[RTOS] Fixed osThreadGetState()

[neilt6]

Fixed regression that caused terminated threads to return an erroneous state value instead of "inactive".

[0xc0170]

is this correct? Why this function returns either osValue or state which is different declaration.

[neilt6]

Ideally, this function should only be returning the state definitions in rt_Task.h, and per the documentation in Thread.h, INACTIVE is the expected state for threads that have been terminated or otherwise don't exist. However, since there's no state definition for the ISR error I left that line alone.

[0xc0170]

Note - In the new upcoming cmsis 5 , they declare osThreadState osThreadGetState (osThreadId thread_id); . I could not locate this function in the current cmsis repository (the version we are using here), thus this was an extension we provided.

[0xc0170]

Fixed regression that caused terminated threads to return an erroneous state value instead of "inactive".

When was this regression introduced? Can you point to the commit?

[neilt6]

@0xc0170 The regression was introduced with 9a68561. I noticed it when one of my programs quit working because it could no longer determine when threads had been terminated.

[0xc0170]

I backported it from cortex-a implementation as I found it useful. That was before we exposed (rt_tid2ptcb). I propose to remote the check if (__get_IPSR() != 0U) return osErrorISR;, that does not make sense there (other thread functions within the file do not check it neither).
Or it might be cleaner just use tcb in the Thread.cpp as we used to and remove this extension from kernel (keep our changes to minimal).

Thanks @neilt6 for pushing this upstream

cc @TomoYamanaka

[sg-]

👍

[TomoYamanaka]

👍

[0xc0170]

I merged it, to fix the regression. We should look at it further, as my comment above indicates. I'll create an issue