Commit Graph
6 Commits
Author SHA1 Message Date
James Laird-Wah 39bd760faf Use a single endpoint for HID reports (#3951)
* Unify multiple HID interfaces into one

This reduces the number of USB endpoints required, which frees them up
for other things.

NKRO and EXTRAKEY always use the shared endpoint.

By default, MOUSEKEY also uses it. This means it won't work as a Boot
Procotol mouse in some BIOSes, etc. If you really think your
keyboard needs to work as a mouse in your BIOS, set
MOUSE_SHARED_EP = no in your rules.mk.

By default, the core keyboard does not use the shared endpoint, as not
all BIOSes are standards compliant and that's one place you don't want
to find out your keyboard doesn't work.. If you are really confident,
you can set KEYBOARD_SHARED_EP = yes to use the shared endpoint here
too.

* unify endpoints: ChibiOS protocol implementation

* fixup: missing #ifdef EXTRAKEY_ENABLEs

broke build on AVR with EXTRAKEY disabled

* endpoints: restore error when too many endpoints required

* lufa: wait up to 10ms to send keyboard input

This avoids packets being dropped when two reports are sent in quick
succession (eg. releasing a dual role key).

* endpoints: fix compile on ARM_ATSAM

* endpoint: ARM_ATSAM fixes

No longer use wrong or unexpected endpoint IDs

* endpoints: accommodate VUSB protocol

V-USB has its own, understandably simple ideas about the report formats.
It already blasts the mouse and extrakeys through one endpoint with
report IDs. We just stay out of its way.

* endpoints: document new endpoint configuration options

* endpoints: respect keyboard_report->mods in NKRO

The caller(s) of host_keyboard_send expect to be able to just drop
modifiers in the mods field and not worry about whether NKRO is in use.
This is a good thing. So we just shift it over if needs be.

* endpoints: report.c: update for new keyboard_report format
2018-11-15 22:22:05 -08:00
James Laird-Wah 0c3137d8ff rgb_matrix: continue calling rgb_matrix_indicators() when toggled off
This allows user code to continue to use the matrix for indication, even
when the RGB toggle is off, using rgb_matrix_set_color().

Without this change, it's impossible for user code to use the matrix
when the toggle is off, as any changes get overwritten with black on the
next task cycle, and the indicator code is not called at all.
2018-10-04 21:43:31 -04:00
James Laird-Wah b382076ad1 Keyboard: keyboardio model01: rgbmatrix support (#3989)
* model01: implement RGB matrix feature

LEDs can still be individually controlled if this is switched off.

* model01: use fast banked LED setting commands

* model01: update default keymap to use RGB matrix

* model01: update RGB matrix support to match common_features.mk

* rgb_matrix: include <string.h>

This was missing after commit 4d5705ea6c
which introduces a use of memset().

* model01: make g_rgb_leds weak

This allows users to override which LEDs are configured as
modifiers, or tweak the x/y coordinates, should they so choose.
2018-09-27 20:53:36 -07:00
James Laird-Wah f70f45ee67 RGB Matrix refactoring to open up for new drivers (#3913)
* rgb_matrix: use a driver ops struct

This is intended to avoid #ifdef proliferation on adding more drivers,
eg. model01, which use different architectures.

* rgb_matrix: document driver struct members

* rgb_matrix: remove unused LED testing code

* rgb_matrix: don't build into IS31x drivers unless being used

* rgb_matrix: refactor make config options

This ensures that the necessary files are included for any custom
RGB_MATRIX_ENABLE value, without having to add entries here for specific
boards. This particularly affects model01 because its controller is
integrated and won't be used anywhere else, so it's preferable not to
put it in common_features.mk.

This now validates the value of RGB_MATRIX_ENABLE.

It was necessary to fix an error in ergodox_ez rules.mk using the wrong
comment separator, yielding an invalid value.

* IS31x drivers: don't write the control registers all the time

This is only needed when they are changed. This is done in init() and
board- or keymap-specific code is free to make further changes.

* rgb_matrix: move structs from chip drivers to rgb_matrix_drivers.c

This approach is specific to the rgb_matrix functionality, so keep it
neatly separated from the raw chip drivers.
2018-09-27 10:40:18 -04:00
James Laird-Wah 03b8ce206d Keyboard: add Keyboardio Model 01 (#3900)
* Keyboard: add Keyboardio Model01

This port implements key scanning and everything in their default
keymap.

It doesn't implement mouse warping; QMK can't do that (yet).

LED control is mostly not implemented. The ability to set all LEDs is
included because they can get stuck on coming from the bootloader
otherwise. Single LED control is also implemented for numpad indication.
The scanners also support batch LED transfer which is what you'd need if
you wanted to do fast effects. Gamma correction is also not implemented,
but is present in the original firmware. The necessary info for further
implementation is in the KeyboardioScanner module for Kaleidoscope.

To install on your keyboard:

    make model01:avrdude

When prompted, hold the "prog" key on the keyboard to put it into
programming mode. This can also be achieved by holding the "prog" key
while plugging in the keyboard. This works even if the firmware is
corrupt or missing.

Hot plugging the halves seems to work fine, but there is no explicit
support for eg. making sure the matrix scan rate is reconfigured.

* model01: clean up includes and include guards

Uses #pragma once everywhere.

* model01: split LED and matrix code

This makes space if someone wants to implement better LED support
later on, the keyboard is a lot more capable than the current code.

* model01: separate I2C timeouts for matrix vs. LED

If the scanners have no data they don't ACK reads and just time out.
So we want a pretty short timeout to keep scan rates high.

Meanwhile the LED transfers might take longer - I don't know though,
so here we are conservative.

* model01: implement better LED control

- gamma correction from the manufacturer's firmware
- suitable delays to allow back to back LED writes
    - this is fast enough to write the whole keyboard without noticeable
    delay, in my experience
- minor bug fix: RHS Fn key was not addressable

* model01: add license to wire-protocol-constants.h

* model01: replace gamma LUT

The original was of unclear license origin. This one is functionally
identical and includes a generator should people wish to adjust it.

* model01: use the already-present CIE1931 lightness curve

...rather than baking in another gamma curve.

It's said that CIE1931 is the right thing to be using rather than gamma.
OK. Let it be so.

* model01: call matrix_init_user() from matrix_init_kb()

* model01: remove inapplicable config options from rules.mk

* model01: readme.md: update build environment links
2018-09-17 15:49:20 -07:00
James Laird-Wah b1f11636c6 docs/contributing: add #pragma once to Coding Conventions
This supersedes the older include guard mechanism since all the
compilers now support it. It's easier for new users and less prone
to error.
2018-09-15 09:37:06 -07:00