Don't check for unreliable grub2 needle #4652
Conversation
tests/caasp/first_boot.pm
Outdated
| @@ -20,7 +20,11 @@ use caasp 'process_reboot'; | |||
|
|
|||
| sub run { | |||
| # On VMX images bootloader_uefi eats grub2 needle | |||
| assert_screen 'grub2' unless is_caasp('VMX'); | |||
| if (is_caasp 'DVD') { | |||
| if (!check_screen('grub2')) { | |||
There was a problem hiding this comment.
As you probably know I recommend to always prevent the use of check_screen with non-zero timeout. Also, a record_soft_failure should normally be only used for product issue tracking. If you do not need this check and do not even abort the waiting time within the grub bootloader why not just delete the whole section and just wait in the next step for linux-login-casp? What Panos mentioned in https://progress.opensuse.org/issues/28648#note-6 is that he does not want to change the autoyast profile to disable the grub timeout. Maybe you want to rethink this? In the other tests we are quite happy with the decision to disable the grub timeout and always explicitly check for the grub screen :)
There was a problem hiding this comment.
So if I understood it right, the installation/autoyast.pm is going to check also for the linux-login-casp needle among others. That's great because this needle is the final stage which means the system has been installed and booted. Well, in case this happens, it means that first_boot has already been tested by the installation/autoyast.pm which makes itself a bit useless, but I am fine with it. I guess this is why you need to exclude the grub needle from the first_boot.pm, because it's too late for that. This makes sense, and it should work in case openQA gets stalled again.
PS: As for the record_soft_failure I would agree with Oliver. Please feel free to use record_info('poo#28648', "Stall prevented matching of BIOS needle");.
Verification run for maintenance: http://skyrim.qam.suse.de/tests/overview?build=4652-kravciak&version=2.0&distri=caasp&groupid=98
tests/autoyast/installation.pm
Outdated
| @@ -133,6 +133,8 @@ sub run { | |||
| = qw(bios-boot nonexisting-package reboot-after-installation linuxrc-install-fail scc-invalid-url warning-pop-up inst-betawarning autoyast-boot); | |||
| push @needles, 'autoyast-confirm' if get_var('AUTOYAST_CONFIRM'); | |||
| push @needles, 'autoyast-postpartscript' if get_var('USRSCR_DIALOG'); | |||
| # grub2 needle check is unreliable (stalls during timeout) - poo#28648 | |||
| push @needles, 'linux-login-casp' if is_caasp; | |||
There was a problem hiding this comment.
Hmm, I'm not sure why we need it here at all as we return after first stage and this tag is not explicitly used anywhere. So I guess this change should not be here. It's not related to the problem at least and we can create needle with reboot-after-installation.
There was a problem hiding this comment.
@rwx788 - we don't have reboot after installation 10 timeout on CaaSP, system just reboots. If there is stall then autoyast/installation.pm does not notice installation is finished (bios-boot needle disappears during stall)
There was a problem hiding this comment.
I still do not fully understand how it helps. Is it enough just to put extra needle into the array? I mean we don't use this tag anywhere later in the module. And regarding reboot-after-installation we have 3 tags which work as a condition to exit installer loop and return in case of caasp. Could you please explain how this particular change is related?
There was a problem hiding this comment.
I finally understood your point - thanks :) I added linux-login-casp to exit conditions, is it ok like this?
kravciak
force-pushed
the
grub2
branch
5 times, most recently
from
2681daa
to
8749f2f
Compare
tests/autoyast/installation.pm
Outdated
| @@ -149,6 +151,8 @@ sub run { | |||
| mouse_hide(1); | |||
| check_screen \@needles, $check_time; | |||
| until (match_has_tag('reboot-after-installation') || match_has_tag('bios-boot') || match_has_tag('autoyast-stage1-reboot-upcoming')) { | |||
| last if match_has_tag('linux-login-casp'); | |||
There was a problem hiding this comment.
Let's just put in the line above. Other than that let's think about more scalable solution, which we can introduce later.
|
Let's try it out. Maybe disabling grub timeout in the profile could resolve this kind of issues (not the first_boot). WDYT? |
|
We don't have static profiles, we use profiles generated on admin dashboard (parallel job): |
|
We still can modify it on server side then, so it's possible. And that's why I'm not saying we should do it now. But we with sle15 we don't have a single static profile, as need to inject reg info, etc. |
Fix for: https://openqa.suse.de/tests/1554587
Local run: http://dhcp165.suse.cz/tests/2095#step/first_boot/2