[widget audit] Slider

[mhsmith]

Audit checklist

• Core tests ported to pytest
• Core tests at 100% coverage
• Core docstrings complete, and in Sphinx format
• Documentation complete and consistently formatted
• cocoa backend at 100% coverage
• winforms backend at 100% coverage
• gtk backend at 100% coverage
• iOS backend at 100% coverage
• android backend at 100% coverage
• Widget support table updated

[mhsmith]

I've rebased this PR onto #1707 in order to get the macOS probe classes and the exit status line. Once that PR is merged, the diff for this PR should become a lot smaller.

[mhsmith]

Replaced get_value and set_value with simpler implementations based on the WIndows backend. There's no need to check for tick_count being less than 2, because the interface layer now prevents that.

[mhsmith]

@freakboy3742: Since #1707 is stabilizing, I think it would be worth you reviewing this PR now. You can see the diff against the upstream branch by selecting only the last few commits like this.

[freakboy3742]

The core of the Android implementation makes sense; however, the examples/slider demo shows some gaps in the Android implementation. The second last slider doesn't appear to report correctly for values <0 or >9. The last slider doesn't work at all - I'm guessing that's missing functionality, though.

I'm not sure I'm following why the sync_values change is needed, though. Some of the code that has been promoted to the interface layer makes sense (like the "tick count < 2" checks), but the rest of the sync approach seems really roundabout - an event occurs on the impl, which tells the interface to sync, which asks the impl for it's value. The int->float conversion is still needed on each backend, so there's no avoidance of duplication there; and it doesn't seem to impact on the float vs double distinction with iOS (although I'm not sure that's a distinction we need to care about - is anyone going to be able set a slider with precision where the difference between double and float will matter?) From what I can tell, the sync suppression stuff is only required because of this round trip as well. What am I missing here.

Some other things I noticed:

• The tests only have 32% coverage of the winforms slider module; 82% on macOS. I'm guessing that means coverage is low on Android as well, but we're not reporting on that at present. It looks like at least some of this is because of on_press/on_release not being tested
• The sync_value changes break the iOS implementation of slider (possibly a moot point depending on where we land on making this change)

[mhsmith]

The second last slider doesn't appear to report correctly for values <0 or >9.

This was actually a problem with the Label: see #1289 (comment).

[mhsmith]

There's a failure in test_flex_horizontal_widget_size on a couple of platforms which I'll look at tomorrow, but everything else is now ready for review.

[freakboy3742]

I've dropped a bunch of comments, but they're almost entirely cosmetic - the broad strokes look really solid. A good chunk of the comments are about introducing redraws into the testbed tests; plus a handful of requests for explanatory docs; but in terms of core functionality, there's not much to fault here.

[freakboy3742]

Looks good.

I've made a small tweak to the slider __init__ docstring, and merged with main to make sure that the progress bar changes didn't break anything; but otherwise - I think our long nightmare of slider testing is over :-)