configure problem on Solaris/Intel

Torbjorn Granlund tg at swox.com
Mon Dec 22 21:35:17 CET 2008


"Steven Dirkse" <sdirkse at gams.com> writes:

  I am building GMP on a Solaris/Intel machine:
  
  -bash-3.2# uname -a
  SunOS sigvm 5.11 snv_95 i86pc i386 i86pc
  
  I could build GMP 4.2.1 very simply, using
  
  ./configure ABI=32
  where the output of config.guess was:
  
  bash-3.2# cd gmp-4.2.1
  -bash-3.2# ./config.guess
  pentium3-pc-solaris2.11
  
  But moving to 4.2.[2,3,4] I get different output from config.guess and
  the build fails with an assembler problem.  I can fix it by manually
  putting in the --build.
  -bash-3.2# cd ../gmp-4.2.4
  -bash-3.2# ./config.guess
  core2-pc-solaris2.11
  -bash-3.2# cd mpz/
  -bash-3.2# gcc -DHAVE_CONFIG_H -I. -I. -I.. -D__GMP_WITHIN_GMP -I..
  -m32 -O2 -fomit-frame-pointer -mtune=k8 -march=k8 -c cmp_d.c  -fPIC
  -DPIC -o .libs/cmp_d.o
  Assembler: cmp_d.c
          "/var/tmp//ccH2vmCb.s", line 59 : Illegal mnemonic
          Near line: "    ffreep  %st(0)"
          "/var/tmp//ccH2vmCb.s", line 59 : Syntax error
          Near line: "    ffreep  %st(0)"
          "/var/tmp//ccH2vmCb.s", line 61 : Illegal mnemonic
          Near line: "    ffreep  %st(0)"
          "/var/tmp//ccH2vmCb.s", line 61 : Syntax error
          Near line: "    ffreep  %st(0)"
          "/var/tmp//ccH2vmCb.s", line 125 : Illegal mnemonic
          Near line: "    ffreep  %st(0)"
          "/var/tmp//ccH2vmCb.s", line 125 : Syntax error
          Near line: "    ffreep  %st(0)"
  -bash-3.2# gcc -DHAVE_CONFIG_H -I. -I. -I.. -D__GMP_WITHIN_GMP -I..
  -m32 -O2 -fomit-frame-pointer -mtune=k6 -march=k6 -c cmp_d.c  -fPIC
  -DPIC -o .libs/cmp_d.o
  -bash-3.2#
  
  ./configure ABI=32 --build=k6-pc-solaris2.11    works, for example.
  
This is not a GMP problem, but a compiler/assembler problem.

When valid C code makes the compiler generate assembly that the assember
cannot handle, then things are quite broken.

See also http://gmplib.org/list-archives/gmp-bugs/2007-August/000816.html.

-- 
Torbjörn


More information about the gmp-bugs mailing list