This is the mail archive of the
gsl-discuss@sources.redhat.com
mailing list for the GSL project.
Re: Patch for histogram/stat2d.c
- From: Achim Gaedke <achim at zpr dot uni-koeln dot de>
- To: Brian Gough <bjg at network-theory dot co dot uk>
- Cc: gsl discussion list <gsl-discuss at sources dot redhat dot com>
- Date: Tue, 5 Mar 2002 21:22:00 +0100 (MET)
- Subject: Re: Patch for histogram/stat2d.c
Here is a first guess for explanation:
It depends on the floating point unit initialisation and some other
circumstances:
The FPU does not fail and gives NaN. On GNU system it defaults to 0 if
added to something else. This behaviour is useful in this case, but not
good to rely on.
I am used to such platform dependent errors, just because some systems
clear memory before allocation, and other do not. So some programs fail
when being built on another system.
I deliberate over the release strategy:
Is it possible to make a prerelease that is announced? The developers seem
to use x86 machines and there are a lot of very helpful users and sys
admins with a variety of machines. A prerelease is an offer to participate
in quality assurance without having developer tools at hand.
At least it is my fault, that I hav'nt tested it on SUN earlier.
Achim
Achim Gaedke, ZPR
Weyertal 80, 50931 Köln
Tel: +49 221 470 6021