional][y] tests should detect that the wrong desktop will show up
0%
Description
Motivation¶
openQA test in scenario sle-15-SP1-Installer-DVD-aarch64-create_hdd_gnome@aarch64 fails in
first_boot expecting a desktop but only the text login shows up.
installation_overview already showed that the target will be "Text mode" instead of "Graphical mode" where the test could already fail. But even earlier in https://openqa.suse.de/tests/2317867#step/system_role/1 only the "Minimal" role is shown and preselected even though it should be "SLES with GNOME".
Acceptance criteria¶
- AC1: The wrong desktop to be expected is already detected before installation of packages starts
Suggestions¶
Check within the installation summary for the expected boot target. Out of scope: Explicit system role checks. That is covered in #37504
- I can think of three ways how to handle this:
- Extend existing needles in the installation summary to cover the systemd target
- Add an explicit needle check for the systemd target
- Check the logs from the installation system for the right systemd target before rebooting
Further details¶
Always latest result in this scenario: latest
Updated by okurz over 5 years ago
- Due date set to 2019-02-12
pre-fill last sprint in M22 with all tickets within milestone not yet assigned to sprints
Updated by riafarov about 5 years ago
- Due date changed from 2019-02-12 to 2019-03-12
Updated by okurz about 5 years ago
- Target version changed from Milestone 22 to Milestone 23
Updated by okurz about 5 years ago
Taking a look in the ticket again: I think it is still valid however the scenario does not fail in the same way anymore. Please keep in mind that the category is "Enhancement to existing tests". The idea was to improve the failing test to make the problem more obvious, e.g. fail early with an explicit message
Updated by JRivrain about 5 years ago
I think, the fact that only "minimal install" was available at that time was a rare product bug, or that the iso image that was used had been wrongly labelled ( as if we had inserted the wrong dvd ) so therefore the default choice was not the expected. So I think we could just close this ticket with this.
However, we could improve the way it currently works
In the case of these failures, the variable SYSTEM_ROLE was set to default, the test does not try to select anything if the role is set to default.
assert_screen('system-role-default-system', 180);
my $system_role = get_var('SYSTEM_ROLE', 'default');
change_system_role($system_role) if (get_var('SYSTEM_ROLE') && !check_var('SYSTEM_ROLE', 'default'));
In latest runs, the system_role is not set, so that type of error would not get detected either, we'd just select the default entry and go ahead.
So for the first solution that is mentioned in description, the system_role should be set explicitly to "gnome", not default or null. And we should create two needles tagged with "system-role-gnome(-selected)".
I think this solution is simpler than the second one, which would involve to re-factor installation_overview to detect many different cases and create as many needles.
Updated by okurz about 5 years ago
- Assignee changed from JRivrain to okurz
As discussed in the refinement meeting I will crosscheck if we do not already have another ticket for explicit system role checking
Updated by okurz about 5 years ago
- Related to action #37504: Assert expected default roles for different scenarios added
Updated by okurz about 5 years ago
- Description updated (diff)
- Assignee deleted (
okurz)
Updated by JRivrain about 5 years ago
Just a note : here is what we get in y2log while going back and forth in the "select role" module and selecting the different options. This could be used for defining the system roles in advance in Openqa.
# grep "Applying system role" y2log
2019-03-05 08:54:44 <1> install(3345) [Ruby] installation/select_system_role.rb:209 Applying system role 'server'
2019-03-05 09:34:04 <1> install(3345) [Ruby] installation/select_system_role.rb:209 Applying system role 'generic_desktop'
2019-03-05 09:38:22 <1> install(3345) [Ruby] installation/select_system_role.rb:209 Applying system role 'gnome'
2019-03-05 09:45:51 <1> install(3345) [Ruby] installation/select_system_role.rb:209 Applying system role 'kde'
2019-03-05 09:46:49 <1> install(3345) [Ruby] installation/select_system_role.rb:209 Applying system role 'serverro'
Updated by JRivrain about 5 years ago
Updated by riafarov about 5 years ago
- Due date changed from 2019-03-12 to 2019-03-26
Missing VR on production.
Updated by JRivrain about 5 years ago
VRs done on production, all good.
https://openqa.suse.de/tests/2537188
https://openqa.suse.de/tests/2537189
https://openqa.suse.de/tests/2537187
Updated by JERiveraMoya about 5 years ago
PR (2nd try): Detect systemd target in installation_overview
Updated by JERiveraMoya about 5 years ago
- Status changed from Feedback to In Progress
Let's check the module in O3 and OSD until the end of the sprint, that everything was fine, no more breaks hopefully :)
Updated by JRivrain about 5 years ago
Test failing: https://openqa.suse.de/tests/2706786
DESKTOP is set to gnome, but then at some point it gets magically changed to textmode.
[2019-03-22T13:10:36.213 CET] [debug] BUG: bsc#1058071 - No VNC server available in SUT, disabling X11 tests. Re-enable after bug is fixed
[2019-03-22T13:10:36.213 CET] [debug] usingenv DESKTOP=textmode
Comes from https://github.com/os-autoinst/os-autoinst-distri-opensuse/blob/master/products/sle/main.pm#L142
Updated by JERiveraMoya about 5 years ago
PR: Use record_info instead of die -> Merged
VR: sle-15-SP1-allmodules+allpatterns (re-triggered one job failing previously)
It detect the problem but the info box is empty. You might want to fix record_info
parameters, first param is a short message (a few characters displayed in the box), and the other one where you put your message, probably is in the logs, but as we need to take a a look to many jobs, I guess it worth to fix it.
We might not see all those info boxes if some build has not recently triggered, therefore searching for them could be part of next ticket.
Updated by JERiveraMoya about 5 years ago
- Copied to action #49622: [functional][y] Verify the wrong desktop will show up added
Updated by JERiveraMoya about 5 years ago
- Copied to deleted (action #49622: [functional][y] Verify the wrong desktop will show up)
Updated by JERiveraMoya about 5 years ago
- Related to action #49622: [functional][y] Verify the wrong desktop will show up added
Updated by mloviska about 5 years ago
There is a problem with remote installations.
We are switching to install the console on the controller, which is a SUT with already installed OS.
I guess, we should omit this check, or change console using chvt [NUM] on remote installations.
- remote_ssh_controller@64bit
REMOTE_CONTROLLER={ssh,vnc}
Updated by JERiveraMoya about 5 years ago
Thanks @mloviska, I will add that info to the follow-up ticket. Ok, I got it now, it is a break, so we need to fix it now.
Updated by JERiveraMoya about 5 years ago
PR: Fix record_info in installation_overview -> Merged
Updated by JRivrain about 5 years ago
- Status changed from In Progress to Feedback
It seems fine now:
https://openqa.suse.de/tests/2730775
I propose a follow-up ticket to improve the way we switch between ttys, by using chvt [NUM] as mloviska suggested.
Updated by JERiveraMoya about 5 years ago
- Status changed from Feedback to Resolved
It's already added to #49622. Let's resolve the detection then.
Updated by JRivrain about 5 years ago
- Related to action #49703: [functional][y][timebox:8h] Evaluate how to switch between TTYs in specific scenarios (hyperv, ssh,vnc) added