Adapt password typing for slower arch #20097

mloviska · 2024-08-30T14:00:00Z

In https://openqa.opensuse.org/tests/4444816#step/firstrun/15 the password typing results in incomplete typing string which yields a cracklib error that the string is too short.

ticket: https://progress.opensuse.org/issues/165800

Verification runs

opensuse-Tumbleweed-JeOS-for-RPi-aarch64-Build20240829-jeos@RPi3 -> https://openqa.opensuse.org/tests/4445634
opensuse-Tumbleweed-JeOS-for-RPi-aarch64-Build20240829-jeos@RPi4 -> https://openqa.opensuse.org/tests/4445635
opensuse-Tumbleweed-JeOS-for-RPi-aarch64-Build20240829-jeos@RPi4 -> https://openqa.opensuse.org/tests/4445636
opensuse-Tumbleweed-JeOS-for-RPi-aarch64-Build20240829-jeos@RPi3 -> https://openqa.opensuse.org/tests/4445637

mloviska · 2024-08-30T14:28:40Z

@Vogtinator, the password for the user fails as it is too short, but the same one works just fine for root. Is this worth to file a bug that root screen accepts even short passwords?

Vogtinator · 2024-08-30T18:14:15Z

@Vogtinator, the password for the user fails as it is too short, but the same one works just fine for root. Is this worth to file a bug that root screen accepts even short passwords?

Currently it's intentional behaviour, similar to YaST, which just shows a warning that can be ignored. When the root pw dialog in jeos-firstboot gets rewritten some time in the future it'll probably do the same.

Vogtinator · 2024-08-30T18:16:09Z

In https://openqa.opensuse.org/tests/4444816#step/firstrun/15 the password typing results in incomplete typing string which yields a cracklib error that the string is too short.

Hm, that particular failure is because of the username. I also wonder how ssh can drop input, that shouldn't happen...

mloviska · 2024-09-02T08:45:47Z

Hm, that particular failure is because of the username. I also wonder how ssh can drop input, that shouldn't happen...

The username needle was matched, I see this error after the community jeos password was set. Which contains only 5 characters

Vogtinator · 2024-09-02T08:50:31Z

The username needle was matched,

Ok, there's just no screenshot for the fully typed in dialog or the error message.

I see this error after the community jeos password was set. Which contains only 5 characters

IMO it shouldn't ever set "linux" as password, only use it for login if necessary.

Vogtinator · 2024-09-02T09:38:18Z

tests/jeos/prepare_firstboot.pm

@@ -45,7 +45,7 @@ sub run {
            # Handle default credentials for ssh login
            # On SLE we use an image preprocessed by openQA where the default
            # $testapi::password was set
-            $testapi::password = $default_password;
+            local $testapi::password = $default_password;
        }
        # 'root-ssh' console will wait for SUT to be reachable from ssh
        select_console('root-ssh');


I guess it should just reset $testapi::password after the select_console?

It should have https://github.com/os-autoinst/os-autoinst-distri-opensuse/blob/master/tests/jeos/prepare_firstboot.pm#L78

alright, even the original $testapi::password is changed outside this module.
@ggardet, is it possible that the password somewhere in the testapi code is changed in RPi tests?

Default password is linux on JeOS images. I guess we can change it in firstboot config interface.

that password fails the cracklib-check and it seems that the osado's default password in testapi also defaults to linux

I have done a PoC, https://openqa.opensuse.org/tests/4457380#step/firstrun/14 shows that $testapi::password is set to linux. In the PoC run I have changed it, the UI test object passed. Later the password was restored to linux (default) hence the test fails in another point.
If we keep linux as default, I would rather skip this test object by setting WIZARD_SKIP_USER in community jeos RPi test runs

pevik · 2024-09-04T10:48:37Z

@mloviska are verifications just outdated or problem not fixed yet? I tried your fork to verify os-autoinst/opensuse-jobgroups#516, but it failed as well.

Vogtinator · 2024-09-23T14:25:07Z

I had a look. What currently happens is this:

The TW job group hardcodes PASSWORD=linux for all RPi tests. This is set as initial $testapi::password by main_common.pm. When consoles are added by init_consoles, $testapi::password is passed to the various ssh instances and stored there. In prepare_firstboot, the initial $testapi::password is still linux due to the test variable, but it also does $testapi::password = "linux" early on.

My initial attempt was to remove the PASSWORD=linux setting (os-autoinst/opensuse-jobgroups#520) and rely on prepare_firstboot to set the image password, but that doesn't work because the consoles already have the password stored at that point. This could be rectified by doing init_consoles again or overwriting the password parameters, but this results in the messy situation that there can be intermediate states, especially if there is no reboot with console reconnection.

Not sure how this can be addressed the best way. IMO the PASSWORD=linux variable makes sense as it's image specific, but belongs into the medium configuration instead of the job group. Maybe prepare_firstboot should drop $default_password and rely on the variable instead, and set $testapi::password to a stronger one just before jeos-firstboot is started and update the consoles?

mloviska · 2024-09-30T08:47:31Z

Not sure how this can be addressed the best way. IMO the PASSWORD=linux variable makes sense as it's image specific, but belongs into the medium configuration instead of the job group. Maybe prepare_firstboot should drop $default_passwordand rely on the variable instead, and set$testapi::password` to a stronger one just before jeos-firstboot is started and update the consoles?

Or we can simply skip this dialog as it is tested everywhere. Anyway, I am fine with both ways but not sure who should make the call when it comes to community images.

Vogtinator · 2024-09-30T08:50:47Z

Not sure how this can be addressed the best way. IMO the PASSWORD=linux variable makes sense as it's image specific, but belongs into the medium configuration instead of the job group. Maybe prepare_firstboot should drop $default_passwordand rely on the variable instead, and set$testapi::password` to a stronger one just before jeos-firstboot is started and update the consoles?

Or we can simply skip this dialog as it is tested everywhere. Anyway, I am fine with both ways but not sure who should make the call when it comes to community images.

That's also an option, but means a separate code path for creation of the $testapi::username user again. It's the easiest option for now though.

mloviska · 2024-09-30T09:16:29Z

That's also an option, but means a separate code path for creation of the $testapi::username user again. It's the easiest option for now though.

The code path is already present. We just need to set WIZARD_SKIP_USER=1 in the medium

Vogtinator · 2024-09-30T09:18:57Z

There's also #20212 with a similar issue FTR

mloviska · 2024-09-30T09:38:25Z

There's also #20212 with a similar issue FTR

I would rather blame yaml schedule over here :D

mloviska · 2024-09-30T10:00:54Z

That's also an option, but means a separate code path for creation of the $testapi::username user again. It's the easiest option for now though.

The code path is already present. We just need to set WIZARD_SKIP_USER=1 in the medium

PoC https://openqa.opensuse.org/tests/4522171#details, no code changes

pevik · 2024-10-11T11:03:18Z

PoC https://openqa.opensuse.org/tests/4522171#details, no code changes

I don't see CASEDIR of your fork in https://openqa.opensuse.org/tests/4522171#settings, https://openqa.opensuse.org/tests/4522171/file/autoinst-log.txt points to d9fcc37.

mloviska · 2024-10-11T12:22:52Z

PoC https://openqa.opensuse.org/tests/4522171#details, no code changes

I don't see CASEDIR of your fork in https://openqa.opensuse.org/tests/4522171#settings, https://openqa.opensuse.org/tests/4522171/file/autoinst-log.txt points to d9fcc37.

Why is there a need of casedir when there are no code changes. Does the problem still persists?

pevik · 2024-10-11T13:11:49Z

PoC https://openqa.opensuse.org/tests/4522171#details, no code changes

I don't see CASEDIR of your fork in https://openqa.opensuse.org/tests/4522171#settings, https://openqa.opensuse.org/tests/4522171/file/autoinst-log.txt points to d9fcc37.

Why is there a need of casedir when there are no code changes. Does the problem still persists?

Ah, I'm sorry, now I understand "no code changes meaning". No, problem don't persist any more (got fixed by WIZARD_SKIP_USER=1 in JeOS-for-RPi flavor in https://openqa.opensuse.org/admin/products as you mentioned).

I wonder if we need this PR for other jobs. If yes, shouldn't be the second commit removed? If not, it should be closed, right?

In https://openqa.opensuse.org/tests/4444816#step/firstrun/15 the password typing results in incomplete typing string which yields a cracklib error that the string is too short. - ticket: https://progress.opensuse.org/issues/165800

mloviska · 2024-10-11T14:02:40Z

I wonder if we need this PR for other jobs. If yes, shouldn't be the second commit removed? If not, it should be closed, right?

I guess it is not necessary to have, but it is Hacktober afterall, every PR counts :D

mloviska · 2024-10-11T14:05:41Z

VRs:

opensuse-Tumbleweed-JeOS-for-kvm-and-xen-x86_64-Build20241010-jeos@uefi_virtio-2G -> https://openqa.opensuse.org/tests/4552263
microos-Tumbleweed-MicroOS-Image-x86_64-Build20241010-microos-wizard@64bit -> https://openqa.opensuse.org/tests/4552264
opensuse-Tumbleweed-JeOS-for-AArch64-aarch64-Build20241007-jeos@aarch64-HD20G -> https://openqa.opensuse.org/tests/4552265
microos-Tumbleweed-MicroOS-Image-aarch64-Build20241007-microos-wizard@aarch64 -> https://openqa.opensuse.org/tests/4552333

mloviska requested a review from ricardobranco777 as a code owner August 30, 2024 14:00

mloviska added WIP Work in progress Verification-Runs-needed labels Aug 30, 2024

mloviska force-pushed the fix_ui_user_pass_arm branch from 52e1c0e to a1bc127 Compare August 30, 2024 14:11

mloviska force-pushed the fix_ui_user_pass_arm branch from 94f6ce2 to b60e751 Compare September 2, 2024 09:35

Vogtinator reviewed Sep 2, 2024

View reviewed changes

mloviska force-pushed the fix_ui_user_pass_arm branch 2 times, most recently from 13ef71d to 837e719 Compare September 2, 2024 10:12

pevik mentioned this pull request Sep 4, 2024

Tumbleweed: RPi4: Add LTP containers, commands, syscalls os-autoinst/opensuse-jobgroups#516

Merged

mloviska force-pushed the fix_ui_user_pass_arm branch 2 times, most recently from 73d4023 to 2c2d093 Compare September 5, 2024 11:05

Adapt password typing for slower arch

c956ecb

In https://openqa.opensuse.org/tests/4444816#step/firstrun/15 the password typing results in incomplete typing string which yields a cracklib error that the string is too short. - ticket: https://progress.opensuse.org/issues/165800

mloviska force-pushed the fix_ui_user_pass_arm branch from 2c2d093 to c956ecb Compare October 11, 2024 14:00

mloviska removed the WIP Work in progress label Oct 11, 2024

mloviska removed the Verification-Runs-needed label Oct 11, 2024

mloviska closed this Dec 4, 2024

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Adapt password typing for slower arch #20097

Adapt password typing for slower arch #20097

mloviska commented Aug 30, 2024 •

edited

Loading

mloviska commented Aug 30, 2024

Vogtinator commented Aug 30, 2024

Vogtinator commented Aug 30, 2024

mloviska commented Sep 2, 2024

Vogtinator commented Sep 2, 2024

Vogtinator Sep 2, 2024

mloviska Sep 2, 2024

mloviska Sep 2, 2024

ggardet Sep 2, 2024 •

edited

Loading

mloviska Sep 2, 2024

mloviska Sep 5, 2024

pevik commented Sep 4, 2024

Vogtinator commented Sep 23, 2024 •

edited

Loading

mloviska commented Sep 30, 2024

Vogtinator commented Sep 30, 2024

mloviska commented Sep 30, 2024

Vogtinator commented Sep 30, 2024

mloviska commented Sep 30, 2024

mloviska commented Sep 30, 2024

pevik commented Oct 11, 2024

mloviska commented Oct 11, 2024

pevik commented Oct 11, 2024 •

edited

Loading

mloviska commented Oct 11, 2024

mloviska commented Oct 11, 2024 •

edited

Loading

Adapt password typing for slower arch #20097

Adapt password typing for slower arch #20097

Conversation

mloviska commented Aug 30, 2024 • edited Loading

Verification runs

mloviska commented Aug 30, 2024

Vogtinator commented Aug 30, 2024

Vogtinator commented Aug 30, 2024

mloviska commented Sep 2, 2024

Vogtinator commented Sep 2, 2024

Vogtinator Sep 2, 2024

Choose a reason for hiding this comment

mloviska Sep 2, 2024

Choose a reason for hiding this comment

mloviska Sep 2, 2024

Choose a reason for hiding this comment

ggardet Sep 2, 2024 • edited Loading

Choose a reason for hiding this comment

mloviska Sep 2, 2024

Choose a reason for hiding this comment

mloviska Sep 5, 2024

Choose a reason for hiding this comment

pevik commented Sep 4, 2024

Vogtinator commented Sep 23, 2024 • edited Loading

mloviska commented Sep 30, 2024

Vogtinator commented Sep 30, 2024

mloviska commented Sep 30, 2024

Vogtinator commented Sep 30, 2024

mloviska commented Sep 30, 2024

mloviska commented Sep 30, 2024

pevik commented Oct 11, 2024

mloviska commented Oct 11, 2024

pevik commented Oct 11, 2024 • edited Loading

mloviska commented Oct 11, 2024

mloviska commented Oct 11, 2024 • edited Loading

mloviska commented Aug 30, 2024 •

edited

Loading

ggardet Sep 2, 2024 •

edited

Loading

Vogtinator commented Sep 23, 2024 •

edited

Loading

pevik commented Oct 11, 2024 •

edited

Loading

mloviska commented Oct 11, 2024 •

edited

Loading