action #31048
closed
action #33208: [sle15sp1][desktop] Integrating of fix for openQA firefox test cases issues.
[sle15][desktop] test fails in firefox_smoke due to performance issue
Added by qkzhu almost 7 years ago.
Updated almost 7 years ago.
Category:
Bugs in existing tests
Description
Observation¶
openQA test in scenario sle-15-Installer-DVD-x86_64-desktopapps-firefox@64bit-virtio-vga fails in
firefox_smoke
We should replace type_string $site. "\n";
with type_string_slow $site. "\n";
- Status changed from New to In Progress
This is an autogenerated message for openQA integration by the openqa_review script:
This bug is still referenced in a failing openQA test: desktopapps-firefox
https://openqa.suse.de/tests/1475990
- Status changed from In Progress to Resolved
- Parent task set to #33208
- Status changed from Resolved to In Progress
subroutine firefox_check_default
is not stable. We have too many needles with tag firefox_clean
and this needle will be matched before the default browser window pop up.
I will improve the stability by using:
`- assert_screen [qw(firefox_default_browser firefox_trackinfo firefox_readerview_window firefox_clean)], 120;
- check_screen 'firefox_default_browser';
if (match_has_tag('firefox_default_browser')) {
wait_screen_change {
assert_and_click 'firefox_default_browser_yes';`
As commented in https://github.com/os-autoinst/os-autoinst-distri-opensuse/pull/4645#discussion_r175716539 I don't think this is a good idea. Many people - mainly QAM - invested a lot of time to get the firefox tests working properly so we should not accept a revert lightly. Simply stated: Sorry, check_screen with default timeout is not acceptable. I do not understand yet how too many needles for firefox_clean make a problem unless a dialog on the SUT would disappear in time. However, looking at https://openqa.suse.de/admin/needles I can confirm that there are many needles for firefox_clean where many of them seem to have never matched. I will try to clean up some of them. Maybe this helps already.
I deleted many firefox_clean tests. Can you try to retrigger tests and see if this helps?
Hi Oliver,
Thanks for your advice, FF cases works fine in QAM tests however it seems more unreliable in SLED15.
I thought the reason is gdm will eat more RAM after 3.16 (according to: https://bugzilla.gnome.org/show_bug.cgi?id=747339).
After some discussion with Yifan, we agree that we should allocate more RAM to the worker of SLED15, and the situation got much better now (with QEMURAM=2048):
https://openqa.suse.de/tests/1560033
My modification to firefox_check_default
is not necessary anymore, I will update my PR.
Thanks!
Ah, that's also a very good finding
- Status changed from In Progress to Resolved
Also available in: Atom
PDF