Possible GMP bug

Torbjorn Granlund tg at gmplib.org
Fri Jun 18 12:15:07 CEST 2010

"Davies, Daniel" <Daniel.Davies at xerox.com> writes:

  *       make[4]: Entering directory `/tool/gcc/4.5.0/buildgmp/tests/cxx'
  *       FAIL: t-assign
  *       FAIL: t-constr
  *       FAIL: t-rand
  *       make[4]: Leaving directory `/tool/gcc/4.5.0/buildgmp/tests/cxx'
  Is there anything else you'd like to know, or do you want to write it
  off as a compiler bug?
  I haven't exercised the g++ compiler - I'll try to download whatever's
  needed to run gcc's "make check".   I have extensively exercised the gcc
  compiler and that seems ok.
Many things can be wrong here, and I'd shamelessly claim the least
likely is some undefined GMP behaviour (i.e., a GMP bug).

Perhaps the C++ front end of GCC 4.5.0 is buggy.  Perhaps your compiler
was miscompiled (did you perform a 3-stage compile?).  Perhaps your
setup causes incompatible C++ libraries to be used with the new compiler.

The only way to know is to debug it.  Please do that!


More information about the gmp-bugs mailing list