Ease integration of RTOS, we need cmsis-compatible exceptional handler naming or API to install exceptional handler in RAM

[xiongyu0523]

I'm porting Azure RTOS threadx to PICO and I can't submodule the threadx repo because I need to hook the pendsv and systick exceptional handlers.

In crt0.S, the SDK declare as isr_pendsv and isr_systick which are private and not conform to CMSIS standard, so the cm0 processor port of Azure RTOS need to be modified.

Or if you can provide a API similar to irq_set_exclusive_handler but work for all entries include 0-15 exceptional handlers. Then I can install by user code after main. This is also good.

Now, I can only modify the source of threadx to align with SDK naming, which is not flexible

[kilograham]

yes we plan to add ability to set exception handlers (an omission)

also we can make these public; we may not rename them, but at that point you can inject a header via cmake to rename them via #define.

[lurch]

Links to #21 and #127

[michael-ring]

Where can I put my vote that at least the cortex specific handlers get proper CMSIS names?

[kilograham]

Actually since these were private, we can likely rename them.

[michael-ring]

That would be really great! #21 has all necessary links to cmsis docs.

[xiongyu0523]

yes we plan to add ability to set exception handlers (an omission)

also we can make these public; we may not rename them, but at that point you can inject a header via cmake to rename them via #define.

So your suggestion is to add below global macro definition in CMakeLists.txt to replace the symbol, right?

add_definitions(-DPendSV_Handler=isr_pendsv)
add_definitions(-DSysTick_Handler=isr_systick)

[kilograham]

that was an option that would have workedd - albeit ugly-ly - but i've decided to add the real names now

[kilograham]

(setting variables yes, via global CMake add_defitions - which is eveil - no!)

[xiongyu0523]

at least it works and I can now submodule my threadx repo without any change. I left this ticket open and let you decide when a better solution is ready

[michael-ring]

Could this be a candidate for Release 1.2?
#127 seems to be so this one will propably make sense to have in, too