problem with std::numeric_limits<float>::epsilon() in gcc 3.0.1

[Kevin Ediger]

Hello All,

I'm having a problem with a file, libstdc++-v3/include/bits/std_limits.h,
included with gcc 3.0.1. The problem is with
std::numeric_limits<float>::epsilon() which is as follows starting on line 1773:

      ....
      static const int radix = 2;
      static float epsilon() throw()
      { return __glibcpp_float_radix; }
      static float round_error() throw()
      { return __glibcpp_float_round_error; }
      ....

It looks like epsilon() incorrectly returns __glibcpp_float_radix instead of
__glibcpp_float_epsilon. Is this a known problem?

I know that std_limits.h was generated in gcc 3.0, but this one appears to be
included directly in gcc-3.0.1.tar.gz.

Thanks,
Kevin Ediger
LI-COR, Inc.