Implement DirtyValues on Tile, Quick Settings, Steps and Batt Info screens

[SteveAmor]

Implemented as an alternative to #2257

[github-actions]

Build size and comparison to main:

Section	Size	Difference
text	385280B	48B
data	944B	0B
bss	22640B	0B

Run in InfiniEmu

[SteveAmor]

@mark9064 I would very much appreciate your review of this.

I see that statusIcons.Update() already uses dirty values to only update if there is a change

InfiniTime/src/displayapp/widgets/StatusIcons.cpp

Line 34 in 71d1f5b

void StatusIcons::Update() {

So I only needed to update the time at the top of the screen to use dirty values.

I thought I might have to use override for the UpdateScreen but the compiler didn't like that

InfiniTime/src/displayapp/screens/settings/QuickSettings.h

Line 33 in 71d1f5b

void UpdateScreen();

If I got it right, I'll move on to a few other screens - like the apps screen which has a similar layout with status icons and the time.
It works as expected on my watch.

[mark9064]

This is a nicer solution!

The logic with dirtyvalue all looks good. When you said you thought you might have to use override, what did you mean by that?

There's a few other places that have Refresh() in the wrong place - maybe you will find this useful for your changes: mark9064@43bfddc

[SteveAmor]

This is a nicer solution!
The logic with dirtyvalue all looks good.

Thank you

When you said you thought you might have to use override, what did you mean by that?

I saw in other screens that use DirtyValues that the header file has void Refresh() override; and I'm getting out of my depth with that one.

There's a few other places that have Refresh() in the wrong place - maybe you will find this useful for your changes: mark9064@43bfddc

Very useful, thank you.

I have updated Tile.cpp (not pushed yet) and then started on BatteryInfo and the lvgl was not lining up. Now I can see why!
BatterInfo needs a lot of work to use DirtyValues. I think I need to track the charging/discharging status to see if that changes which is more of a refactor than I thought it would be. Rather than track changes in charging.

[mark9064]

Ahhhhh right I see

The Screen class, which (almost?) all app screens inherit from provides a dummy Refresh method and a short helper which you can pass to LVGL (see Screen.cpp). Refresh is purposefully a virtual method: if a method is virtual, it means that classes inheriting from it are allowed to override it (i.e. replace the contents of the method with something else), and when you do override it you need to use the override keyword.

Since QuickSettings is using its own helper for LVGL and its own UpdateScreen, it's not overriding anything. But it would probably make sense to use Refresh and the normal helper here like other apps, so maybe worth changing?

[SteveAmor]

Got it. Thank you. I had started to refactor using Refresh() and backed out for two reasons. First was the "override" or not question and not really understanding it (and not wanting to use the excuse "well it compiled so must be ok") and second because I wanted to go back and tidy up the few cases that use UpdateScreen() in a PR just for that change. And then I saw your lvgl PR and all I could think about was the merge conflicts 🤣

[mark9064]

LOL don't worry that isn't a PR yet. I can rebase around any conflicts, I don't mind

[SteveAmor]

@mark9064

Forced push above was to refactor for Refresh().

I now have tested changes to Tile.cpp, Steps.cpp and BatteryInfo.cpp that include dirty values for screen changes, lvgl refactoring and screenUpdate replaced with Refresh() where required. I can add them to this PR, raise individual PRs for each screen or a new PR with those three screens. What would you prefer me to do?

[mark9064]

I think rolling them into this PR should be fine - sounds like the overall change isn't going to be huge

[SteveAmor]

After the discussion about dirty values in #2450 (comment) I have updated the steps goal to be a dirty value.