[PATCH 7/7] rtems: Add rtems_task_build()

[Chris Johns]

On 31/8/20 4:43 pm, Sebastian Huber wrote:
> On 31/08/2020 02:34, Chris Johns wrote:
> 
>> On 31/8/20 12:49 am, Sebastian Huber wrote:
>>> On 22/08/2020 09:49, Chris Johns wrote:
>>>
>>>> On 21/8/20 9:51 pm, Sebastian Huber wrote:
>>>>> In contrast to rtems_task_create() this function creates a task with a
>>>>> user-provided task storage area.
>>>> The name is build but it creates a task? I am wondering about
>>>> rtems_task_create_static or something along this line?
>>> A function to do a static initialization is a contradiction from my point of
>>> view. Static initialization means for me that you statically initialize a data
>>> structure and then it is ready to use (it may involve a static constructor).
>>>
>>> The function builds a task from user-provided (stack, attributes, etc.) and
>>> system-provided (thread control block) components.
>> Build and create are both verbs which means both contradict the idea of
>> something being static. By tradition we assume a function's naming is in the
>> run-time context and we do not consider the fact a compiler may optimise the
>> operation and prepare the result before the code runs.
> Yes, the task creation or building through a function call is not a static
> initialization. This is why I don't like rtems_task_create_static().
>>
>> I am concerned there maybe doubt about how the calls are to be used if you are
>> not familiar with the API and it's history. Do I need to create a task then
>> build it before I start it?
> Yes, such a confusion is possible, but I think this can be solved by the
> documentation. 

Yes it can but why not take a moment to consider what we could use. :)

> Also both functions return an identifier. You cannot use them
> with an identifier.

That implies some knowledge of the API. For us this is apparent.

>> The call names I proposed both create a task, one is
>> static and by default the other is not.
> 
> If you really want something with create in it, then I suggest
> rtems_task_create_with_config() and rtems_message_queue_create_with_config(). I
> think these names are a bit long.

This is hard. "With" implies having a config, ie created with it attached to the
task. What about rtems_task_create_by_config() as a shorten version of
rtems_task_create_by_means_of_a_config() or even rtems_task_create_using_config()?

Chris