There are a number of places in our code where clang-format tries too hard, and
destroys human readability, so I protected them with `clang-format off`
directives.
Signed-off-by: Michael Richters <gedankenexperimenter@gmail.com>
First, we add an `uinitialized_byte` prop to `BaseProps`, so implementations
that don't use `0xff` as default can set it to whatever they use. Do keep in
mind that plenty of code still assumes `0xff` is the uninitialized byte, but
this way we have a starting point.
Then, we add an `isSliceUninitialized()` function, which checks whether a given
slice is completely made up from uninitialized bytes or not.
Fixes#1145.
Signed-off-by: Gergely Nagy <algernon@keyboard.io>
This standardizes namespace closing brackets for namespace blocks. Each one is
on its own line, with a comment clearly marking which namespace it closes.
Consecutive lines closing namespace blocks have no whitespace between them, but
there is one blank line before and after a set of namespace block closing lines.
To generate the namespace comments, I used clang-format, with
`FixNamespaceComments: true`. But since clang-format can't exactly duplicate
our astyle formatting, it made lots of other changes, too. To isolate the
namespace comments from the other formatting changes, I first ran clang-format
with `FixNamespaceComments: false`, committed those changes, then ran it again
to generate the namespace comments. Then I stashed the namespace comments,
reset `HEAD` to remove the other changes, applied the stashed namespace
comments, and committed the results (after examining them and making a few minor
adjustments by hand).
Signed-off-by: Michael Richters <gedankenexperimenter@gmail.com>
I ran `include-what-you-use` wherever possible, and made a number of manual
changes required to get things working with the new header organization.
Signed-off-by: Michael Richters <gedankenexperimenter@gmail.com>
By specifying the type of the `EventHandlerResult` enum as `uint8_t`, it only
takes up one byte instead of two, saving a significant amout of PROGMEM.
Signed-off-by: Michael Richters <gedankenexperimenter@gmail.com>
This removes the pre-`KeyEvent` handler hooks for `onKeyswitchEvent()` and
`beforeReportingState()`. Any third-party plugins that still use either should
be updated to use the new handlers instead.
Signed-off-by: Michael Richters <gedankenexperimenter@gmail.com>
This removes all the core code that called the pre-`KeyEvent` hook functions
`onKeyswitchEvent(key, addr, state)` and `beforeReportingState()`. Any plugins
that were still using these functions should use the newer equivalents instead:
`onKeyEvent(event)` & `beforeReportingState(event)`.
Signed-off-by: Michael Richters <gedankenexperimenter@gmail.com>
This removes the `key_events.*` files that once contained the main
`handleKeyswitchEvent()` function, and all references to it. Because
`key_events.h` was included in the main `Kaleidoscope.h` header file,
`key_defs.h` and `keyswitch_state.h` were added to that header so that other
code that relies on those things being included via `Kaleidoscope.h` will
continue to work.
Signed-off-by: Michael Richters <gedankenexperimenter@gmail.com>
This function was mainly intended for internal use by Kaleidoscope's event
handling functions. The logic that it used to provide is now handled by
`Runtime.handleKeyEvent()`, and it is not generally necessary or recommended for
plugin or sketch code to directly modify HID reports any longer.
Signed-off-by: Michael Richters <gedankenexperimenter@gmail.com>
This removes several vestigial functions from the `Layer` class:
- `.handleKeymapKeyswitchEvent()` & `.eventHandler()`, which have been replaced
with the `.handleLayerKeyEvent()` function that operates on `KeyEvent` objects.
- `.lookup()`, which has been replaced by either `Runtime.lookupKey()` or
`Layer.lookupOnActiveLayer()`, depending on the specific purpose.
- `.updateLiveCompositeKeymap()`, which referred to a structure that has been
replaced. `live_keys.activate()` serves a similar purpose, but in most cases
shouldn't be called directly by user code.
Signed-off-by: Michael Richters <gedankenexperimenter@gmail.com>
Storage must be page aligned and must span full pages. Adjust the default
storage size accordingly, so it is a multiple of 4096.
Signed-off-by: Gergely Nagy <algernon@keyboard.io>
The commit method of the GD32Flash storage driver was accidentally left empty,
disabling the functionality. Call the parent's commit method to restore it.
Signed-off-by: Gergely Nagy <algernon@keyboard.io>
This change prevents `KeyAddrEventQueue::remove()` from shifting values in
memory out of bounds of its arrays if `shift()` is called on an empty queue. It
also adds a check to be sure that the entry removed is in the queue.
Signed-off-by: Michael Richters <gedankenexperimenter@gmail.com>
This lets us remove some awkward code from `handleKeyswitchEvent()`, because as
long as the default value (which triggers a keymap lookup) was the same as
`Key_Masked`, it wasn't sufficient for an `onKeyswitchEvent()` handler to change
`event.key` to `Key_Masked`, because that would be interpreted by
`handleKeyEvent()` as a signal to do a keymap lookup.
This also makes it more consistent with other parts of the code. The values
`Key_Undefined`, `Key_Inactive`, and `Key_Transparent` are all the same, and
with this change they are each interpreted the same way by code that encounters
them. In a keymap lookup, if an active layer has a `Key_Transparent` value, we
continue searching for a different value on another layer. In the live keys
array, if we find a `Key_Inactive` entry, we look for a value from the keymap.
And with this change, when processing a key event, if it has a `Key_Undefined`
value, we look for a value from `active_keys` (and then from the keymap layers,
if nothing is found there).
Signed-off-by: Michael Richters <gedankenexperimenter@gmail.com>
To allow us to use any other HID than KeyboardioHID, the base device _must_ use
something else (practically, the Base HID), otherwise we'll get a compile error
when building on a platform that KeyboardioHID does not support, even if we do
not use KeyboardioHID. We get that error, because the base class references it
anyway.
As such, lets use the base HID as default, and adjust all users of KeyboardioHID
to explicitly set that: Virtual, Dygma Raise, and ATmega32U4Keyboard. Everything
else derives from the last one, so they're covered with just the change to
ATmega32U4Keyboard.
Signed-off-by: Gergely Nagy <algernon@keyboard.io>
If `HID_BOOT_PROTOCOL` is undefined, define it ourselves (the value is set by
the USB standard, so we can do that). This allows compiling the base class
without KeyboardioHID, letting our HID base driver be completely independent of
it.
Signed-off-by: Gergely Nagy <algernon@keyboard.io>
This function is completely unused, and should not have been exposed to begin
with. Neither in the base class, nor in specific implementations.
The `getReport()` method ties us to a specific report datatype, both in name,
and in shape. That's not something we want, we'd rather have the base class be
HID-library agnostic, which it can't be with `getReport()` present.
Luckily, the function is completely unused, and as such, is safe to remove
without deprecation.
Signed-off-by: Gergely Nagy <algernon@keyboard.io>
Instead of doing the stop in the base class, delegate it to the specific
implementation. Do this to avoid depending on the exact shape and layout of the
mouse report within the base class.
Signed-off-by: Gergely Nagy <algernon@keyboard.io>
Other parts of Kaleidoscope require the NKRO and Boot keyboard classes to have
an `isKeyPressed()` method, which our base classes did not provide - until now.
Signed-off-by: Gergely Nagy <algernon@keyboard.io>
The `layer_state_` bitfield is no longer necessary now that we have
activation-order layers. Removing it reduces clutter and saves a modicum of
PROGMEM & RAM.
Signed-off-by: Michael Richters <gedankenexperimenter@gmail.com>
This event handler is useful for plugins that need to react to events, but
should wait until after those events are fully processed before doing so. This
is useful for OneShot, which needs to keep keys active until after events that
trigger their release. The `afterEachCycle()` hook is unfortunately
insufficient for this purpose, because the same event could trigger multiple
plugins (e.g. TapDance & OneShot) to resolve events, and the OneShot should
apply only to the first ensuing report.
Signed-off-by: Michael Richters <gedankenexperimenter@gmail.com>
This fixes a problem with a plugin that returns `ABORT` from its
`onKeyswitchEvent()` handler, for a masked key addr. I'm convinced that a
better solution is to switch from using `Key_NoKey` (a.k.a. `Key_Masked`) to
using `Key_Transparent` as the default for new events, and thus, the value that
signals that a lookup should be done, but this at least fixes the bug for now,
with a much less intrusive change.
Signed-off-by: Michael Richters <gedankenexperimenter@gmail.com>
`Runtime.hasTimeExpired()` had a minor flaw. Because it was comparing two values
using `>` instead of `>=`, it meant that a timeout set at 20ms wouldn't actually
time out until 21ms elapsed.
Signed-off-by: Michael Richters <gedankenexperimenter@gmail.com>
Jesse Vincent
Michael Richters
Gergely Nagy
Florian Bruhin