New boost regressions since yesterday for gcc 3.0.1

[Gabriel Dos Reis]

Peter Schmid <schmid@snake.iap.physik.tu-darmstadt.de> writes:

| There are two new failures when running the boost regression test
| suite vs gcc 3.0.1 on the i686-pc-linux-gnu target. These problems were
| not present yesterday. 

Peter,

  Please, could you tell me where those DBL_MAX are coming from?  We
don't use those symbols in <limits>

[...]

| **** test failed: traits::const_max == traits::max(), file: ../libs/integer/integer_traits_test.cpp, line: 50
| Checking short; min is -32768, max is 32767
| Checking unsigned short; min is 0, max is 65535
| Checking int; min is -2147483648, max is 2147483647
| Checking unsigned int; min is 0, max is 4294967295
| Checking long; min is -2147483648, max is 2147483647
| Checking unsigned long; min is 0, max is 4294967295
| Checking int64_t (possibly long long); min is -9223372036854775808, max is 9223372036854775807
| Checking uint64_t (possibly unsigned long long); min is 0, max is 18446744073709551615

What is wrong with the above?

-- Gaby