QA » openQA Project » openQA Tests

I think keymap_or_locale should be started after consoletest_finish

In order to rule out a product issue, I retriggered the KDE-Live test of the previously passed snapshot 0115 - the test now fails the exact same way as 0121 does, implying that a change on the openQA side must cause the failure

0115 verification run: https://openqa.opensuse.org/tests/836012

Oliver reverted the change from https://github.com/os-autoinst/os-autoinst-distri-opensuse/pull/6603

Post revert, the KDE Live test passed the keymap_or_locale test: e.g. https://openqa.opensuse.org/tests/836049

[23/01/2019 12:17:22] <DimStar> okurz: re https://openqa.opensuse.org/tests/835543#step/keymap_or_locale/3 and the tagging you did: is that fix you propose really right? Looking at the screens, we might simply not have the right focus set for 'ret' (which, according the code, is what we should be pressing)
[23/01/2019 12:20:19] <okurz> DimStar: uhm, what fix did I propose? I only changed something for the textmode case, not the x11 scenarios
[23/01/2019 12:20:35] <DimStar> okurz: ah, then I misread the poo ticket :)
[23/01/2019 12:20:54] <okurz> DimStar: ah, sorry. the label was carried over. That should be a different ticket. Let's see
[23/01/2019 12:21:36] <okurz> DimStar: IMHO it's https://progress.opensuse.org/issues/46532
[23/01/2019 12:21:37] <|Anna|> [functional][u] test fails in keymap_or_locale as the test fails to unlock the screenlock when it should not even care about the displaymanager in openQA Tests (action for unassigned) [Workable] Created on: 2019-01-23 | 0% done.
[23/01/2019 12:22:33] <DimStar> indeed, that matches better the issue seen
[23/01/2019 12:22:44] <okurz> DimStar: and I assume that my change did not really influence that. The problem I see is that keymap_or_locale should not even be the first module in case of x11 because there are so many factors that could fail which are not related to "keymap_or_locale", e.g. the display manager, screensaver, desktop runner, xterm, etc., so the module should be scheduled way later – but only in the case of x11
[23/01/2019 12:23:28] <DimStar> https://openqa.opensuse.org/tests/832423 used to not have this problem at least - but that could have been by pure luck of timing
[23/01/2019 12:25:13] <okurz> DimStar: maybe you can help me and I misunderstood something but I don't see any recent changes by *anyone* that would explain the current behaviour except maybe product changes
[23/01/2019 12:27:58] <DimStar> I hope we still have the live image of 0115 on o3 - I retriggered the test that passed, to check the 'product change' theory
[23/01/2019 12:28:23] <DimStar> the entire issue is limited to the KDE Live image, so it's well possible something there changed)
[23/01/2019 12:28:58] <DimStar> like the entire plasma framework version :)
[23/01/2019 12:32:32] <okurz> great, appreciated. If it's likely *not* product changes then please comment in the above ticket https://progress.opensuse.org/issues/46532 so that we know we need to change something
[23/01/2019 12:32:34] <|Anna|> [functional][u] test fails in keymap_or_locale as the test fails to unlock the screenlock when it should not even care about the displaymanager in openQA Tests (action for unassigned) [Workable] Created on: 2019-01-23 | 0% done.
[23/01/2019 12:32:49] <DimStar> will do
[23/01/2019 12:37:05] <DimStar> okurz: https://github.com/os-autoinst/os-autoinst-distri-opensuse/pull/6603 would be a candidate imho (if we're coming to the non-product conclusion - test of 0115 live image is still running)
[23/01/2019 12:37:06] <|Anna|> Github project os-autoinst/os-autoinst-distri-opensuse pull request#6603: "Confirm unlocked screen for consoletest_finish", created on 2019-01-22, status: closed on 2019-01-22, https://github.com/os-autoinst/os-autoinst-distri-opensuse/pull/6603
[23/01/2019 12:37:49] <okurz> DimStar: *ugh*, exactly! So. should I revert?
[23/01/2019 12:38:02] <DimStar> let's wait for https://openqa.opensuse.org/tests/836012#
[23/01/2019 12:41:05] <okurz> DimStar: yep, stuck there. That's enough for me. I did not like the PR anyway as you can read from my comment.
[23/01/2019 12:45:47] <okurz> I retriggered all kde-live tests and will now update the o3 webui :)
[23/01/2019 12:57:20] <DimStar> okurz: thanks for the quick reaction on that

so tests are currently more stable again. Next step: Schedule the "x11"-relevant part only after e.g. "desktop_runner" and "xterm". I understood from mloviska that he does not want to be a test module maintainer anymore because of not having enough recent experience with the test module anymore. Also, as agreed, fits better in the scope of QSF-u

ggardet_arm wrote:

I think keymap_or_locale should be started after consoletest_finish

This will most likely fail because the console will be "used", logged in and cluttered. If we again would have an "extra" console we should be able to trigger this module way later, sure.

okurz wrote:

ggardet_arm wrote:

I think keymap_or_locale should be started after consoletest_finish

This will most likely fail because the console will be "used", logged in and cluttered. If we again would have an "extra" console we should be able to trigger this module way later, sure.

I was thinking of X11 case.

true. But currently the module fulfills both cases. So we could either schedule the same module on different places depending on the DESKTOP variable or use explicit modules for the two specific cases. I favor the second option.

WIP PR created: https://github.com/os-autoinst/os-autoinst-distri-opensuse/pull/6634
feedback highly appreciated

PR should be mergeable now

okurz wrote:

merged.

Created new needle, e.g. for https://openqa.opensuse.org/tests/878013 , and retriggered about 6 upgrade scenarios in https://openqa.opensuse.org/tests/overview?groupid=1&distri=opensuse&build=20190311&version=Tumbleweed

Other than this it looks quite good, e.g.

• textmode: https://openqa.opensuse.org/tests/878045
• kde: https://openqa.opensuse.org/tests/878039

thanks