[U-Boot] NDS32 toolchain?

[Tom Rini]

On Tue, Apr 24, 2018 at 03:10:02AM +0000, rick at andestech.com wrote:
> 
> 
> > -----Original Message-----
> > From: Tom Rini [mailto:trini at konsulko.com]
> > Sent: Tuesday, April 24, 2018 4:17 AM
> > To: u-boot at lists.denx.de; Rick Jian-Zhi Chen(陳建志)
> > Subject: NDS32 toolchain?
> >
> > Hey,
> >
> > So I'm looking over my setups again at what we do and do not have build testing
> > and run-time testing, and I see that for nds32 I just cannot find a good toolchain.
> > If I try for example
> > https://mirrors.edge.kernel.org/pub/tools/crosstool/files/bin/x86_64/7.3.0/x86_
> > 64-gcc-7.3.0-nolibc-nds32-elf.tar.xz
> > it first fails on the -G0 flag that we have set in and then it complains about a lack
> > of position independent code support.  Is the kernel.org toolchain perhaps
> > configured incorrectly?
> >
> 
> Hi Tom
> 
> You may try the following paths:
> 
> 1.
> github version
> https://github.com/andestech/build_script.git
> https://github.com/greentime/prebuilt-nds32-toolchain/releases

This binary gets a lot farther:
  CC      drivers/mtd/cfi_flash.o
drivers/mtd/cfi_flash.c: In function 'write_buff':
drivers/mtd/cfi_flash.c:1435:1: internal compiler error: in change_address_1, at emit-rtl.c:2126
 }
 ^
0x81fecd6 change_address_1
        /NOBACKUP/sqa2/greentime/contrib/travis/build_script/src_pkg/gcc-6.3.0/gcc/emit-rtl.c:2126
0x81ff26d adjust_address_1(rtx_def*, machine_mode, long long, int, int, int, long long)
        /NOBACKUP/sqa2/greentime/contrib/travis/build_script/src_pkg/gcc-6.3.0/gcc/emit-rtl.c:2264
0x8762488 nds32_spilt_doubleword(rtx_def**, bool)
        /NOBACKUP/sqa2/greentime/contrib/travis/build_script/src_pkg/gcc-6.3.0/gcc/config/nds32/nds32-md-auxiliary.c:3238
0x886be2f gen_split_7(rtx_insn*, rtx_def**)
        /NOBACKUP/sqa2/greentime/contrib/travis/build_script/src_pkg/gcc-6.3.0/gcc/config/nds32/nds32-doubleword.md:209
0x81fdf74 try_split(rtx_def*, rtx_insn*, int)
        /NOBACKUP/sqa2/greentime/contrib/travis/build_script/src_pkg/gcc-6.3.0/gcc/emit-rtl.c:3658
0x8450a37 split_insn
        /NOBACKUP/sqa2/greentime/contrib/travis/build_script/src_pkg/gcc-6.3.0/gcc/recog.c:2865
0x8450d0e split_all_insns()
        /NOBACKUP/sqa2/greentime/contrib/travis/build_script/src_pkg/gcc-6.3.0/gcc/recog.c:2955
0x8450d6a rest_of_handle_split_before_sched2
        /NOBACKUP/sqa2/greentime/contrib/travis/build_script/src_pkg/gcc-6.3.0/gcc/recog.c:3995
0x8450d6a execute
        /NOBACKUP/sqa2/greentime/contrib/travis/build_script/src_pkg/gcc-6.3.0/gcc/recog.c:4034
Please submit a full bug report,
with preprocessed source if appropriate.
Please include the complete backtrace with any bug report.
See <http://gcc.gnu.org/bugs.html> for instructions.
/home/trini/work/u-boot/u-boot/scripts/Makefile.build:280: recipe for target 'drivers/mtd/cfi_flash.o' failed
make[2]: *** [drivers/mtd/cfi_flash.o] Error 1
/home/trini/work/u-boot/u-boot/Makefile:1355: recipe for target 'drivers/mtd' failed
make[1]: *** [drivers/mtd] Error 2
make[1]: Leaving directory '/tmp/adp-ag101p'
Makefile:150: recipe for target 'sub-make' failed
make: *** [sub-make] Error 2

> 2.
> Arnd Bergmann built
> https://lkml.org/lkml/2018/4/17/395
> www.kernel.org/pub/tools/crosstool/files/bin/x86_64/6.4.0/x86_64-gcc-6.4.0-nolibc-nds32le-linux.tar.xz

Oddly, his 6.4.0 gets farther along than his 7.3.0 ones, but still
doesn't work as I get things like this:
  AS      arch/nds32/cpu/n1213/start.o
arch/nds32/cpu/n1213/start.S: Assembler messages:
arch/nds32/cpu/n1213/start.S:256:
Error: Incorrect syntax, sethi $ta,hi20(_start at GOTOFF).
arch/nds32/cpu/n1213/start.S:256:
Error: Incorrect syntax, ori $ta,$ta,lo12(_start at GOTOFF).
arch/nds32/cpu/n1213/start.S:256:
Warning: relax hint unrecognized instruction: line 254.

and it just keeps going from there.  You may want to drop him an email
about how he's configuring for nds32le.  Last time I talked with him, he
wasn't planning to re-spin the 7.3.x ones, but he might do gcc-8 once
it's ready.

> > Related, is there a QEMU target for nds32 that we could leverage so that once
> > the toolchain issue is resolved we can update .travis.yml to run tests on it?
> > Thanks!
> 
> I am applying the QEMU offering permit.
> If it is ok. I will prepare it for you.

Thanks!

-- 
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: not available
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20180424/a994e226/attachment.sig>