[RTOS] Added idle hook API

[neilt6]

Added a new API for attaching an idle hook to be executed by the idle task. This will allow developers to add custom logic to the idle task (e.g. to call sleep()) without modifying the library code.

Updated version of #1463.

[neilt6]

@adamgreen, I've created a new pull-request for my RTOS idle hook API, and incorporated your recommendations from #1463.

[adamgreen]

Why did you not just #include "rtos_idle.h" instead to get the function prototype for rtos_idle_loop() since that header contains the prototype as well? Or maybe consider removing the prototype for rtos_idle_loop() from rtos_idle.h so it isn't exposed from the more public header and then keep these forward declarations in the RTX_Conf* sources.

[neilt6]

I was mirroring the way mbed_die() is forward declared for the error functions, and forgot to remove the unnecessary declaration from rtos_idle.h. Fixed in c52e0dc.

[adamgreen]

👍

[0xc0170]

Looks good to me.

cc @sg-

[sg-]

👍

As a future improvement it'd be nice to have the same interface that supports https://github.com/mbedmicro/mbed/blob/master/libraries/mbed/api/FunctionPointer.h
so member functions are also compatible.

[0xc0170]

@neilt6 Referenced PR is having troubles, I'll investigate why CI did not catch this

[neilt6]

@0xc0170 That's interesting... What if we added a prototype for void rtos_idle_loop(void) to rtos_idle.h, and included that header in RTX_Conf_CM.c instead of using extern void rtos_idle_loop(void)?

Proposed rtos_idle.h:

#ifndef RTOS_IDLE_H
#define RTOS_IDLE_H

#include <stddef.h>

#ifdef __cplusplus
extern "C" {
#endif

void rtos_attach_idle_hook(void (*fptr)(void));

void rtos_idle_loop(void);

#ifdef __cplusplus
}
#endif

#endif

Proposed RTX_Conf_CM.c:

/*----------------------------------------------------------------------------
 *      OS Idle daemon
 *---------------------------------------------------------------------------*/
#include "rtos_idle.h"

void os_idle_demon (void) {
    /* The idle demon is a system thread, running when no other thread is      */
    /* ready to run.                                                           */
    rtos_idle_loop();
}

[adamgreen]

I suspect that this has something to do with a circular dependency. Previously the C++ code in rtos/rtos just depended on the rtos/rtx code but this change introduced a circular dependency where rtos/rtx now depends on rtos/rtos as well. Can the rtos_idle.* files not be placed in the RTX directory with the rest of the C code?

[neilt6]

Yes, they just need to be duplicated for the three "versions" of RTX.

[neilt6]

@0xc0170, were you expecting me to submit the patch for #1639? I can write it, but I would be doing so blindly since I'm unable to reproduce the phenomenon on the online compiler...

[0xc0170]

@0xc0170, were you expecting me to submit the patch for #1639? I can write it, but I would be doing so blindly since I'm unable to reproduce the phenomenon on the online compiler...

I haven't got time to reproduce it, I'll have a look and comment to your proposal