Bug 1217618

Created attachment 871046 [details]
screenshot

Created attachment 871047 [details]
nis_server config

we have another test case that have similar issue with refresh the network card setup page.

https://openqa.suse.de/tests/12923972#step/yast2_lan_restart_bridge/26

As always for each and every YaST bug report, we need y2logs.

In the video at 0:18, I see that screenshot, and this is clearly not a refresh problem (and the button is "Finish", not "Next", so there is no more step after this), something is hanging. A refresh problem would be UI-related which this is clearly not.

Probably it attempts to start something in the background, and that hangs.

ARGH. and it's an NFS SERVER that you are trying to configure here, not a NIS server. That's something completely different.

So that YaST NFS server module is called after you configured a NIS server.

And the last step (writing the configuration and starting the NFS server) hangs after you just configured a NIS server.

Do you know if the NIS server is now active? If you now have a working NIS server? Are there any tests if that worked? I don't see anything. A lot of weird things might happen if that didn't work.

What is this test actually testing with the NFS server after the NIS server?

but we use the dev mode and it works https://openqa.suse.de/tests/12924231 . so that might be some refresh issue for the yast2 I think.

The NIS server is now active and works and it can pass with the OpenQA develop mode. https://openqa.suse.de/tests/12924231

(In reply to Huajian Luo from comment #8)
> but we use the dev mode and it works https://openqa.suse.de/tests/12924231 .
> so that might be some refresh issue for the yast2 I think.

All that is known that that time is that something hangs.

There is no refresh problem; that would mean partial screen refresh which is not what we see here.

And please attach y2logs, as requested. Without logs, we can't do anything.

OK, I'll attach one tomorrow, looks like there isn't one in the openQA job's logs, https://openqa.suse.de/tests/12923929#downloads. thanks

And please stop changing the bug subject back to the wrong thing.
It's NFS server that hangs, not NIS server.

I've tried to get the y2log from a branch and uploaded to the attachment.
the job is: https://openqa.suse.de/tests/12931336#step/nis_server/53

thanks.

Created attachment 871078 [details]
yast2log

Created attachment 871081 [details]
y2log for the yast nfs-server call

Extracted from the attached y2logs tarball with

  y2log-merge
  y2log-split

from https://github.com/shundhammer/y2log-util

From that partial y2log:

It starts with calling "yast2 nfs_server qt ...".

> 2023-11-30 02:49:03 <1> susetest(5695) [Ruby] bin/y2start(<main>):22
>   y2base called with
>   ["nfs_server", "qt", "-name", "YaST2", "-icon", "yast"]
> ...
> ...

This can be used to identify the last dialog that was visible:

> 2023-11-30 02:49:08 <0> susetest(5695) [ui-shortcuts] YShortcutManager.cc(checkShortcuts):66
>   Checking keyboard shortcuts
>   Shortcut conflict: 'C' used for YSelectionBox "Directories"
>   Shortcut conflict: 'E' used for YPushButton "Edit"
>   Shortcut conflict: 'H' used for YPushButton "Add Host"
>   Shortcut conflict: 'H' used for YQWizardButton "Help"
>   Shortcut conflict: 'E' used for YQWizardButton "Release Notes"
>   Shortcut conflict: 'C' used for YQWizardButton "Cancel"
> 2023-11-30 02:49:08 <0> susetest(5695) [ui-shortcuts] YShortcutManager.cc(resolveAllConflicts):166
>   Resolving shortcut conflicts
>   Keeping preferred shortcut 'H' for YQWizardButton "Help"
>   Keeping preferred shortcut 'C' for YQWizardButton "Cancel"
>   Keeping preferred shortcut 'E' for YQWizardButton "Release Notes"
>   Couldn't resolve shortcut conflict for YPushButton "Edit" at 0x7faa645a2bf0 - assigning no shortcut
>   Reassigning shortcut 'A' to YPushButton "Add Host"
>   Reassigning shortcut 'R' to YSelectionBox "Directories"
> ...
> ...
> 2023-11-30 02:49:06 <0> susetest(5695) [ui-wizard] YCPWizardCommandParser.cc(isCommand):223
>   Recognized wizard command SetNextButtonID(any) : `SetNextButtonID (`next)
> 2023-11-30 02:49:06 <0> susetest(5695) [ui-wizard] YCPWizardCommandParser.cc(isCommand):223
>   Recognized wizard command SetFocusToNextButton() : `SetFocusToNextButton ()
> ...  
> 2023-11-30 02:49:06 <0> susetest(5695) [Ruby] binary/Yast.cc(ycp_module_call_ycp_function):395
>   Call WaitForEvent
> ...


Finally, this is the last call in the log:

> 2023-11-30 02:49:08 <0> susetest(5695) [Ruby] binary/Yast.cc(ycp_module_call_ycp_function):395
>   Call WaitForEvent


This is simply waiting for an event - a mouse click, a key press, or a synthetic event sent by the libyui-rest-api which is heavily used by OpenQA (I don't know if that is the case for this test).

AFAICS simply no event ever arrives. That's not a "refresh problem" by any stretch of the imagination, nor is it hanging: It's normal UI operation.

I am pretty sure when you do the same thing interactively, you can simply press that "Finish" button, and it will continue.

https://openqa.suse.de/tests/12915763/modules/nis_server/steps/1/src

> sub run {
>     my ($self) = @_;
>     x11_start_program('xterm -geometry 155x45+5+5', target_match => 'xterm');
>     turn_off_gnome_screensaver if check_var('DESKTOP', 'gnome');
>     become_root;
>     setup_static_mm_network($setup_nis_nfs_x11{server_address});
>     zypper_call 'in yast2-nis-server yast2-nfs-server';
> 
>     # we have to stop the firewall, see bsc#999873 and bsc#1083487#c36
>     systemctl 'stop ' . $self->firewall;
> 
>     my $module_name = y2_module_consoletest::yast2_console_exec(yast2_module => 'nis_server');
>     nis_server_configuration();
>     wait_serial("$module_name-0", 360) || die "'yast2 nis server' didn't finish";
>     assert_screen 'yast2_closed_xterm_visible';
>     $module_name = y2_module_consoletest::yast2_console_exec(yast2_module => 'nfs_server');
>     nfs_server_configuration();
      ^^^^^^^^^^^^^^^^^^^^^^^^^^^
      This is where this is stuck

>     wait_serial("$module_name-0", 360) || die "'yast2 nfs server' didn't finish";
>     assert_screen 'yast2_closed_xterm_visible', 200;
>     # In order for the hostname to get the set value via yast2 nis_server, a restart is needed. Otherwise "make"
>     # command won't work as in Makefile, there is a variable that gets it's value from "domainname" command
>     systemctl 'restart network';
>     # NIS and NFS Server is configured and running, configuration continues on client side
>     mutex_create('nis_nfs_server_ready');
>     my $children = get_children();
>     my $child_id = (keys %$children)[0];
>     mutex_wait('nis_nfs_client_ready', $child_id);
>     # Read content of a file created by the client
>     setup_verification();
>     enter_cmd "killall xterm";    # game over -> xterm
> }

Now we add 'SHIFT+F3" as a workaround for it.

```
sub apply_workaround_poo124652 {
    my ($mustmatch) = shift;
    my $timeout;
    $timeout = shift if (@_ % 2);
    my %args = (timeout => $timeout // 0, @_);
    if (!check_screen($mustmatch, %args)) {
        record_info('poo#124652', 'poo#124652 - gtk glitch not showing dialog window decoration on openQA');
        send_key('shift-f3', wait_screen_change => 1);
        check_screen('style-sheet-selection-popup', 10);
        send_key('esc', wait_screen_change => 1);
        # in some verification tests this didn't work, so let's check
        if (!check_screen($mustmatch)) {
            record_info('Retry', "shift-f3 workaround did not solve the problem");
            send_key('alt-f10', wait_screen_change => 1);
            send_key('alt-f10', wait_screen_change => 1);
        }
    }
}

```
and it can pass now.
https://openqa.suse.de/tests/12931367

In the video at 0:16 I can see that the NIS server is configured to handle /etc/hosts, too. That may or may not affect the network for subsequent network operations (like the NFS server configuration).

At 0:17 I see the xterm where those commands were started; first a 'zypper -n yast2-nis-server yast2-nfs-server' that refreshed repos and then installed new versions of those packages:

  nfs-kernel-server-2.1.1-150500.22.3.1.x86_64.rpm
  yast2-nis-server-4.6.0-150600.1.1.noarch.rpm
  yast2-nfs-server-4.6.0-150600.1.1.noarch.rpm

Then it started

  systemctl --no-pager stop firewalld
  yast2 nis-server     (no screenoutput at all)
  yast2-nfs-server

which immediately showed the screenshot of the first attachment here:

--------------------------------------------
  NFS Server Configuration

    NFS Server
      ( ) Start
      (x) Do Not Start

    Firewall not configurable
      ...
      ...

    Enable NFSv4
      [x] Enable NFSv4

      Enter NFSv4 domain name:
      [localdomain              ]

    [ ] Enable GSS Security

             [Cancel] [Back] [OK]

--------------------------------------------  

OpenQA changed the text for the comain name to
   [nfs.openqa.suse.de           ]

The [OK] button changed to [Next] a fraction of a second later, then to [Finish]. By 0:17, the changes were done, and from that point on, the screen was static - no more change.

Some of those details were NOT recorded with OpenQA screenshots, but the change of the button label and the input field content were:

https://openqa.suse.de/tests/12915763#step/nis_server/50
https://openqa.suse.de/tests/12915763#step/nis_server/51
https://openqa.suse.de/tests/12915763#step/nis_server/52

(In reply to Huajian Luo from comment #18)
> Now we add 'SHIFT+F3" as a workaround for it.
> 
> ```
> sub apply_workaround_poo124652 {
...
>     if (!check_screen($mustmatch, %args)) {
>         record_info('poo#124652', 'poo#124652 - gtk glitch not showing
> dialog window decoration on openQA');
> 

What does that mean? That a Gtk error changed the actual screenshot vs. the expected one?


> and it can pass now.
> https://openqa.suse.de/tests/12931367

So we can close this one now?

You can close if you are confirmed it's not a bug. we now applied this workaround for all these screenshot checking and passed now.

Thanks.

If this is a bug, it's not on the YaST side; either on the networking setup or with REST API events not being sent.

Closing.