TestServices.testFilter pixel change

[cockpituous]

The job fedora-40/updates-testing failed on commit c7fddca.

Log: https://cockpit-logs.us-east-1.linodeobjects.com/pull-0-c7fddca2-20241019-013057-fedora-40-updates-testing/log.html

[martinpitt]

Uninteresting

[martinpitt]

Meh, this is just the scrolling, and I can't think of what could influence this in updates-testing. But it repeats.

[jelly]

This issue rings a bell, I recall this was introduced in a PF update and we have a workaround for it but it seems the workaround doesn't work well enough?

[martinpitt]

I also see this in "normal" PRs, like in https://cockpit-logs.us-east-1.linodeobjects.com/pull-21121-a74fe387-20241021-073359-fedora-40-other/log.html

[martinpitt]

@jelly I assume you mean PR #20962 ? Not sure if it's really the reload, it's "just" some different scroll position?

[martinpitt]

This reproduces locally when I revert commit f2f381a. But neither with nor without that hack, the actual nav_scroll_btn loop never runs.

I can interactively reproduce it with Firefox: Switch "Galaxy S10/S10+ Android 11", or just resize the window. I can reload as often as I want, and there are no scroll buttons.

With interactive Chromium I see the buttons after a reload, but sometimes I can reproduce the failure with TEST_SHOW_BROWSER=pixels or default chromium-headless.

[jelly]

Ok, so my F5 doesn't help anymore or accidentally helped. Then we should rework this, keep in mind that I think this was a PF bug?

[martinpitt]

This is https://www.patternfly.org/components/navigation#tertiary-deprecated . The example in the docs is way too wide to fit on a single screen, so the bug can't easily be reproduced there. But with it being deprecated, I suppose it doesn't make much sense to file a bug. We could try porting to https://www.patternfly.org/components/navigation#horizontal-subnav

I forked the codesandbox example to only have 4 entries, and can't reproduce the issue with that either. Perhaps we messed up our breakpoints?

[jelly]

I created patternfly/patternfly-react#10897 back then and this has been fixed apparently.

[martinpitt]

Ooh, thanks for that pointer! patternfly/patternfly-react#11099 is included in 5.4.2, so let's try an update.