This is the mail archive of the libc-alpha@sourceware.org mailing list for the glibc project.

Index Nav:	[Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav:	[Date Prev] [Date Next]	[Thread Prev] [Thread Next]
Other format:	[Raw text]

Re: [PATCH] Fix bogus exceptions on exp(double)

From: Jeff Law <law at redhat dot com>
To: "Joseph S. Myers" <joseph at codesourcery dot com>
Cc: Aurelien Jarno <aurelien at aurel32 dot net>, libc-alpha at sourceware dot org
Date: Fri, 24 Feb 2012 09:58:51 -0700
Subject: Re: [PATCH] Fix bogus exceptions on exp(double)
References: <20120221135013.GA29215@hall.aurel32.net> <Pine.LNX.4.64.1202220118280.22118@digraph.polyomino.org.uk>

On 02/21/2012 06:20 PM, Joseph S. Myers wrote:

On Tue, 21 Feb 2012, Aurelien Jarno wrote:

This patch fixes exp(double) triggering bogus underflow and overflow
exceptions. More details are in:
http://sourceware.org/bugzilla/show_bug.cgi?id=13705

2012-02-21 Aurelien Jarno<aurelien@aurel32.net>
	[BZ #13705]
	* sysdeps/ieee754/dbl-64/w_exp.c(__exp): Use __kernel_standard
	instead of __kernel_standard_f.


This seems pretty obvious as fixing something that's clearly using the
wrong API to use the right one - but shouldn't there be a test added to
the testsuite for this bug, if it isn't triggered by existing tests?

(Do we have any libm experts reviewing patches?  My rint and nearbyint
patches could do with review.)

I certainly thought it was pretty obvious when I looked at the original downstream Fedora report. However, not knowing the code at all, I figured it was best to defer in case there was a reason for explicitly using the _f variants.

jeff

References:
- [PATCH] Fix bogus exceptions on exp(double)
  - From: Aurelien Jarno
- Re: [PATCH] Fix bogus exceptions on exp(double)
  - From: Joseph S. Myers

Index Nav:	[Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav:	[Date Prev] [Date Next]	[Thread Prev] [Thread Next]