Building mpir svn trunk (which for the Sun's is the same as mpir-1.2.2)
./configure && make -j && make -j check
ld.so.1: t-modlinv: fatal: /usr/local/gcc-4.2.4-sun-linker/lib/libgcc_s.so.1:
wrong ELF class: ELFCLASS32
/bin/bash: line 1: 25188 Killed ${dir}$tst
FAIL: t-modlinv
ld.so.1: t-popc: fatal: /usr/local/gcc-4.2.4-sun-linker/lib/libgcc_s.so.1:
wrong ELF class: ELFCLASS32
/bin/bash: line 1: 25208 Killed ${dir}$tst
FAIL: t-popc
ld.so.1: t-parity: fatal: /usr/local/gcc-4.2.4-sun-linker/lib/libgcc_s.so.1:
wrong ELF class: ELFCLASS32
/bin/bash: line 1: 25228 Killed ${dir}$tst
FAIL: t-parity
ld.so.1: t-sub: fatal: /usr/local/gcc-4.2.4-sun-linker/lib/libgcc_s.so.1:
wrong ELF class: ELFCLASS32
/bin/bash: line 1: 25248 Killed ${dir}$tst
FAIL: t-sub
=============================================================
9 of 9 tests failed
Please report to http://groups.google.co.uk/group/mpir-devel/
=============================================================
running source doesn't help but setting
LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/sparc-solaris-toolchain/lib/sparcv9
LD=/usr/ccs/bin/sparcv9/ld
does the job.
What should pass , and what should not ?
Thanks
Jason
On Friday 31 July 2009 01:16:46 Jason Moxham wrote:
> On Wednesday 29 July 2009 01:28:08 David Kirkby wrote:
> > 2009/7/28 jason <ja...@njkfrudils.plus.com>:
> > > This issue of building mpir on freebsd64 has been fixed in release
> > > 1.2.2 , which should be on the mpir main site in an hour or two. As
> > > far as we are aware it should work on Suns , although on T2 , you
> > > can't use the default linker with c++ and Suns cc , or at least that
> > > used to be the case .
> >
> > Can you be more precise about the Sun issue? I found the mpir version
> > in Sage would not build on Solaris x86, but would on Solaris SPARC.
> >
> > Here are a few facts, which you might find helpful if you are not
> > familiar with Suns and Solaris.
> >
> > 0) Sun build machines based on both their own SPARC and Intel/AMD x86
> > processors. SPARC and x86 processors are totally incompatible at the
> > assembly code level, so if you write assembly code, you need to write
> > different versions.
> >
> > However, well written 'user-land' code writen in a high-level
> > language should only need to be recompiled to run on x86 or SPARC
> > hardware.
> >
> > 1) Sun provide a linker /opt/ccs/bin/ld which is installed with a full
> > Solaris install.
> >
> > 2) Sun provide an assember /usr/ccs/bin/as which is installed with a
> > full install of Solaris.
> >
> > 3) /opt/SUNWspro/bin/cc is a C compiler, not a C++ compiler.
> >
> > 4) /opt/SUNWspro/bin/CC is a C++ compiler.
> >
> > 5) Neither /opt/SUNWspro/bin/cc (C compiler) or /opt/SUNWspro/bin/CC
> > (C++ compiler) are installed by default. You have to install them.
> > Downloading SunStudio (free) is the way to do that.
> >
> > 6) Recent versions of Solaris provide /usr/sfw/bin/gcc (C compiler)
> > and /usr/sfw/bin/g++ (C++ compiler). These are installed by default.
> >
> > 7) gcc on Solaris can be configured to use either the Sun or GNU
> > linker and assembler. Different people configure GCC different ways,
> > so you should not assume anything about the linker or assembler unless
> > you test to see what gcc/g++ is using.
> >
> > 8) Sun do provide a GNU versions of a linker, as /usr/sfw/bin/gld
> > Note the 'g' in front of the normal name for a linker.
> >
> > 9) Sun provide a GNU version of an assembler as /usr/sfw/bin/gas -
> > again note the 'g'
> >
> > 10) In Solaris 10 update 7 (as used on 't2'), both /usr/sfw/bin/gcc
> > and /usr/sfw/.bin/g++ use the Sun linker and assembler and ignore the
> > GNU ones. But in general, you should not make assumptions about the
> > linker or assembler.
> >
> > 11) Whilst the GNU and Sun compilers, assemblers and linkers share
> > some options (compilers use -g for debugging), 95% of the options that
> > can be used on the GNU tools can not be used on the Sun tools, and
> > visa versa.
> >
> > Dave
>
> We test on T2 , mark , mark2 , gcc54
>
> I know T2 has changed since our last test (which was when we released
> mpir-1.2) , but our previous problems have been
>
> On T2
>
> LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/sparc-solaris-toolchain/lib/spa
>rcv9 LD=/usr/ccs/bin/sparcv9/ld
> CC=cc
> CXX=CC
> ./configure --enable-cxx
> /home/jasonmoxham/mpir/mpir/trunk/mpirxx.h", line 3080: Error: Ambiguous
> partial specialization for __gmp_expr<__mpf_struct[1],
> __gmp_binary_expr<long, __gmp_expr<__mpf_struct[1], __mpf_struct[1]>,
> __gmp_binary_divides>>, __gmp_expr and __gmp_expr.
> "/home/jasonmoxham/mpir/mpir/trunk/mpirxx.h", line 3080: Error: Ambiguous
> partial specialization for __gmp_expr<_
>
> on gcc54 a c++ error ?
>
> on mark ?
> cc: Warning: -xarch=v9 is deprecated, use -m64 to create 64-bit programs
> ld: warning: file ../../.libs/libmpir.so: linked to
> /tmp/mark2/.libs/libmpir.so: attempted multiple inclusion of file
>
> Now that T2 has been updated I will retest it , I use a script so I will
> try all possible combinations of linkers/compilers , as long as it takes <
> 1day If there is too many combinations then I will have to prune them ,
> clearly I will try a "all gnu" toolchain , "all sun" toochain , if there
> are any other common/usefull combinations then I can just add them to my
> script , and they will be always be tested from now on.
>
> Jason
>
>
>
--~--~---------~--~----~------------~-------~--~----~
To post to this group, send email to sage-devel@googlegroups.com
To unsubscribe from this group, send email to
sage-devel-unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URLs: http://www.sagemath.org
-~----------~----~----~----~------~----~------~--~---
