gen405 status?

[Joel Sherrill]

Keith Robertson wrote:

> Chris Caudle wrote:
>
>> On Wednesday 01 February 2006 12:01, you wrote:
>>
>>> For C++ to work completely, the bsp_specs and linker
>>> script have to be just so.  C++ works on the powerpc
>>> but this BSP might have a tweak or two.
>>
>>
>> When the CVS sources are built on the test machine, are all options 
>> still turned on?  I found that --enabledocs also stops the build on 
>> the CVS head sources, looking for a VERSION variable which was not set.
>>
>> But back to the original question I had about --enable-cxx failing, I 
>> believe on closer examination that actually the build of the RTEMS 
>> executive completes, and the iostream sample application is failing.
>>
>> Here is my configure call:
>> ../rtems/configure \
>> --target=powerpc-rtems4.7 \
>> --disable-itron \
>> --enable-cxx \
>> --enable-rdbg \
>> --disable-docs \
>> --enable-networking \
>> --enable-rtemsbsp="gen405"
>>
>> Here are the tools I have installed:
>>
>>> rpm -qa | grep rtems | grep powerpc
>>
>> rtems-4.7-powerpc-rtems4.7-gdb-6.3-1
>> rtems-4.7-powerpc-rtems4.7-binutils-2.16.1-3
>> rtems-4.7-powerpc-rtems4.7-gcc-gcc4.0.1newlib1.13.0-4
>> rtems-4.7-powerpc-rtems4.7-c++-gcc4.0.1newlib1.13.0-4
>>
>> Here is the last part of the output of "make all"
>> powerpc-rtems4.7-g++ -B../../../../../gen405/lib/ -specs bsp_specs 
>> -qrtems -mcpu=403 -D_OLD_EXCEPTIONS -Dp
>> pc405 -O4 -fno-keep-inline-functions -g       -o iostream.exe init.o 
>> ../../../../../gen405/lib/no-dpmem.re
>> l ../../../../../gen405/lib/no-event.rel 
>> ../../../../../gen405/lib/no-msg.rel ../../../../../gen405/lib/no
>> -mp.rel ../../../../../gen405/lib/no-part.rel 
>> ../../../../../gen405/lib/no-signal.rel ../../../../../gen40
>> 5/lib/no-timer.rel ../../../../../gen405/lib/no-rtmon.rel
>> /opt/rtems-4.7/lib/gcc/powerpc-rtems4.7/4.0.1/../../../../powerpc-rtems4.7/bin/ld: 
>> error: no memory region
>>  specified for loadable section 
>> `.gnu.linkonce.sb._ZGVNSt7collateIcE2idE'
>> collect2: ld returned 1 exit status
>> gmake[5]: *** [iostream.exe] Error 1
>> gmake[5]: Leaving directory 
>> `/home/chris/Src/RTEMS/rtems-4.7-build-ppc/powerpc-rtems4.7/c/gen405/testsuite 
>>
>> s/samples/iostream'
>>
>
> Hi Chris.
>
> I've used the gen405 bsp extensively on 4.6.x.  However, I've not 
> migrated to 4.7 yet.  I've been planning to look at this, but haven't 
> yet had the time.
>
> At first glance, your error above is because your linkcmds script 
> doesn't include a directive for where to place .gnu.linkonce.sb.*  I 
> don't have access to my gen405 work currently, but I'd make the guess 
> that this section is introduced by the newer gcc tools and that the 
> existing gen405 linkcmds script doesn't account for it.
>
That's what it is all right.  I just fixed the linkcmds.

FWIW RTEMS testing builds all BSPs in a semi-random configuration.  That 
configuration stays set for a while
and then I change it.  No particular magic.  Chris was right, C++ was 
disabled in it.  We do build 64 configurations
of psim.  Honestly, I have been testing lots of gcc versions lately -- 
4.0 branch, 4.1 branch, and the head and most of the
BSP build time has been just to make sure the compilers aren't 
completely brain dead.  I am starting a build of all
BSPs with just Classic API and C++ to see what else is broken.

Someone mentioned an issue with a configuration with docs enabled.  Can 
you post the exact configure line?11

> If this is indeed the problem, you'll need to add this section to your 
> linkcmds script so that it gets included correctly in your binary. 
> Examining the linkcmds scripts for the other ppc based targets may 
> provide some help.
>
> Again, I've not tried gen405 on 4.7 yet myself, so I'm only speaking 
> off the top of my head.
>
> Hope this helps.
>
> Keith