Bug#547429: marked as done (gcc: incorrect code with optimisation)

[owner@bugs.debian.org (Debian Bug Tracking System)]

Your message dated Thu, 9 May 2013 12:06:18 +0400
with message-id <CAFwitOHDo1d5mTQ54gHUCMT=kZcuZMLBpLt3Qr3oP=C7sbo8zg@mail.gmail.com>
and subject line Float has not enough precision
has caused the Debian Bug report #547429,
regarding gcc: incorrect code with optimisation
to be marked as done.

This means that you claim that the problem has been dealt with.
If this is not the case it is now your responsibility to reopen the
Bug report if necessary, and/or fix the problem forthwith.

(NB: If you are a system administrator and have no idea what this
message is talking about, this may indicate a serious mail system
misconfiguration somewhere. Please contact owner@bugs.debian.org
immediately.)


-- 
547429: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=547429
Debian Bug Tracking System
Contact owner@bugs.debian.org with problems

--- Begin Message ---

• To: Debian Bug Tracking System <submit@bugs.debian.org>
• Subject: gcc: incorrect code with optimisation
• From: Jörgen Tegnér <jorgen.tegner@telia.com>
• Date: Sat, 19 Sep 2009 19:05:16 +0200
• Message-id: <20090919170516.15818.89869.reportbug@burken.floppyfwsecured.com>

Package: gcc
Version: 4:4.3.3-9
Severity: normal


The code below gives incorrect result for the subtraction when optimisation is used.

Without optimisation, the difference is 2.
With optimisation, the difference is 0.


#include <stdio.h>
#include <limits.h>

int main(void)
{
  int ii;
  float ftest;

  ii=INT_MAX-1;
  ftest=(float)ii;
  printf("%d %.11g %.11g\n",ii,ftest,ftest-ii);

  return 0;
}

Tested on both gcc 4.3 and 4.4, same error.



-- System Information:
Debian Release: squeeze/sid
  APT prefers unstable
  APT policy: (500, 'unstable')
Architecture: i386 (i686)

Kernel: Linux 2.6.24
Locale: LANG=en_GB.UTF-8, LC_CTYPE=en_GB.UTF-8 (charmap=UTF-8)
Shell: /bin/sh linked to /bin/bash

Versions of packages gcc depends on:
ii  cpp                           4:4.3.3-9  The GNU C preprocessor (cpp)
ii  gcc-4.3                       4.3.4-2    The GNU C compiler

Versions of packages gcc recommends:
ii  libc6-dev [libc-dev]          2.9-26     GNU C Library: Development Librari

Versions of packages gcc suggests:
ii  autoconf               2.64-2            automatic configure script builder
pn  automake1.9            <none>            (no description available)
ii  bison                  1:2.4.1.dfsg-3    A parser generator that is compati
ii  flex                   2.5.35-8          A fast lexical analyzer generator.
ii  gcc-doc                5:2               documentation for the GNU compiler
pn  gcc-multilib           <none>            (no description available)
ii  gdb                    6.8.50.20090628-4 The GNU Debugger
ii  libtool                2.2.6a-4          Generic library support script
ii  make                   3.81-6            An utility for Directing compilati
ii  manpages-dev           3.22-1            Manual pages about using GNU/Linux

-- no debconf information

--- End Message ---

--- Begin Message ---

• To: 547429-done@bugs.debian.org
• Subject: Float has not enough precision
• From: Aleksandr P <alexpikptz@gmail.com>
• Date: Thu, 9 May 2013 12:06:18 +0400
• Message-id: <CAFwitOHDo1d5mTQ54gHUCMT=kZcuZMLBpLt3Qr3oP=C7sbo8zg@mail.gmail.com>

Hello.

I try this with gcc-4.3.5-4 and gcc-4.4.5-8. I try -O0, -O1, -O2, -O3 and -Os optimization levels. In every case result was

2147483646 2147483648 0

Float has too little�precision - only 23 bits. Whereas int is 32 bits. If you use double result will be correct.

#include <stdio.h>

#include <limits.h>

int main(void)

{

��int ii;

��double ftest;

��ii=INT_MAX-1;

��ftest=(double)ii;

��printf("%d %.11g %.11g\n",ii,ftest,ftest-ii);

��return 0;

}

2147483646 2147483646 0

Best wishes,

Aleksandr

--- End Message ---