Bug#448527: libc6: hypotf underflows for small inputs

["Duncan Po" <Duncan.Po@mathworks.com>]

Package: libc6
Version: 2.3.6.ds1-13
Severity: normal

The hypotf function in libm produces incorrect results for the code
below.

Comparing Debian 4 glibc source with current upstream source code shows
a change in the constants used for bit fiddling in the function. It
should be possible to backport the fix.

Current implementation on Debian 4 produces a result of 0 instead of
2.009718e-14.

#include <math.h>
#include <stdio.h>

int main() {
   float a,b,c;

   a = -1.81341284e-16;
   b = 2.00963666e-14;

   c = hypotf(a,b);
   
   printf("result of hypotf(%e,%e) is %e.\n", a, b, c); }

-- System Information:
Debian Release: 4.0
  APT prefers stable
  APT policy: (500, 'stable')
Architecture: amd64 (x86_64)
Shell:  /bin/sh linked to /bin/bash
Kernel: Linux 2.6.18-4-amd64
Locale: LANG=C, LC_CTYPE=C (charmap=ANSI_X3.4-1968)

Versions of packages libc6 depends on:
ii  tzdata                        2007b-1    Time Zone and Daylight
Saving Time

libc6 recommends no packages.

-- no debconf information