Re: Patch adding POWER9 cross compile support

[Tobias Platen <trisquel@platen-software.de>]

I had to work on the core-updates branch, or in the past on the wip-gcc7 
branch. I mentioned this some month ago. The current branch has an 
outdated gcc version.

On 03/12/2019 03:10 PM, Ludovic Courtès wrote:
> Hi!
> 
> Efraim Flashner <efraim@flashner.co.il> skribis:
> 
>> On Sun, Mar 10, 2019 at 09:20:04PM +0100, Tobias Platen wrote:
>>> I ran configure on my Talos II, and got the following error message.
>>>
>>> checking for the Guix system type... powerpc64le-linux
> 
> Sure but that’s the next step.  My question was about the triplet you
> passed to ‘--target’ when cross-compiling from your Intel(?) machine,
> but I think I can derive the answer now:
> 
>    guix build --target=powerpc64le-linux-gnu …
> 
>>> Guix already knows about this architecture, but building glibc will fail if
>>> gcc does not have the float128 datatype. Once I saw this link[1] on the guix
>>> mailing list, I knew how to solve the build error.
>>>
>>> For the second question I could not find an answer.
>>>
>>> [1] http://lists.busybox.net/pipermail/buildroot/2017-September/201379.html
> 
> Alright.
> 
> Thanks for the info.
> 
>> In the off chance we ever wish to support powerpc64 big endian, I
>> suggest instead using (string-prefix? "powerpc64*-" target)
> 
> Yes.
> 
> Unfortunately, with just this extra --with-long-double-128, this:
> 
>    guix build sed --target=powerpc64le-linux-gnu
> 
> eventually leads to a glibc build failure:
> 
> --8<---------------cut here---------------start------------->8---
> running configure fragment for sysdeps/unix/sysv/linux/powerpc
> checking whether powerpc64le-linux-gnu-gcc -g -O2 -mlong-double-128 uses IBM extended format... yes
> running configure fragment for sysdeps/unix/sysv/linux
> checking installed Linux kernel header files... 3.2.0 or later
> configure: WARNING: minimum kernel version reset to 3.10.0
> checking for kernel header at least 3.10.0... ok
> running configure fragment for sysdeps/gnu
> running configure fragment for sysdeps/powerpc/powerpc64/le
> checking if powerpc64le-linux-gnu-gcc supports binary128 floating point type... no
> checking if the target machine is at least POWER8... yes
> configure: error: ***  binary128 floating point type (GCC >= 6.2) is required on powerpc64le.
> --8<---------------cut here---------------end--------------->8---
> 
> See ‘config.log’ excerpt below.  Do you happen to have an additional
> change needed?
> 
> Thanks,
> Ludo’.
> 
> --8<---------------cut here---------------start------------->8---
> configure:6698: result: running configure fragment for sysdeps/powerpc/powerpc64/le
> configure:5: checking if powerpc64le-linux-gnu-gcc supports binary128 floating point type
> configure:31: powerpc64le-linux-gnu-gcc -c -g -O2 -Werror -mfloat128  conftest.c >&5
> powerpc64le-linux-gnu-gcc: error: unrecognized command line option '-mfloat128'
> configure:31: $? = 1
> configure: failed program was:
> | /* confdefs.h */
> | #define PACKAGE_NAME "GNU C Library"
> | #define PACKAGE_TARNAME "glibc"
> | #define PACKAGE_VERSION "(see version.h)"
> | #define PACKAGE_STRING "GNU C Library (see version.h)"
> | #define PACKAGE_BUGREPORT "https://sourceware.org/bugzilla/"
> | #define PACKAGE_URL "http://www.gnu.org/software/glibc/"
> | #define PKGVERSION "(GNU libc) "
> | #define REPORT_BUGS_TO "<http://www.gnu.org/software/libc/bugs.html>"
> | #define HAVE_TUNABLES 1
> | #define HAVE_CC_NO_STACK_PROTECTOR 1
> | #define STACK_PROTECTOR_LEVEL 0
> | #define USE_MULTIARCH 1
> | #define HAVE_ASM_SET_DIRECTIVE 1
> | #define HAVE_SDATA_SECTION 1
> | #define NO_CTORS_DTORS_SECTIONS 1
> | #define HAVE_Z_COMBRELOC 1
> | #define HAVE_CC_INHIBIT_LOOP_TO_LIBCALL 1
> 
> | #define HAVE_EHDR_START 1
> | #define HAVE_BUILTIN_TRAP 1
> | #define HAVE_ELFV2_ABI 1
> | #define __LINUX_KERNEL_VERSION (3 * 65536 + 10 * 256 + 0)
> | #define __ABI_TAG_VERSION 2,6,32
> | #define HAVE_INLINED_SYSCALLS 1
> | /* end confdefs.h.  */
> |
> | __float128 a, b, c, d, e;
> | int i;
> |
> | __float128
> | foobar (__float128 x)
> | {
> |   a = __builtin_nansq ("0");
> |   b = __builtin_huge_valq ();
> |   c = __builtin_infq ();
> |   d = __builtin_fabsq (x);
> |   e = __builtin_nanq ("0");
> |   i = __builtin_signbit (x);
> |   return __builtin_copysignq (x, x);
> | }
> |
> configure:39: result: no
> configure:47: checking if the target machine is at least POWER8
> configure:61: powerpc64le-linux-gnu-gcc -c -g -O2   conftest.c >&5
> configure:61: $? = 0
> configure:68: result: yes
> configure:75: error: ***  binary128 floating point type (GCC >= 6.2) is required on powerpc64le.
> --8<---------------cut here---------------end--------------->8---
>