* fix(yaml): use find_clang as for datacopy
* fix: use library path provided by clang python package first
* fix: anonymous structs and unions
* fix: typedefd struct attributes
* chores(flysky): separate AFHDS2A and AFHDS3 into 2 different module type
This should allow to remove some of the hacks added earlier.
* chore: Add some EL18 CI bits
* fix(afhds): cmake defs
* fix(ci): add el18 to commit-tests.sh
* chore(cpn): separate AFHDS2A and AFHDS3
* chore(cpn): remove binary import of afhds and legacy fxFreq field for afhds3
* fix(cpn): limit available internal afhds protocols to those supported by hardware
* chore(cpn): update Flysky int module test
* fix(cpn): compile error
* fix(cpn): afhds2a max channels
* fix: TR_MODULE_PROTOCOLS
* enh(cpn): more support for afhds2a and 3
* fix(cpn): rebase compile error
* fix(cpn): bit indexes
* fix(cpn): emi toString
* fix(cpn): limit available internal afhds protocols to those supported by hardware
* enh(cpn): add support for Flysky EL18
* fix(cpn): available protocols
* fix(cpn): rebase moduledata
* chore(cpn): add module fields to model print
* Separate AFHDS2A and AFHDS3 UI settings.
* Fix AFHDS3 conversion from old subtype value.
Fix 'Channel Range' layout on portrait LCD.
* Fix name.
---------
Co-authored-by: Peter Feerick <peter.feerick@gmail.com>
Co-authored-by: elecpower <neilh713@tpg.com.au>
Co-authored-by: Phil Mitchell <phil.a.mitchell@gmail.com>
PR #3451 moved the start of the mixer task to a point in time where the task scheduler is already started (UI task). Before that change, the mixer task would be created before and effectively started after the task scheduler started.
This introduces a race condition on `mixerTaskId.rtos_handle`, as the mixer scheduler might trigger before `xTaskCreateStatic()` returns and the task handle is assigned.
This PR fixes the issue by delaying the start of the mixer scheduler until `xTaskCreateStatic()` returned and the task handler has been properly assigned.
* fix: use keysGetMaxTrims() instead of MAX_TRIMS
* feat: add support for virtual trims
This is necessary for radios like X-lite where half of the trims are simulated
* fix(flysky gimbals): mask first 4 channels when FS gimbals are detected.
* re-adjust center value to 4096 (2^12).This prevents values higher than necessary being displayed.
* fix: pulse driver with 32bit timers
* fix(mpm): clear RX buffer before signature
* fix: add some debouncing on soft-serial start bit detection
This helps coping with spurious impulses on S.PORT created by payload sent over PPM pin (X10E).
* fix: whenever the RXNE bit is set, read the register
... regardless of errors. That allows for reducing the effect of detected errors.
* fix(imu): reset error counter on successful IMU query
Please note that the HAL already handles busy I2C bus, so that we should not run into the case where the bus is used and another query is started. However, if that happens, HAL_BUSY will be returned, which leads to -1 being returned in our functions, thus increasing the error count until the sensor stops working.
* fix(imu): move imu init into mixer task
Fixes a race condition between init and use.
* fix: sort mixer lines after loading model
Fixes#2806
* fix(cpn): sort mixer lines after loading model
* More efficient swap
---------
Co-authored-by: elecpower <neilh713@tpg.com.au>
* chore(lua): revert useless c light function change
This change was part of the eLua LTR patches (based on 5.1), whereby this feature was already integrated in Lua 5.2.
* feat(lua): port of NodeMCU Lua53 read-only tables
# Conflicts:
# radio/src/lua/CMakeLists.txt
# radio/src/lua/api_general.cpp
* fix(nv14): missing SDRAM definition
* fix: l_mathop
* fix: use float functions in drawPower()
* fix: ltable resize
* fix: correct name is "bit32"
* fix: first Lua context must be closed last
As the lookup cache is attached to that one
* Revert to Lua using double
* Ensure byte-code compatibility
* Fix possible compatibility issue with `lua_typename`/`luaL_typename`
* Fix Bitmap symbol table
* Re-add table lib for COLORLCD
* WiP: new module interface
* WiP: softserial timer DMA
* WiP: setup() and send() in one step
and remove methods specific to internal or external module
* fixes and extra port config
* horus targets
* module ports have moved
* T8 support (only internal module)
* fix nv14
* stupid GCC....
* disable BT
* remove nv14 telem driver
* remove old aux serial driver
* Move module buffers out of pulses.h
* Soft serial fixes and PXX1 support
* Fixed PPM module output
* Fixed CLI
* compilation fixes
* Ported bluetooth and flysky gimbal to new serial driver
* Re-enable soft-serial RX
* fixes and cleanup
* nv14 fixes
* Fix flysky gimbals
* Ghost module
* 16 bytes of RAM
* Remove getByte from module API
* Ghost cleanup
* fix: _half_duplex_input() must be called after usart init
* WiP
* more cleanup
* Remove board specific settings from PXX1 / PXX2 drivers
* A bit of PXX2 header cleanup
* Add support for OVER8 and ONEBIT
* Add some support for inverters (NV14 / EL18)
* Support NV14 power fixes in generic board
* Fix polarity handling
* Fix RX DMA
* Fix mixer not running
* Add soft-serial polarity control
* Quick fix for telemetry polling while module port is changed
* Fix bootloader DEBUG
* Fix MPM telemetry
* Fix pulse start
* DSM2 soft-serial fallback
* DSMP async module restart
* Fix CRSF LUA upstream
* Remove global variable telemetryProtocol
* Use HAL/LL constants
* Fix various issue with AFHDS3 and NV14/EL18 internal module
* Fixed S.PORT unit tests
* fix: update widx after writing at current index
* fix: init error condition
* Fix: possible infinite loop in _find_port()
* Add support for bidir S.PORT init
* Add support for "getLastByte" in HW serial driver
* PXX1: Add support for S.PORT based on IDLE callback
* Fix rebase error
* Fix softserial _rx_fifo_push()
* Add softserial support for waitForTxCompleted()
* Fix MPM flashing (internal / external)
* Fixed external MPM
* Activate RX DMA on internal module
Seems to work with serial passthrough without issues, at least on the L3R PRO.
* Fix softserial DMA on stdlcd F4 targets
* Add softserial support for 32bit timers (mostly TIM2)
* More precise soft-serial
* Fix MPM update using inverters & HW uart
* Raise S.PORT DMA/USART IRQ prio to achieve faster switching to input
* Remove useless delays
* Fix flashing external ELRS module
* Added a retry mechanism for FrSky S.PORT update
* Remove useless enableRx()
* External FrSky modules are all flashed via S.PORT
* Remove power management unit update
This was only used on the first batch of X-lite.
* Enable RX DMA on external module USART
* feat(ACCESS): configure AUX serial port as "External module" (#2562)
This allows configuring the external access mod on runtime rather than by compilation option.
* fix: fetch external module driver
* CMake cleanup and simplifications
* Position Independent Code is need for radiolib_native
* Some more cleanup & repair
* Fix SIMU_SRC scope issue
* fix: avoid warnings due to assembler statements from StdPeriph on arm64 native
* Lua array indexes start at 1
* LUA model.setFlightMode: use array index to compute trim index
This allows for providing an incomplete table like this:
```
-- Disable T5 & T6 on FM0
modes = {}
modes[5] = 31
modes[6] = 31
t = {}
t["trimsModes"] = modes
model.setFlightMode(0, t)
```
* Remove warning & cleanup
* Fix trim value constraints in model.setFlightMode()
Prior to this fix, the confirmation dialog would crash because of an issue in libopenui and even then would not be displayed for the lack of the proper parent window.
Fixes#2741
Suppress `EVT_KEY_BREAK(KEY_PGDN)` after `EVT_KEY_LONG(KEY_PGDN)` on targets which don't have a physical `KEY_PGUP`.
This way, the sequence is now clean: no more `EVT_VIRTUAL_NEXT_PAGE` emitted after `EVT_VIRTUAL_PREV_PAGE` on these targets.
