RTEMS 4.12 gcc-7.1.0/binutils-2.28 test results

[Joel Sherrill]

On Wednesday, May 31, 2017, Chris Johns <chrisj at rtems.org> wrote:

> Hi,
>
> I am testing RSB patches to move 4.12 to gcc 7.1.0/binutils-2.28.
>
> The SH arch fails to build on FreeBSD 11.0. The issue is
> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80923.
>
> RTEMS BSP Builder results summary for:
>
> $ rtems-bsp-builder --rtems-tools=/build/rtems/tools/4.12-7.1
> --rtems=/opt/work/chris/rtems/kernel/rtems.git
> --log=everything-standard-7.1
> --profile=everything --build=standard --jobs=7/6
>
> Total: Warnings:8682  exes:1996  objs:249031  libs:11855
> Failures:
>    1 standard powerpc/mcp750 build:
>       configure: /opt/work/chris/rtems/kernel/rtems.git/configure
> --target\
>       =powerpc-rtems4.12 --enable-rtemsbsp=mcp750 --prefix=/opt/rtems/4.12
>
>    2 standard powerpc/mtx603e build:
>       configure: /opt/work/chris/rtems/kernel/rtems.git/configure
> --target\
>       =powerpc-rtems4.12 --enable-rtemsbsp=mtx603e\
>       --prefix=/opt/rtems/4.12
>
>    3 standard powerpc/mvme2100 build:
>       configure: /opt/work/chris/rtems/kernel/rtems.git/configure
> --target\
>       =powerpc-rtems4.12 --enable-rtemsbsp=mvme2100\
>       --prefix=/opt/rtems/4.12
>
>    4 standard powerpc/mvme2307 build:
>       configure: /opt/work/chris/rtems/kernel/rtems.git/configure
> --target\
>       =powerpc-rtems4.12 --enable-rtemsbsp=mvme2307\
>       --prefix=/opt/rtems/4.12
>
>    5 standard powerpc/qemuprep-altivec build:
>       configure: /opt/work/chris/rtems/kernel/rtems.git/configure
> --target\
>       =powerpc-rtems4.12 --enable-rtemsbsp=qemuprep-altivec\
>       --prefix=/opt/rtems/4.12
>
>    6 standard powerpc/qemuprep build:
>       configure: /opt/work/chris/rtems/kernel/rtems.git/configure
> --target\
>       =powerpc-rtems4.12 --enable-rtemsbsp=qemuprep\
>       --prefix=/opt/rtems/4.12
>
>    7 standard sh/gensh1 build:
>       configure: /opt/work/chris/rtems/kernel/rtems.git/configure
> --target\
>       =sh-rtems4.12 --enable-rtemsbsp=gensh1 --prefix=/opt/rtems/4.12
>
>    8 standard sh/gensh2 build:
>       configure: /opt/work/chris/rtems/kernel/rtems.git/configure
> --target\
>       =sh-rtems4.12 --enable-rtemsbsp=gensh2 --prefix=/opt/rtems/4.12
>
>    9 standard sh/gensh4 build:
>       configure: /opt/work/chris/rtems/kernel/rtems.git/configure
> --target\
>       =sh-rtems4.12 --enable-rtemsbsp=gensh4 --prefix=/opt/rtems/4.12
>
>   10 standard sh/simsh1 build:
>       configure: /opt/work/chris/rtems/kernel/rtems.git/configure
> --target\
>       =sh-rtems4.12 --enable-rtemsbsp=simsh1 --prefix=/opt/rtems/4.12
>
>   11 standard sh/simsh2 build:
>       configure: /opt/work/chris/rtems/kernel/rtems.git/configure
> --target\
>       =sh-rtems4.12 --enable-rtemsbsp=simsh2 --prefix=/opt/rtems/4.12
>
>   12 standard sh/simsh2e build:
>       configure: /opt/work/chris/rtems/kernel/rtems.git/configure
> --target\
>       =sh-rtems4.12 --enable-rtemsbsp=simsh2e --prefix=/opt/rtems/4.12
>
>   13 standard sh/simsh4 build:
>       configure: /opt/work/chris/rtems/kernel/rtems.git/configure
> --target\
>       =sh-rtems4.12 --enable-rtemsbsp=simsh4 --prefix=/opt/rtems/4.12
>
> Average BSP Build Time: 0:00:17.431351
> Total Time 0:49:23.329735
> Passes: 157   Failures: 13
>
> The SH failures are expected because there are no tools. I have reached
> out to
> find a solution here.
>
> The PowerPC failures are due to using binutils-2.28 and linking the
> bootloader.
> It is broken. The linker is generating lots of messages like:
>
> /opt/work/chris/rtems/kernel/rtems.git/c/src/lib/libbsp/
> powerpc/motorola_powerpc/bootloader/../../../powerpc/
> shared/bootloader/head.S:140:(.text+0x8a):
> dangerous relocation: generic linker can't handle R_PPC_SECTOFF_HA
>
> This bug exposes a PowerPC BSP build system error where running make again
> does
> not generate the error. It seems the RALF file rules are broken.
>
> I am happy to push the patches to update to gcc-7.1.0 and binutils-2.28
> and keep
> SH on the current tools and the PowerPC on binutils-2.27.


This sounds reasonable and unfortunately it is common to have a couple of
architectures which have issues and have to be held back for a while.

Did you ask about the binutils issue? It appears to be motorola_powerpc
BSPs which turn the executable into a compressed binary image and link it
to the bootloader. Maybe the objcopy messes up the section type.  One
solution if this isn't fixable is to use the RTEMS utility which converts a
file into a C array which can then just be compiled so it gets the right
sections, CPU flags, etc. Same result in the end but a bit more complicated
for the build system.

>
> I will run a build of everything with tests.
>
> Chris
> _______________________________________________
> devel mailing list
> devel at rtems.org <javascript:;>
> http://lists.rtems.org/mailman/listinfo/devel
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rtems.org/pipermail/devel/attachments/20170531/b8f0ebd2/attachment.html>