The way ChibiOS and our CMSIS/StdPeriphs handle platform flags clashes. This is the easiest workaround with the least changes.

…t 16000000 with the F446.

The RE1 sets the HSE clock to 16MHz. ChibiOS 17.x does some more rigorous checks on the mcuconf.h settings than old ChibiOS 2.6, and considers this invalid. We already shipped it like that, though. So I assume this will work anyway, so this check is neutered for our only F446 target.

Untested, because my RE1 is broken.

…d for new ChibiOS.

chMtxUnlock() can't be called without arguments anymore.

Memcore is now enabled, new ChibiOS makes this a dependency for Mempools.

…biOS

…use $STM32_TYPE

STM32_TYPE is being defined for all targets, but only the F4 makefiles reference it, the others specify the hosted value statically inside their makefile. Changed that for consistency.

Added STM32_SUBTYPE, to support the stm32_registry stuff in ChibiOS, to be able to define both STM32F405xx and STM32F40_41xxx cleanly together.

…luded makefile.

…dler.

Assuming we're going generally F405 all the way for the F4.

We don't clear the heap (yet?).

…tloader makefiles.

…dler.

… savings.

… space.

Storm32Bgc and VtolPathFollower got axed to make it fit the flash image.

…dler.

I'm not sure what specific MCU the pipXtreme uses to fix it properly, but STM32F10X_MD needs to be defined anyway for the older CMSIS, so there's a need to touch that header anyway.

…between cmsis_system.c and HAL.

Fixes CPU time accounting for modules in TaskInfo.RunningTime.

…cation.

ChibiOS 17 moved the thread_t structure from the top of the stack to the bottom. Need to know where stack starts for PIOS_Thread_Get_Stack_Usage. Fixes TaskInfo.StackRemaining.

This is fixed upstream apparently.