I'm wondering, since it's easy to create side effects in this code, if conditions like that should be use instead.

if (p_event->is_pressed() && (mouse_button.is_null() || status.hovering)) {
	status.press_attempt = true;
	status.pressing_inside = true;
	if (!status.pressed_down_with_focus) {
		status.pressed_down_with_focus = true;
		emit_signal(SNAME("button_down"));
	}
}

if (!p_event->is_pressed()) {
	status.press_attempt = false;
	status.pressing_inside = false;
	if (status.pressed_down_with_focus) {
		status.pressed_down_with_focus = false;
		emit_signal(SNAME("button_up"));
	}
}

Worst case scenario, the button_down or button_up are not emitted but they are a lot less used.

I tried and works, however i'll agree with Hilderin that button_up should be emitted after any press event. I modified the code a bit to use an emit_button_up boolean which is checked at the end of the function above queue_draw (using the existing code) and it seemed to work fine, so i'd recommend that.

I'm wondering, since it's easy to create side effects in this code, if conditions like that should be use instead.

Great suggestion, thanks. I've done exactly this. I've removed the special-case button-up signal for toggle buttons with ACTION_MODE_BUTTON_PRESSED, so with respect to the above discussion about signal order, this is currently preserving the somewhat inconsistent 4.3 behavior. If a fix/change is desired there, I can take a look at it.

I've removed the special-case button-up signal for toggle buttons with ACTION_MODE_BUTTON_PRESSED

Great!

so with respect to the above discussion about signal order, this is currently preserving the somewhat inconsistent 4.3 behavior. If a fix/change is desired there, I can take a look at it.

I'm not sure it causes any issue right now. I suggest we wait for an actual bug report on this.

I'll test a bit more your last commit but it looks good.

Thanks!