Bugzilla Bug 12554

When using idle.py I got next erro, so I belive that this is a bug inside
Python
2.2.2

A similar error happens when you import LinearAlgebra (in Numeric) from the
prompt.

Python 2.2.2 (#1, Dec 22 2002, 17:44:21) 
[GCC 3.2.1] on linux2
Type "copyright", "credits" or "license" for more information.
IDLE 0.8 -- press F1 for help
>>> Exception in Tkinter callback
Traceback (most recent call last):
  File "/usr/lib/python2.2/lib-tk/Tkinter.py", line 1300, in __call__
    return apply(self.func, args)
  File "/usr/lib/python2.2/tools/idle/TreeWidget.py", line 124, in flip
    self.item.OnDoubleClick()
  File "/usr/lib/python2.2/tools/idle/ClassBrowser.py", line 206, in
OnDoubleClick
    edit.gotoline(self.cl.methods[self.name])
  File "/usr/lib/python2.2/tools/idle/EditorWindow.py", line 397, in gotoline
    self.center()
  File "/usr/lib/python2.2/tools/idle/EditorWindow.py", line 488, in center
    top, bot = self.getwindowlines()
  File "/usr/lib/python2.2/tools/idle/EditorWindow.py", line 497, in
getwindowlines
    bot = self.getlineno("@0,65535")
  File "/usr/lib/python2.2/tools/idle/EditorWindow.py", line 506, in getlineno
    return int(float(text.index(mark)))
OverflowError: float too large to convert

This looks like a very strange bug indeed. The bug is very simple to
demonstrate (for me at least):     $ python    Python 2.2.2 (#1, Dec 31 2002,
13:29:44)    [GCC 3.2.1 20021207 (Gentoo Linux 3.2.1-20021207)] on linux2   
Type "help", "copyright", "credits" or "license" for more information.    >>>
int(2.2)    0    >>> int(4)    4    >>> int(5.4)    0    >>> int(10.2)   
Traceback (most recent call last):      File "<stdin>", line 1, in ?   
OverflowError: float too large to convert    >>>^D  I've tried compiling python
with the following sets of CFLAGS:     CFLAGS="-march=pentium4 -O3 -pipe"   
CFLAGS="-mcpu=pentium4 -O3 -pipe"    CFLAGS="-O2 -pipe  (this is all with
gcc-3.2.1-r6 of the "unstable" gentoo portage).  

Subject: Re: [gentoo-user] int(float(1)) == 0 ? in Python 2.2.2
From: Nicholas Hockey <tilt@bluecherry.net>
To: gentoo-user@gentoo.org
Date: 02 Jan 2003 01:13:40 -0500
X-Mew: Text/Html in Multipart/Alternative as a singlepart
       Text/Html in Multipart/Alternative as a singlepart
       Text/Html in Multipart/Alternative as a singlepart
X-Mew: <8 body> No his/her public key. ID = not found
X-Mew: <12 body> No his/her public key. ID = not found
X-Mew: <14 body> No his/her public key. ID = not found

I think you got bit by a gcc-3 bug,
http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=8988
we got similare seg faults in BitchX, and if you notice the one of three ppl who
said it worked
fine, was the one using an odd version of gcc, i compiled it with the optamizer
flags set to -O3
-march=i686 which defuses the bug... another quick fix would be putting
-fno-strength-reduce in
your CFLAGS till you can upgrade

[tilt@blackhole] ~ > python
Python 2.2.2 (#1, Dec 10 2002, 14:18:33)
[GCC 3.2.1] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> int(1)
1
>>> int(1.0)
1
>>> int(9.01754349826526)
9
>>> int(9.01754)
9
>>> int(10)
10
>>>

On Sun, 2002-12-29 at 23:44, Cong wrote:

    Could you help me to figure out what happened !
    
    $ python
    Python 2.2.2 (#1, Dec 30 2002, 13:03:14) 
    [GCC 3.2.1] on linux2
    Type "help", "copyright", "credits" or "license" for more information.
    >>> int(1)
    1
    >>> int(1.0)
    0
    >>> int(9.01754349826526)
    Traceback (most recent call last):
      File "<stdin>", line 1, in ?
    OverflowError: float too large to convert
    >>> int(9.01754)
    0
    >>> int(10)
    10
    >>>
    
 

The int() bug is a showstopper for me doing any Python development on Gentoo.

I also tried using -march-i686 -O3, and that didn't help any.  Rob.

Ok, can you please attach the output of 'emerge info'? thanks!

# emerge info 
Portage 2.0.46-r9 (default-x86-1.4, gcc-3.2.1, glibc-2.3.1-r2) 
================================================================= 
System uname: 2.4.20-gentoo-r1 i686 Intel(R) Pentium(R) 4 CPU 2.40GHz 
GENTOO_MIRRORS="http://www.ibiblio.org/pub/Linux/distributions/gentoo" 
CONFIG_PROTECT="/etc /var/qmail/control /usr/kde/2/share/config 
/usr/X11R6/lib/X11/xkb /usr/kde/3.1/share/config /usr/kde/3/share/config 
/usr/share/config" 
CONFIG_PROTECT_MASK="/etc/gconf /etc/env.d" 
PORTDIR="/usr/portage" 
DISTDIR="/opt/tmp/distfiles" 
PKGDIR="/usr/portage/packages" 
PORTAGE_TMPDIR="/opt/tmp" 
PORTDIR_OVERLAY="/opt/portage" 
USE="x86 oss apm avi crypt cups encode gif jpeg libg++ mmx mpeg nls pdflib png 
quicktime spell truetype xmms xv zlib directfb gtkhtml alsa readline arts tetex aalib 
nas bonobo tcltk java guile ruby X gpm tcpd pam libwww ssl perl python esd gnome 
gtk qt kde opengl mozilla gphoto2 cdr scanner -3dnow acpi atlas -berkdb canna cjk 
dvd freewnn -gdbm icc icc-pgo imap -imlib jikes leim maildir mbox -mikmod -motif 
mule -ncurses oav -oggvorbis pic plotutils pnp -qtmt -sdl -slang slp sse -svga tiff 
wmf -xml2" 
COMPILER="gcc3" 
CHOST="i686-pc-linux-gnu" 
CFLAGS="-O3 -march=pentium4 -pipe -fomit-frame-pointer" 
CXXFLAGS="-O3 -march=pentium4 -pipe -fomit-frame-pointer" 
ACCEPT_KEYWORDS="x86" 
MAKEOPTS="-j2" 
AUTOCLEAN="true" 
SYNC="rsync://rsync.gentoo.org/gentoo-portage" 
FEATURES="sandbox ccache" 
 

Could it be that the python ebuild script overrides the -O3 compiler flag with
-O2 exposing this problem?

I'm also running into this problem in the emerge of media-sound/lilypond (bug
13396).

Jon, If you look closely at the output of emerge python, then it is clear that
it does pass all the arguments in the CFLAGS. In fact the ebuild teaches that
only an option of "-malign-double" would be filtered out. I believe Cong has the
proper answer to this problem if only I could manage to reproduce the problem
somehow...

Cong, I've recompiled python a few times now including exactly your settings
from emerge info, I haven't been able to reproduce this issue. Have you verified
that your system is "up to date"? (The only difference between my tries and
yours is that I had to use -march=i686 since I run on an Athlon and P4 code
wouldn't run on it.)

Hi ,  I'd like to keep SYSTEM and my favorite apps up-to-date.
Have a look at my /var/cache/edb/world too.

$ emerge  system -up --deep

These are the packages that I would merge, in order:

Calculating system dependencies ...done!

$ emerge world -up --deep

These are the packages that I would merge, in order:

Calculating world dependencies ...done!
[ebuild    UD] media-video/nvidia-kernel-1.0.3123-r2 [1.0.4191-r1]
[ebuild    UD] media-video/nvidia-glx-1.0.3123 [1.0.4191]
[ebuild    U ] media-libs/glut-3.7.1 [3.7-r2]
[ebuild    U ] app-text/openjade-1.3.1-r6 [1.3.1-r5]
[ebuild    UD] app-text/docbook-sgml-dtd-4.0-r1 [4.1-r1]
[ebuild  N   ] net-libs/openslp-1.0.9a
[ebuild    UD] kde-base/arts-1.0.5a [1.1.0_rc6]
[ebuild    UD] kde-base/kdelibs-3.0.5a [3.1_rc6]
[ebuild    UD] media-libs/xine-lib-0.9.13-r2 [1_beta2]
[ebuild    UD] media-libs/libpng-1.0.15 [1.2.5-r3]
[ebuild    U ] media-libs/libsdl-1.2.5-r1 [1.2.4.20020601]
[ebuild    UD] media-video/vlc-0.4.4 [0.4.5]
[ebuild    UD] app-cdr/cdrtools-1.11.33 [2.0]
[ebuild    UD] app-cdr/xcdroast-0.98_alpha10-r2 [0.98_alpha13]
[ebuild    UD] x11-wm/metacity-2.4.1 [2.4.8]
[ebuild    U ] x11-libs/wxGTK-2.4.0 [2.3.3]
[ebuild    U ] dev-python/wxPython-2.4.0.1 [2.3.3.1-r2]
[ebuild  N   ] media-video/mplayer-0.90_rc2
[ebuild    UD] dev-util/anjuta-1.0.0 [1.0.1]
[ebuild    UD] media-video/transcode-0.6.0 [0.6.3.20021205]
[ebuild    UD] media-video/dvdrip-0.46 [0.48.5]
[ebuild    UD] sys-apps/eject-2.0.10 [2.0.12]
[ebuild    U ] x11-libs/libzvt-2.0.1-r2 [2.0.1-r1]
[ebuild    U ] gnome-extra/gconf-editor-0.4.0 [0.3.1]
[ebuild  N   ] media-sound/sox-12.17.3-r3
[ebuild    UD] kde-base/kdebase-3.0.5a-r1 [3.1_rc6]

Created an attachment (id=7489) [details]
the WORLD file

> (The only difference between my tries and yours is that I had to 
> use -march=i686 since I run on an Athlon and P4 code wouldn't run on it.) 
 
This seems to be the exact problem. I suspect that gcc-3.2.1-r7 (which 
was patched with patches that supposedly should have fixed this problem, but apparently 
didn't) is miscompiling glibc-2.3.1, which in turn causes this problem in python. 
 
I can do 
 
	jed /etc/make.conf (changing the CFLAGS to "-march=p4 -O3 ...") 
	emerge glibc python 
	<do python test> 
	jed /etc/make.conf (change CFLAGS to "-mcpu=p4 -O3 ...") 
	emerge glibc python 
	<do python test> 
 
and the int() function *will* fail in the first instance, but int() will work fine in the 
second test. So it seems that there are still problems with glibc. (I haven't actually 
tried it, but the above should also reproduce the errors without recompiling python 
since it's dynamically linked with glibc). 
 

I have verified both 1 and 2 on my box:

1.
> and the int() function *will* fail in the first instance, but int() will work
fine in the 
> second test. So it seems that there are still problems with glibc. 

2.
> (I haven't actually 
> tried it, but the above should also reproduce the errors without recompiling
python 
> since it's dynamically linked with glibc). 
 

Known problems with -march=pentium4 -O3 ... please recompile glibc with
"-march=pentium3 -O2 -pipe".  Its actually a gcc3 problem.

With 
CFLAGS="-O3 -mcpu=pentium4 -pipe -fomit-frame-pointer"

it works too. Now I use it to compile glibc and the next for compiling others:
CFLAGS="-O3 -march=pentium4 -pipe -fomit-frame-pointer"

Since this has been confirmed as a gcc/libc bug I am going to close this
particular bug as it contains a workaround which will resolve the problem. This
will hopefully be corrected permanently with the next gcc/glibc release.