[PATCH v2 00/10] Rework work area initialization

[Chris Johns]

On 3/2/20 4:27 pm, Sebastian Huber wrote:
> ----- Am 3. Feb 2020 um 0:01 schrieb Chris Johns chrisj at rtems.org:
> 
>> On 3/2/20 12:28 am, Sebastian Huber wrote:
>>> This change set reworks the work area initialization carried out by the BSPs to
>>> initialize the workspace and C program heap only on demand, e.g. in case these
>>> components are used by the application.
>>>
>>> Currently, the workspace is always used, however, this may change in a follow up
>>> change set.
>>
>> Is the purpose of change to have a more efficient use of the workspace because a
>> defined area is reserved and allocated when the manager is initialised?
> 
> No, this patch set changes the way how BSPs provide the memory area used for the workspace and the C program heap. It doesn't change how the workspace itself is used.

OK and thanks.

>> I am not sure what the "demand" is and where it is coming from.
> 
> Currently, the workspace and C program heap area always initialized.  With this patch, they are only initialized if they are used. So, a future application which only uses statically allocated resources, will not have the heap handler in its executable.

Is the reason why and approach for making a fully static application documented
anywhere? How will a user know this exists?

I assume this helps low memory targets?

>> How are allocation errors handled? Traditionally this has been at compile time
>> but I am not sure if this is still valid.
> 
> Workspace allocation errors were never handled at compile time, they result in fatal run time errors or failed directives. This patch set changes nothing in this respect.

Sure.

Chris