- In general this shouldn't cause any issues
- BIOSs should ignore
- OSs should happily use
Still, this masks the EHCI hand-off that Windows and Linux are starting to do.
A better solution is needed to force devices to re-negotiate their mode.
- Use commit number as the basis for bcdDevice field (maximum is 2^16)
- Change iSerial from commit date to chip Unique ID
- Update both Firmware and Bootloader to use the same scheme (bcdDevice and iSerial)
- Cleanup of Bootloader code
- Split out build info CMake commands to a new file
USB - General
- Refactored descriptors
- Enabled/Disable USB endpoints
- Added debug flags for special features
- Code cleanup
- Interface count calculation based off of enabled endpoints
- Delayed wTotalLength calculation to simplify descriptor offsets
- Re-ordered endpoints and interfaces
- Added more debug output
- Added usbInitTime to show how long keyboard initialization took
(Useful when debugging bad init sequences)
- Added function for usb_resume() which takes care of the resume sequence
* Resume is now only called if packets are starting to timeout
USB - Special Options
- Added enableDeviceRestartOnUSBTimeout
* A last resort hammer for bad USB Chipsets/OSs, don't use if you can help it
* Disabled
- Added enableUSBResume
* Enables host resume wake-up signalling, required to wake a computer from sleep
* Enabled
- Added enableUSBLowPowerNegotiation
* Enables power negotiation hack
* Required to use firmware with an IPad and other hard-limit low-power USB hosts
* Hasn't been tested with the recent changes
* Disabled
- Added enableUSBSuspend
* Enables power down events on host USB bus suspend
* Enabled
USB - Keyboard
- Attempted to cleanup HID SET_REPORT
* Works much better
* Still has an issue under Linux which generates *a lot* of NAKs (initializes quickly regardless)
+ Not present on other keyboards
+ SETUP -> OUT -> IN : This sequence is the problem
+ Specifically during the OUT phase
- Enabled
USB - CDC Virtual Serial Port
- Code cleanup
- Added convenience struct USBCDCLineCoding for easier debugging
- Attempted to cleanup CDC_SET_LING_CODING
* Works much better
* Still has an issue under Linux which generates *a lot* of NAKs (initializes quickly regardless)
+ SETUP -> OUT -> IN : This sequence is the problem
+ Specifically during the OUT phase
+ Likely the same issues as HID SET_REPORT
- Enabled
USB - Mouse
- Enabled
USB - Joystick
- Disabled
USB - RawIO
- Initial code, API not used yet
- Disabled
DFU
- Updated load script, now faster
- Somewhat aggresive, may cause restarts if the keyboard/OS hasn't fully intialized the keyboard
- Added GET_IDLE handling and correct usage of SET_IDLE
- Initial implementation of idle send, commented out as it causes issues on Mac OSX for sleeping
(keyboard has been working without it)
- MacOSX seems to have some sort of data corruption on the USB link, not sure why (other OSs have no issues)
- Cleaned up some code
- Added a longer sleep after the resume sequence to prevent possible issues sending keys too soon
(may need to be increased more)
Ipad support now seems flaky, though Mac, Windows seems solid.
Init sequence on Linux seems slow, even though there are no errors.
- Maintains a sleep state variable (may be able to use as a signal in the future)
- On the first keypress, hold the RESUME signal for 10 ms (spec says between 1 and 15 ms)
- Removed some messages that were affecting sleep state
- Odd that this went unnoticed for sooooo long
- Fixes Mac OSX issue where keyboard freezes/hangs on a reboot
- Also adding missing code for usbMuxUart
- Still very basic (lots of room for improvement)
- Capability format will likely change at some point
- 16 bit movement control, however repeat rate limits usability (will need KLL 0.4 to make better)
- Full Mouse support will have to wait for KLL 0.6
* This will include dynamic HID descriptor generation for many wheels and axis depending on the KLL needs
- HID descriptor is currently limited to 8 buttons
- Technically mouse movement also works (tested by accident), but it's disable for now (needs some API thought)
- Adding additional udev rules
- Added KRO mode default define
- Each scan module now has a current change callback which passes the available current as a parameter
- No longer attempts to use the max 500 mA immediately, starts with 100 mA then goes to 500 mA after enumeration
- If enumeration fails due to bMaxPower of 500 mA, then attempt again at 100 mA (might also be possible to go even lower to 20 mA in certain cases)
- Now working with the Apple Ipad (no over-power messages)
- Fixed Wake-up behaviour on Apple Ipad (and likely other iOS devices)
- More effecient set_feature/clear_feature handling (device handler)
- Initial power handling via Interconnect (still needs work to get it more dynamic)
- Rx buffers weren't fast enough, had to use DMA :D
- Basic LCD remote capabilities are working, single node
- Multi-node broadcast seems to have a bug still
- DMA ring buffer allowed for significant processing simplification
* There is an overrun risk, but the buffer is large and generally there isn't too much data being sent (just very quickly)
- Split out LCD layer stack capability into itself and an "exact" version used for updating remote nodes
- Implementation was completely wrong (has never worked properly)
- Now working in both NKRO and 6KRO modes
- Added more debugging code if things go wrong again
- Adds proper flashMode support for all keyboards and microcontrollers (usb and serial)
- flashModeEnabled must be set to 1 otherwise it will only show an error
* This is on purpose (somewhat dangerous feature as it allows remote flashing)
- Capability cleanup
- Media keys tested working on Linux/Windows/Mac (use Consumer control)
- Fixed enumeration delays
- Fixed virtual serial port configuration issues
- Fixed GET_REPORT and SET_REPORT
- Added intial descriptors and endpoints for Mouse and Joystick devices
- Split out the consumer and system control endpoint
- Added more fault debugging messages
- Added interface names to endpoints (visible in Windows Device Manager)
- Added KLL define for keyboard locale
Use a consistent standard - Tabs in front for indenting, spaces after for anything else. This way everything stays nice and lined up while also letting users change there prefered indent level. Most of the new files from Haata where already in this format.
- CLEAR_FEATURE unset is currently causing the keyboarod to freeze
- Though not ideal, it seems harmless to ignore this USB Control Packet
- Tested working on a Mac Mini (Yosemite)
- Had to re-write descriptor to support Windows 8.1 while still working with Mac OSX
- NKRO keyboard is now more bandwidth intensive as all the keys must be updated at the same time
(due to 8.1 bug in the HID spec)
- OSX doesn't properly follow the spec when it comes to padding bits (when using bitfields)
- Everything seems to work when I don't use them
- NKRO tested working on OSX
