At this point, I consider the batch scripts @IBNobody and I worked on to mostly be a failure. They've proven to be unreliable, too dependent on the environment they're being run in, and I've seen far too many examples of people having frustrating issues with them that I haven't been able to help them with. They can also produce misleading and confusing error messages. I've been pointing people to use the WSL for a while now. Eventually, I think we should make a proper replacement for the batch scripts, possibly with an environment in msys2. For now, the WSL method in Windows 10 is far more reliable, and is easy to set up.
I also cleaned up some things in the WSL instructions themselves.
This reworks how the tap-dance feature works: instead of one global
state, we have a state for each tap-dance key, so we can cancel them
when another tap-dance key is in flight. This fixes#527.
Since we have a state for each key, we can avoid situation where a keyup
would mess with our global state. This fixes#563.
And while here, we also make sure to fire events only once, and this
fixes#574.
There is one breaking change, though: tap-dance debugging support was
removed, because dumping the whole state would increase the firmware
size too much. Any keymap that made use of this, will have to be
updated (but there's no such keymap in the repo).
Also, there's a nice trick used in this rework: we need to iterate
through tap_dance_actions in a few places, to check for timeouts, and so
on. For this, we'd need to know the size of the array. We can't discover
that at compile-time, because tap-dance gets compiled separately. We'd
like to avoid having to terminate the list with a sentinel value,
because that would require updates to all keymaps that use the feature.
So, we keep track of the highest tap-dance code seen so far, and iterate
until that index.
Signed-off-by: Gergely Nagy <algernon@madhouse-project.org>
