From: Beman Dawes (bdawes_at_[hidden])
Date: 2003-12-31 12:44:24
At 07:07 AM 12/31/2003, John Maddock wrote:
> ====== BEGIN OUTPUT ======
>> Running 18 test cases...
>> ../../../libs/math/quaternion/quaternion_test.cpp(576): error in
>> "::boost::bind(static_cast < void (*) (const char *) >(&exp_test<long
>> double>), "long double")": test
>> 2*numeric_limits<T>::epsilon()) failed for (1.22465e-16, 2.1684e-19)
>> Which all look rather worrying, anyone any ideas what's going on here?
>Further testing suggests that this is a numeric_limits issue: the tests
>pass on linux and with mingw32 gcc 3.3, but fail as above on cygwin,
>strange... anyway patches applied (to main branch only).
On the Win32 tests, the change:
* Had no effect on the gcc 3.3.1 results.
* Caused two Intel 8.0 tests to change from fail to pass.
* Caused four Metrowerks tests to change from pass to fail.
See the results just posted at
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk