Build problems with source-builder compiling 4.11-arm with ADA

[Joel Sherrill]

On February 8, 2015 4:07:27 PM CST, Chris Johns <chrisj at rtems.org> wrote:
>On 9/02/2015 8:46 am, Joel Sherrill wrote:
>> On February 8, 2015 3:20:16 PM CST, Chris Johns <chrisj at rtems.org>
>wrote:
>>> On 8/02/2015 7:46 pm, Jan Sommer wrote:
>>>
>>>>
>>>> According to the log file it seems that the script is now compiling
>>> with
>>>> multilib support, but in the end it fails with the same error.
>>>> Attached is the rsb-report I get (now with gcc-4.9.2).
>>>>
>>>
>>> I do not follow what is happening because there is no
>--enable-multilib
>>>
>>> support in the RTEMS kernel configuration script and I cannot see
>how
>>> you enabled a kernel build because there is no --with-rtems on the
>>> command line. Note --enable-multilib is broken and should not be
>used.
>>
>> Actually --enable-multilib is in RTEMS and used to be used to ensure
>cpukit rules were followed.
>>
>
>It is by default disabled and I cannot see where it is being enabled. 
>There is no multilib in the RSB configuration files.

My old shell scripts used to enable it for testing RTEMS and as part of building Ada.

I doubt it had any other users. Its primary value was checking cleanliness of cpukit but with Arm having variants which don't build and no one wants to fix even as workaround, it is of diminishing utility.


>> Ada can be built though by installing the desired BSP and ensuring
>that it is used via CFLAGS_FOR_TARGET  passed to the GCC Ada build.
>
>Sure, so this means it currently cannot be built by the RSB without 
>someone adding support.

Agreed. Ada and Go require being built after RTEMS. I don't recall for gcj but the scripts know. :)

>Chris

--joel