fix(split): Make locality work for nested behaviors

[caksoylar]

This is another attempt to solve #1494. My understanding is that #1630 only covers the case of macros invoking global behaviors. I incorporated the changes in it and attempted to address the existing comments.

In addition to the other PR, this should cover sticky keys, hold-taps, mod-morphs and work with source+global locality behaviors. I tested the reset behaviors with hold-taps and tap dances so far.

I am not familiar with these parts of the code; so far I stuck an extra field source wherever I saw a position and passed it around. Specifically, it is a new field in zmk_behavior_binding_event. Any recommendations are welcome on how better to do it.

TODO:

• Test source local behaviors (reset)
  • macros
  • mod-morphs
  • hold-taps
  • tap dances
  • sticky keys
• Test global behaviors (backlight)
  • macros
  • mod-morphs
  • hold-taps
  • tap dances
  • sticky keys
• Handle combos (at least get them working with global behaviors)

[caksoylar]

This is ready for review: I tested all the combinations I can. (And fixed one bug where I learned the purpose of the struct copy the hard way.)

One thing missing is sensor rotate since I don't have hardware with encoders but it is using the behavior queue just like the macros, so I assume it should be fine. I'll ask for testers on Discord just in case.

[Nick-Munnich]

Missing a docs adjustment to the new split keyboards page

[caksoylar]

@Nick-Munnich removed the note, thanks!

One thing missing is sensor rotate since I don't have hardware with encoders but it is using the behavior queue just like the macros, so I assume it should be fine.

This was naive, I looked into it and passing through the source information in the sensor events is not trivial and I don't feel confident enough to do it given I can't even test them. In 4e12392 I hardcoded them to trigger on the central; compared to status quo it should fix using global behaviors like &rgb_ug, at least. I think making source local ones work can be tackled in the future.

[petejohanson]

A few comments after a first pass. Thanks!

[caksoylar]

I got confirmation from a user on Discord that encoders also work with a global behavior now. To sum up, this is the current state after this PR, according to my testing and others'.

Working with global and source-local behaviors:

• Mod-morphs
• Hold-taps
• Macros
• Tap dances
• Sticky keys

Working with global behaviors:

• Combos¹
• Sensors/encoders²

I am not planning to tackle above two for source-local in this PR (see below) so it is complete from my POV.

Footnotes

1. Making source local work with combos requires deciding the locality of combos from the set of key positions that trigger it, some discussion here ↩

2. The code change required for this is separate and since I don't have the hardware to iterate on, I think it'd be better to tackle this separately ↩

[petejohanson]

Thanks for sticking with this! A few thoughts on the implementation.

[petejohanson]

Suggested change

	#include <zmk/ble.h>
	#if ZMK_BLE_IS_CENTRAL
	#if ZMK_BLE_IS_CENTRAL
	#include <zmk/ble.h>

[caksoylar]

This #include <zmk/ble.h> is actually to get ZMK_BLE_IS_CENTRAL itself (and I believe nothing else, actually).

[caksoylar]

Went over the review items, squashed commits to be a bit saner and gave it a drive on my test split to check the previously tested items (no encoder, but everything else).