GCC Buildbot

[Gedare Bloom]

On Wed, Nov 27, 2013 at 9:27 AM, Ralf Corsepius
<ralf.corsepius at rtems.org> wrote:
> On 11/27/2013 02:56 PM, Joel Sherrill wrote:
>>
>> On 11/27/2013 5:10 AM, Ralf Corsepius wrote:
>>>
>>> On 11/27/2013 09:08 AM, Sebastian Huber wrote:
>>>>
>>>> The --enable-threads GCC configure option can be controlled via
>>>> "gcc/config.gcc":
>>>>
>>>> *-*-rtems*)
>>>>     case ${enable_threads} in
>>>>       yes) thread_file='rtems' ;;
>>>>     esac
>>>>
>>>> Does it make sense to build RTEMS without the RTEMS thread model
>>>> enabled?
>>
>> Yes. It makes sense to make it the default. I don't think this
>> matters much for C but it does for other languages including C++.
>>
> I could be wrong, but AFAIK, these days, threads are more or less mandatory
> in C++.
>
>>> IMO, it doesn't nor do I think it makes sense to build RTEMS with
>>> --disable-posix.
>>
>> For those writing C Classic API applications, there is no reason to
>> enable POSIX at this point.
>>
>> That said, the community that does that may find themselves having to
>> enable POSIX if they move to SMP or C++.
>>
>>> However, in the past, there had been strong resistance to make threading
>>> the default or mandatory. because of the overhead it causes in RTEMS.
>>
>> Where?
>
> Well, the origin of resistance sits in front of your desk ;)
>
The overhead of posix vs classic API would need to be examined
thoroughly before any move that mandates RTEMS must contain posix
support. With dynamic linking it may be possible (?) to remove the
overhead of these "optional" components, but we don't have that
feature quite yet.

>
> Ralf
>
> _______________________________________________
> rtems-devel mailing list
> rtems-devel at rtems.org
> http://www.rtems.org/mailman/listinfo/rtems-devel