QA » openQA Project » openQA Tests

I agree with Petr's original comment. Upon opening the application, the first shotwell module (shotwell_import) handles the first-launch window and closes it, while the next two modules (shotwell_edit and shotwell_export) do not expect the welcome window any more and merely assert that the application is launched.

When everything goes well, the welcome screen does not appear any more, even though every shotwell* module calls clean_shotwell() at the end, because the clean function only removes shotwell library files and does not reset shotwell completely.

However, if any of the previous modules fail, the SUT state is restored to the last anchored snapshot, which in this case is right after boot_to_desktop, therefore the next shotwell* module after the failure behaves again as first launch and the welcome window appears again.

I opted for removing the code to start shotwell out of the individual shotwell* tests and instead added a start_shotwell() function into the x11regression library, in accordance with the DRY principle. The function starts the application and subsequently checks whether there is a welcome window pop up, in order to prevent failures when the state is restored from the last good snapshot.

PR: https://github.com/os-autoinst/os-autoinst-distri-opensuse/pull/3530

Verification runs:
SLED12 SP2: http://dreamyhamster.suse.cz/tests/746
SLED12 SP3: http://dreamyhamster.suse.cz/tests/744
Note that in both runs, the shotwell_edit module fails due to another issue (21048), however, the following module, shotwell_export, detects the welcome screen and runs fine.

Needles MR: Not applicable - no new needles were needed.

The shotwell tests are now working as expected in production:

SLED12 SP2: https://openqa.suse.de/tests/1148427#step/shotwell_export/4
SLED12 SP3: https://openqa.suse.de/tests/1148275#step/shotwell_export/4

The failure in shotwell_edit is unrelated to this issue and is owing to poo#21048.