From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <gcc-bugs-return-465801-listarch-gcc-bugs=gcc.gnu.org@gcc.gnu.org>
Received: (qmail 6517 invoked by alias); 5 Nov 2014 14:34:45 -0000
Mailing-List: contact gcc-bugs-help@gcc.gnu.org; run by ezmlm
Precedence: bulk
List-Id: <gcc-bugs.gcc.gnu.org>
List-Archive: <http://gcc.gnu.org/ml/gcc-bugs/>
List-Post: <mailto:gcc-bugs@gcc.gnu.org>
List-Help: <mailto:gcc-bugs-help@gcc.gnu.org>
Sender: gcc-bugs-owner@gcc.gnu.org
Received: (qmail 6470 invoked by uid 48); 5 Nov 2014 14:34:40 -0000
From: "vogt at linux dot vnet.ibm.com" <gcc-bugzilla@gcc.gnu.org>
To: gcc-bugs@gcc.gnu.org
Subject: [Bug go/63269] libgo/math test failures in TestLog2
Date: Wed, 05 Nov 2014 14:34:00 -0000
X-Bugzilla-Reason: CC
X-Bugzilla-Type: changed
X-Bugzilla-Watch-Reason: None
X-Bugzilla-Product: gcc
X-Bugzilla-Component: go
X-Bugzilla-Version: 5.0
X-Bugzilla-Keywords:
X-Bugzilla-Severity: normal
X-Bugzilla-Who: vogt at linux dot vnet.ibm.com
X-Bugzilla-Status: WAITING
X-Bugzilla-Priority: P3
X-Bugzilla-Assigned-To: ian at airs dot com
X-Bugzilla-Target-Milestone: ---
X-Bugzilla-Flags:
X-Bugzilla-Changed-Fields:
Message-ID: <bug-63269-4-ii2F9ze6ev@http.gcc.gnu.org/bugzilla/>
In-Reply-To: <bug-63269-4@http.gcc.gnu.org/bugzilla/>
References: <bug-63269-4@http.gcc.gnu.org/bugzilla/>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: 7bit
X-Bugzilla-URL: http://gcc.gnu.org/bugzilla/
Auto-Submitted: auto-generated
MIME-Version: 1.0
X-SW-Source: 2014-11/txt/msg00273.txt.bz2

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=63269

--- Comment #4 from Dominik Vogt <vogt at linux dot vnet.ibm.com> ---
regarding 2)

> I'm not entirely persuaded by your argument for item 2. ...

Hm, good that you doubted it, because the actual mistake is somehwere else: 
The unpatched code has

  if l != float64(i)

but if you want to use a tolerance here this must become

  if !veryclose(float64(i), l) {

With the argument reversed.  This could/should be cleaned up by renaming the
arguments of the tolerance() function, e.g. a -> expected, b -> result, e ->
maxerr.

> Zero is a special
> value.  When we expect a zero, we should get a zero, not something close to
> zero.  I don't think this change is correct in general.  It may be correct for
> some specific cases, but then we need to investigate those.

Actually, this has nothing to do with 0 being special here, abut with scaling
of the allowed error: Multiplying it by 0 yields zero error tolerance, so the
tolerance() function does not do that.

--> This chunk is not necessary, but a (separate) cleanup patch might help to
avoid future confusion.