lib: smf: check for NULL

[maass-hamburg]

don't execute share_paren() if the target is NULL
in smf_execute_ancestor_exit_actions()

Fixes: #71215

[glenn-andrews]

I'm a bit confused over what situation smf_set_state(ctx, NULL) is a valid function call. Setting state to NULL should be an error, if not a crash.

I feel that exiting in smf_set_state() if new_state == NULL would be a better solution.

[maass-hamburg]

I'm a bit confused over what situation smf_set_state(ctx, NULL) is a valid function call. Setting state to NULL should be an error, if not a crash.

I feel that exiting in smf_set_state() if new_state == NULL would be a better solution.

@glenn-andrews in the docs and header it says explicitly that it is a valid function call:

/**
 * @brief Changes a state machines state. This handles exiting the previous
 *        state and entering the target state. A common parent state will not
 *        exited nor be re-entered.
 *
 * @param ctx       State machine context
 * @param new_state State to transition to (NULL is valid and exits all states)
 */
void smf_set_state(struct smf_ctx *ctx, const struct smf_state *new_state);

[glenn-andrews]

I'm a bit confused over what situation smf_set_state(ctx, NULL) is a valid function call. Setting state to NULL should be an error, if not a crash.
I feel that exiting in smf_set_state() if new_state == NULL would be a better solution.

@glenn-andrews in the docs and header it says explicitly that it is a valid function call:

/**
 * @brief Changes a state machines state. This handles exiting the previous
 *        state and entering the target state. A common parent state will not
 *        exited nor be re-entered.
 *
 * @param ctx       State machine context
 * @param new_state State to transition to (NULL is valid and exits all states)
 */
void smf_set_state(struct smf_ctx *ctx, const struct smf_state *new_state);

My mistake. I apologize.

I'm even more confused about the meaning of 'exits all states'. Does that mean the HSM terminates, or that it calls the exit action, or what?

[maass-hamburg]

I'm a bit confused over what situation smf_set_state(ctx, NULL) is a valid function call. Setting state to NULL should be an error, if not a crash.
I feel that exiting in smf_set_state() if new_state == NULL would be a better solution.

@glenn-andrews in the docs and header it says explicitly that it is a valid function call:

/**
 * @brief Changes a state machines state. This handles exiting the previous
 *        state and entering the target state. A common parent state will not
 *        exited nor be re-entered.
 *
 * @param ctx       State machine context
 * @param new_state State to transition to (NULL is valid and exits all states)
 */
void smf_set_state(struct smf_ctx *ctx, const struct smf_state *new_state);

My mistake. I apologize.

I'm even more confused about the meaning of 'exits all states'. Does that mean the HSM terminates, or that it calls the exit action, or what?

It calls the exit function the current state and if CONFIG_SMF_ANCESTOR_SUPPORT is enabled also the exit actions of the parents of the current state. But it will not terminate the HSM on its own. I recommend to have smf_set_terminate() be executed in one of the exit action. Preferably at the exit action of the most parent state, because all remaining exit actions are not executed after smf_set_terminate()

[glenn-andrews]

My mistake. I apologize.
I'm even more confused about the meaning of 'exits all states'. Does that mean the HSM terminates, or that it calls the exit action, or what?

It calls the exit function the current state and if CONFIG_SMF_ANCESTOR_SUPPORT is enabled also the exit actions of the parents of the current state. But it will not terminate the HSM on its own. I recommend to have smf_set_terminate() be executed in one of the exit action. Preferably at the exit action of the most parent state, because all remaining exit actions are not executed after smf_set_terminate()

Oh. I think I see. You're trying to force the existing state machine to act like a terminate event calls all exit actions.

Would the correct way to do that is to have a STATE_TERMINATED outside all other states, and transition to that and call `smf_set_terminated() in the entry action for that state?