Focus: defer focus border until first directional navigation

[bitcoin3us]

Make add_focus_border() (mpos/ui/focus.py) skip the focus ring on pointer/touch devices.

Why

add_focus_border() draws a focus highlight and adds the widget to the default focus group — a navigation aid for keypad/encoder devices, where it shows which widget the moving highlight is on. On a pointer/touch UI there is no moving highlight: the ring only appears after a tap and then lingers around the tapped widget.

Concretely, this is why the Lightning Piggy balance shows a gold rectangle after you tap it (the balance is add_focus_border-ed in 0.6.0, and tapping it cycles the denomination). On a touchscreen that ring is visual noise, not a cue.

What

def add_focus_border(widget, ...):
    from .input_manager import InputManager
    if InputManager.has_pointer():
        return
    ...

• Reuses the existing InputManager.has_pointer() (no new detection).
• Full no-op on touch: skips both the FOCUSED/DEFOCUSED border callbacks and the focus-group registration. Widgets remain interactive through their own CLICKABLE flag / click handlers, which are independent of the focus group.
• Keypad/encoder devices are unchanged — has_pointer() is False there, so the border is added exactly as before.

Hybrid devices (both touch and a keypad) would also skip the ring; that edge case can be refined later if needed — flagging it explicitly rather than silently.

Testing

• tests/test_focus_border_touch.py added: asserts add_focus_border() registers no callbacks when InputManager.has_pointer() is patched True.
• Verified the logic in a CPython stub harness (lvgl + InputManager stubbed): touch present → 0 callbacks (no-op); no pointer → 2 callbacks registered.
• Confirmed InputManager.has_pointer() returns True on a real touchscreen device (ESP32-S3), so the guard fires there.
• ⚠️ I did not run ./tests/unittest.sh locally — it needs a freshly built desktop binary with this change and my local build is stale. Please run ./tests/unittest.sh tests/test_focus_border_touch.py in CI / on a current build.

Merge checklist

• CHANGELOG: ✅ added under Future release → Frameworks
• App version / MANIFEST: n/a (framework change)
• MAINTAINERS.md / boards: n/a
• Settings migration: n/a
• Unit tests: ✅ added (no-op path); see testing caveat above
• Frameworks docs (docs.micropythonos.com): not updated — add_focus_border behavior is now device-conditional; happy to add a note if there is a docs page for it.

[ThomasFarstrike]

Would indeed be nice to find a solution for this and I guess it would work but indeed:

Hybrid devices (both touch and a keypad) would also skip the ring

...but I was thinking, how about on all devices, we still add the widget to the focusgroup, but don't do anything with the border (no set_style_border_color etc) until the user uses "move_focus_direction" , meaning they pushed the joystick?

So first they don't see anything special, but as soon as they press the joystick once the set_style_border_color , set_style_border_width , set_style_border_opa and set_style_radius are applied so it behaves as normal?

[bitcoin3us]

Good call — adopted your approach. Pushed a rewrite (force-update, rebased onto current main, so the CHANGELOG conflict is resolved).

Instead of skipping add_focus_border() on touch devices, every widget is still added to the focus group, but the border is now deferred until the first directional navigation:

• mpos/ui/focus.py: module flag _focus_nav_active (starts False) + enable_focus_borders(). _focus_border_handler() early-returns while the flag is False, so no border is drawn.
• mpos/ui/focus_direction.py: move_focus_direction() calls enable_focus_borders() at the top, so the first joystick/keypad press flips it on and the resulting FOCUSED event draws the border as normal.

Behavior:

• touch-only (focus by tapping) → never shows a ring;
• keypad/encoder → shows it the moment they navigate;
• hybrid → shows it only once the joystick is actually used — which fixes the caveat you flagged.

add_focus_border itself is back to its original shape (no capability check), so it's a smaller change than my first version.

Testing: tests/test_focus_border_deferred.py asserts no border before nav and a border after enable_focus_borders(). Verified the logic in a CPython stub (both states pass); I did not run ./tests/unittest.sh locally (needs a fresh desktop build), so please run it in CI / on a current build.