QA » openQA Project » openQA Tests

pidgin_IRC failing from 20180915-1 multiple times, seems to be missing the a keypress events:
[2018-09-17T03:22:30.0365 CEST] [debug] /var/lib/openqa/cache/tests/sle/tests/x11/pidgin/pidgin_IRC.pm:46 called testapi::send_key
[2018-09-17T03:22:30.0365 CEST] [debug] <<< testapi::send_key(key='alt-`', do_wait=0)
...
[2018-09-17T03:22:30.0579 CEST] [debug] /var/lib/openqa/cache/tests/sle/tests/x11/pidgin/pidgin_IRC.pm:49 called testapi::send_key
[2018-09-17T03:22:30.0580 CEST] [debug] <<< testapi::send_key(key='ctrl-w', do_wait=0)
...
/var/lib/openqa/cache/tests/sle/tests/x11/pidgin/pidgin_IRC.pm:54 called testapi::assert_screen
...
https://openqa.suse.de/tests/2059231#downloads

SLED12SP4 also met the same issue.
See https://openqa.suse.de/tests/2069539#step/pidgin_IRC/21 for details.

From the below log:

[2018-09-18T05:36:40.0247 CEST] [debug] >>> testapi::_check_backend_response: match=pidgin-ctcp-version,pidgin-irc-sledtesting timed out after 30 (assert_screen)
[2018-09-18T05:36:40.0541 CEST] [debug] no candidate needle with tag(s) 'pidgin-ctcp-version, pidgin-irc-sledtesting' matched

It seems that it failed when executing the below code:
assert_screen([qw(pidgin-ctcp-version pidgin-irc-sledtesting)]);

could it be possible it cannot connect to freenode anymore?

due to the recent spambot attack there?

I tried to look at the video but its not easily visible if the connect fails and how

hmm, it seems to turn the bubble green, so it seems connected.

but the ctcp check seems not to come in for some reason.
also not when i try it on tumbleweed pidgin.

the control flow is weird in the test.

the ctcp handling has assert_screen([qw(pidgin-ctcp-version pidgin-irc-sledtesting)]);
but at this time we have not yet joined the #sledtesting channel it matches on.

should this block move a bit lower?

proposal for a patch https://github.com/os-autoinst/os-autoinst-distri-opensuse/pull/5805

(untested)

This one would be hard to debug as it works locally: http://panigale.suse.cz/tests/277

I re-checked logic sequence of the test and I think it is not correct in assert_screen([qw(pidgin-spoofing-ip pidgin-irc-sledtesting)]). There is a chance of pidgin-spoofing-ip, which may occur (but don't have to). But why is there match for pidgin-irc-sledtesting? This shouldn't be there, because this match is supposed to be done little bit later. I think @msmeissn had good idea to remove it and there should be if (check_screen('pidgin-spoofing-ip')) { instead of assert_screen with match_has_tag. SImilar incorrect logic is also for CTCP check.
I did a test run with check_screen and it looks fine http://10.100.12.105/tests/226#.

I also tried to reproduced it manually in virtual machine and ip spoofing check or CTCP didn't occur. So those steps are really optional.

Suggested code change:

     # Warning of spoofing ip may appear
     if (check_screen('pidgin-spoofing-ip')) {
         wait_screen_change {
             send_key is_sle('<15') ? "alt-tab" : "alt-`";
         };
         wait_screen_change {
             send_key "ctrl-w";    # close it
         };
     }

     # CTCP Version and warning about scan may appear
     if (check_screen('pidgin-ctcp-version')) {
         wait_screen_change { send_key "ctrl-w"; }    # close it
     }

Personally I always prefer multi-tag assert_screen with match_has_tag because I think this is the most efficient way for tests without any unnecessary sleep time. So what is wrong with an assert_screen with multiple tags? E.g. instead of check_screen('pidgin-ctcp-version') why not check for both, this screen and the alternative, and only act upon the warning if it shows up.

@msmeissn I will further continue with this one. Could you please close your PR? Okurz had good point and I would
like to explore the other way to fix.

New PR to properly detect pidgin status(and don't waste time by check_screen) was created:
https://github.com/os-autoinst/os-autoinst-distri-opensuse/pull/5853