Problems in boot_card() with configuration

[Joel Sherrill]

On 2/18/2013 1:04 PM, Joel Sherrill wrote:
> On 2/18/2013 12:05 PM, Matthew J Fletcher wrote:
>> On 15/02/13 14:24, Joel Sherrill wrote:
>>> Hmmm.. So you ended up with the default configuration info.. Just there to ensure things link..
>>>
>>> Folks.. Any ideas?
>> It might be an idea to just remove the magic that creates (in dummy.o)
>> the 'Configuration' and 'Configuration_RTEMS_API' structures. I guess
>> all its doing is making things link, but given that the structure is not
>> initialised in any way it just leads to odd crashes later on.
>>
>> I think it would be better to just allow the link to fail and its then
>> pretty obvious what the problem is.
> The dummy configuration and crt0.c in newlib are needed to make autoconf
> probes
> work.
>
> I think we could add a linker warning.
Following up on my own post. The default dummy.c COULD have a
moderately useful configuration like console and tick drivers, unlimited
resources, and unified memory. That would be enough to run many
simple programs.

It is also ONLY used with library feature probes when you include
the BSP arguments on the package you are building that is probing.
Otherwise, it is strictly a C Library issue (e.g. dummy crt0.c in newlib).

Chris and I discussed libbsp/shared/bspinit.c being used as the basis
for a new dummy configuration.
>> - Matthew
>


-- 
Joel Sherrill, Ph.D.             Director of Research & Development
joel.sherrill at OARcorp.com        On-Line Applications Research
Ask me about RTEMS: a free RTOS  Huntsville AL 35805
Support Available                (256) 722-9985