Re: The glibc math bug. Not on my P-166 system!

[Sven N. Thommesen]

At 11:52 PM 7/14/98 +0100, you wrote:
>
>Paul,
>
>Let's put the lid on this one!
>
>The upgrade to glibc-2.0.7-13 solves the pow() bug. That entails
>installing RPMs for glibc-2.0.7-13, glibc-debug-2.0.7-13,
>glibc-devel-2.0.7-13 and texinfo > 3.11 (due due a dependency
>problem).
>
>Wouldn't it be nice to have a swarm-aware sysadmin!
>
>Steve.

Sorry, Steve, but quite possibly not yet ... ;-(

The rpms you mention are all included in the RedHat 5.1 distribution
(before any errata upgrades are applied), so anyone running 5.1 should be
in the clear.

However, as I reported earlier to Marcus, there is a problem with
DynamicGraph-1.2 and GraphLib-1.2 that *appears* to have something to do
with the pow() function. In particular, dynamicGraph coredumps even when
all of swarm has been compiled with 'CFLAGS="-g -O2
-fno-inline-functions"'. A gdb trace on the core says the problem is in the
pow() function in __math.h. The problem goes away if I replace
pow(something,0.5) with the mathematically equivalent sqrt(something) in
GraphLib-1.2/DiGraph.m (two places) and recompile GraphLib then DynamicGraph.

I'd appreciate it if others who have installed RedHat 5.1 would download
GraphLib and DynamicGraph, and either confirm or disconfirm this behavior.
As Alex says, a bug in one of the core libraries is not something that
inspires confidence!

Cheers,
Sven


                  ==================================
   Swarm-Support is for discussion of the technical details of the day
   to day usage of Swarm.  For list administration needs (esp.
   [un]subscribing), please send a message to <address@hidden>
   with "help" in the body of the message.
                  ==================================