Add scaling tests

[tarek-y-ismail]

No description provided.

[codecov]

Codecov Report

Attention: Patch coverage is 25.00000% with 3 lines in your changes missing coverage. Please review.

Project coverage is 68.37%. Comparing base (1a39b71) to head (31dbc8a).
Report is 24 commits behind head on main.

Files	Patch %	Lines
mir-ci/mir_ci/wayland/virtual_pointer.py	25.00%	3 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #140      +/-   ##
==========================================
- Coverage   68.45%   68.37%   -0.08%     
==========================================
  Files          16       16              
  Lines         875      876       +1     
  Branches      126      126              
==========================================
  Hits          599      599              
- Misses        250      251       +1     
  Partials       26       26              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[Saviq]

Only looked at the CI results for now, but the direct cause for the failure is:

E               AssertionError: gtk4-demo died without being waited for or killed

But that's because it's just not installed (we should improve that error report to say the executable wasn't found).

[tarek-y-ismail]

Only looked at the CI results for now, but the direct cause for the failure is:

E               AssertionError: gtk4-demo died without being waited for or killed

But that's because it's just not installed (we should improve that error report to say the executable wasn't found).

That's weird, I've been using it to test for the past two days. I recall seeing another error in the output of the robot near the top. Have you seen that?

[Saviq]

That's weird, I've been using it to test for the past two days. I recall seeing another error in the output of the robot near the top. Have you seen that?

On your machine, yes - but who installed gtk-4-examples in CI? :)

[Saviq]

@tarek-y-ismail and with these tweaks your tests pass (the failures are only on neverputt, that's a known issue fixed in Mir already.)

You can see the test results better in the "Test results" check. It sometimes gets attached to PreCommit for some reason...

On failure, you can also download the test-results archive from the "Test" workflow and dig for the appropriate "log.html" for your test - e.g. at the bottom here for a previous run:

https://github.com/canonical/mir-ci/actions/runs/10166013391

You'll then see a video of the run as well as the screenshot that failed to match - this way you can cut out exactly the template you need in CI.

All this... will get better soon when we integrate it all with the certification team's tooling.

[Saviq]

I think you should be able to reduce the number of templates needed to just two per scale:

1. top-left portion of the app that fits at the scale in the minimum resolution
2. the expected window title after clicking the button

You can then use a static offset (multiplied by scale) to reach the intended button.

Updating the templates will require less work, then.

---

You got hung up on the "Combo Boxes" button, but at 1024x768 at scale >=1.75 it's not on screen when maximized, so maybe use/expect a different button? One that's on screen across all the cases you want to handle?

---

And I think you might have uncovered a bug in the initial placement logic - we probably should place the window's (0, 0) on screen, not it's lower right corner - when it doesn't fit.

[tarek-y-ismail]

You got hung up on the "Combo Boxes" button, but at 1024x768 at scale >=1.75 it's not on screen when maximized, so maybe use/expect a different button? One that's on screen across all the cases you want to handle?

Regarding this, I wanted to aim for a button on a far edge of the application as that's where issues usually pop up when it comes to scaling, hence why I originally used "simple constraints". I worked around the button not being visible at higher scales by scrolling the list down and matching again if the button is not found

I think you should be able to reduce the number of templates needed to just two per scale:

I think I'll try this out. Thanks for the suggestion :)

[Saviq]

Regarding this, I wanted to aim for a button on a far edge of the application as that's where issues usually pop up when it comes to scaling, hence why I originally used "simple constraints". I worked around the button not being visible at higher scales by scrolling the list down and matching again if the button is not found

I think it would be appropriate to just go for the last visible button, at the minimum supported resolution. You can also aim for different buttons on different scales.

[Saviq]

I think it would be appropriate to just go for the last visible button, at the minimum supported resolution. You can also aim for different buttons on different scales.

Ping on this one.

[Saviq]

Handful of nits.

I'm looking into why they're failing in CI - there's some subtle differences, we may want to avoid having the full screenshots as templates, as e.g. Frame will always maximize.

[tarek-y-ismail]

Additional changes reviewed, LGTM! Will just rebase to clean up the commit history then merge.