problem building gmp on Mac G5

[Torbjorn Granlund]

Kate Minola <kminola at po0.glue.umd.edu> writes:

  > We've tried to make GMP work with recent Apple compilers,
  > but all releases we've tried have had many bugs that cause
  > miscompilation of GMP.  
  
  I am using gcc-3.3.2 built from source, not an Apple
  supplied compiler.

Perhaps you've struck a bug in gcc-3.3.2.  Unfortunately, gcc 3.3
also has problems with the gmp sources, while is is not anywhere
close to the bug count of the Apple compiler.

The GCC team has decided to make sure to test future releases
with gmp as a test case, so this problem should disappear in the
future.
  
  I thought I should at least be able to build a generic C
  version of gmp by using the configure option
  --host=none-unknown-darwin.  I thought this only uses C
  code for gmp and no assembler code - or is this not
  correct?

This is mostly correct.  The longlong.h asm macros will still be
used.

But this sort of failures are not because the GMP assembly code
has bugs.  Disabling some assembly will not resolve it.  With a
buggy compiler, it is surely better to use our well-tested
assembly than to stress the compiler with additional C code.  :-)

  Yet I get the same problem when I run the test suite.  Does
  this indicate that there is a problem with the gcc compiler -
  which I should report to the GCC folks?
  
I'm almost certain we've seen this one and reported it already.
Kevin will remember better.

-- 
Torbjörn