Re: [boost] Bug in numeric_cast<> ??

From: David Abrahams (abrahams_at_[hidden])
Date: 2000-10-19 12:24:23

----- Original Message -----
From: "Swatosh, Joe R NWP" <joe.r.swatosh_at_[hidden]>
To: <boost_at_[hidden]>
Sent: Thursday, October 19, 2000 12:47 PM
Subject: [boost] Bug in numeric_cast<> ??

> Well, I've been lurking a while, and we've been using the bits of boost we
> find useful (thanks!) and we've finally run into a place where we got
> surprised.  Specifically, we've been using numeric_cast<> for a while and
> it's been working slick.  Yesterday we ran into a little problem so today
I
> added the following lines to cast_test.cpp, either one causes it to fail:
>
> assert( 0.0f == numeric_cast<float>( 0.0 ) );
> assert( 0.0 == numeric_cast<double>( 0.0 ) );
>
> Which looks to me like a bug, or am I abusing the notion of
numeric_cast<>?
>
> BTW we're using MSVC6.0sp4 with Dinkum's STL.

Your test looks right to me, though the notion of floating-point equality is
sometimes dicey.
I guess the question is, is this a compiler bug or a bug in the library?
Ah. We're being bitten by the asymmetry of numeric_limits<>::min() for
non-integer (floating) types. I'll check in and post a fix shortly. Thanks
for the bug report.

-Dave

P.S. Beman, could you add this test to the regression suite?