gcc 4.9.2 build from source

[Michael E Goldsby]

Reporting as requested in output of make -k check.

Solaris 11.2-42 x86 running in Virtualbox on Ubuntu Linus 14.04 over Intel CoreDuo.

[Marc Glisse]

Hello,

it appears you are using an old version of GMP, please try a more recent
one. If you still get errors with the latest version, see:
https://gmplib.org/manual/Reporting-Bugs.html
so we can better help you.

--
Marc Glisse

[dclarke blastwave.org]

> On November 23, 2014 at 9:17 AM Marc Glisse <marc....@inria.fr>

> wrote:
>
>
> On Sun, 23 Nov 2014, Michael E Goldsby wrote:
>
> > Reporting as requested in output of make -k check.
> >
> > Solaris 11.2-42 x86 running in Virtualbox on Ubuntu Linus 14.04 over
> > Intel
> > CoreDuo.
>
> Hello,
>
> it appears you are using an old version of GMP, please try a more
> recent
> one.

The funny thing is that the folks in the GCC project and maillist point
to an older version of gmp as the version to use. See
ftp://gcc.gnu.org/pub/gcc/infrastructure/ where gmp 4.3.2 is what they
have there. If and when problems occur I have been pointed in that
direction over and over. I don't like it, but such is life with
bootstrapping gcc.

Dennis

[Torbjörn Granlund]

"dclarke blastwave.org" <dcl...@blastwave.org> writes:

The funny thing is that the folks in the GCC project and maillist point
to an older version of gmp as the version to use. See
ftp://gcc.gnu.org/pub/gcc/infrastructure/ where gmp 4.3.2 is what they
have there. If and when problems occur I have been pointed in that
direction over and over. I don't like it, but such is life with
bootstrapping gcc.

We're sadly aware of GCC's loud recommendation of severely obsolete
software.

The right approach for GCC is to use a modern GMP, compile it with
--disable-assembly, and link to it statically.


Torbjörn
Please encrypt, key id 0xC8601622

[dclarke blastwave.org]

> We're sadly aware of GCC's loud recommendation of severely obsolete
> software.
>
> The right approach for GCC is to use a modern GMP, compile it with
> --disable-assembly, and link to it statically.

I just extract the tarball into the gcc source tree and everything
builds fine without any issues. This rev seems to work well :

node000 $ grep "__GNU_MP_VERSION" /usr/local/include/gmp.h
#define __GNU_MP_VERSION 6
#define __GNU_MP_VERSION_MINOR 0
#define __GNU_MP_VERSION_PATCHLEVEL 0
#define __GNU_MP_RELEASE (__GNU_MP_VERSION * 10000 +
__GNU_MP_VERSION_MINOR * 100 + __GNU_MP_VERSION_PATCHLEVEL)


Not bad results at all :

https://gcc.gnu.org/ml/gcc-testresults/2014-10/msg01022.html


So I don't know what the issue is over at ye old GCC project but I have
seen no issues with gmp in a long while. However, just for the sake of
giving the optimization levels a try I want to see if gmp can build on
Solaris 10 with the latest release of Oracle Studio 12.4.

I'll let you know.

Dennis

[Cédric Bastoul]

Hi,

thanks for the report. For CLooG, the problem is diff on this system does not support the "--ignore-matching-lines" option. I got a report on this a while ago, I'll eventually fix it.

If you need to test it, the fix proposed by Marc Girod is to use in cloog/test/checker.sh, instead of --ignore-matching-lines=CLooG:

        $cloog $options -q $input > cloog_temp;
        diff -u -w cloog_temp $output | grep -v CLooG;

Best regards,

Cedric

On Sun, Nov 23, 2014 at 8:00 AM, Michael E Goldsby <michael...@gmail.com> wrote:

Reporting as requested in output of make -k check.

Solaris 11.2-42 x86 running in Virtualbox on Ubuntu Linus 14.04 over Intel CoreDuo.

--
--
You got this message because you subscribed to the CLooG Development mailing list.
To send messages to this list, use cloog-de...@googlegroups.com
To stop subscribing, send a mail to cloog-developm...@googlegroups.com
For more options and to visit the group, http://groups.google.fr/group/cloog-development?hl=en
---
You received this message because you are subscribed to the Google Groups "CLooG Development" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cloog-developm...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.