Bug 207031 - [x86_64] Floating point exception when starting 32bit C++ binaries (libstdc++-4.1.1-13)
[x86_64] Floating point exception when starting 32bit C++ binaries (libstdc++...
Product: Fedora
Classification: Fedora
Component: gcc (Show other bugs)
x86_64 Linux
medium Severity medium
: ---
: ---
Assigned To: Jakub Jelinek
Depends On:
  Show dependency treegraph
Reported: 2006-09-18 16:43 EDT by Ben Konrath
Modified: 2007-11-30 17:11 EST (History)
1 user (show)

See Also:
Fixed In Version:
Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of:
Last Closed: 2006-09-23 06:42:00 EDT
Type: ---
Regression: ---
Mount Type: ---
Documentation: ---
Verified Versions:
Category: ---
oVirt Team: ---
RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: ---

Attachments (Terms of Use)
oowriter strace (30.22 KB, text/plain)
2006-09-22 15:26 EDT, Ben Konrath
no flags Details
32bit c++ hello world (4.98 KB, application/octet-stream)
2006-09-23 06:18 EDT, Caolan McNamara
no flags Details

  None (edit)
Description Ben Konrath 2006-09-18 16:43:30 EDT
Description of problem:

I get this error when I try to start any openoffice.org program:

% oowriter /usr/lib/openoffice.org2.0/program/soffice: line 140: 23872 Floating
point exception(core dumped) "${sd_prog}"/pagein -L"${sd_prog}"
${sd_pagein_args} /usr/lib/openoffice.org2.0/program/soffice: line 151: 23873
Floating point exception(core dumped) "$sd_prog/$sd_binary" "$@"

% oocalc /usr/lib/openoffice.org2.0/program/soffice: line 140: 23885 Floating
point exception(core dumped) "${sd_prog}"/pagein -L"${sd_prog}"
${sd_pagein_args} /usr/lib/openoffice.org2.0/program/soffice: line 151: 23886
Floating point exception(core dumped) "$sd_prog/$sd_binary" "$@"

Version-Release number of selected component (if applicable):

% rpm -qa | grep openoffice.org

How reproducible:

Comment 1 Caolan McNamara 2006-09-19 03:38:13 EDT
Well, FC-5 OOo is 32bit on 64bit platforms. But otherwise there simply isn't
enough information here to make a guess as to what is wrong.

No crash reporter window ? So must have crashed very early in bootstrapping. Do
you have any other (large) 32bit apps, e.g. 32bit firefox ? Otherwise some
strace or gdb traces might help.
Comment 2 Caolan McNamara 2006-09-22 08:21:03 EDT
I'd need something to work on, it's not reproducable for me, and normally on a
fc-5 crasher like this I'd have a horde of similiar bug reports
Comment 3 Ben Konrath 2006-09-22 15:26:43 EDT
Created attachment 136966 [details]
oowriter strace

I've attached an strace. I don't have a lot of time to debug this right now but
your welcome to log into toast.toronto.redhat.com to debug if you have time.
Comment 4 Caolan McNamara 2006-09-23 06:18:40 EDT
Created attachment 136995 [details]
32bit c++ hello world

Checking on your machine, it seems that any 32bit c++ binary blows up with the
same FPE error, e.g. the attached a.out was created with...

g++ test.c on FC-5 where test.c is just...

#include <stdio.h>

int main(void)
	printf("hello world\n");
	return 0;

and blows up with the same SIGFPE
Comment 5 Caolan McNamara 2006-09-23 06:23:02 EDT
and gcc test.c -lstdc++ is sufficient to reproduce this, so looks like your
32bit libstdc++.so.6 is hosed somehow.

Not OOo specific, moving to gcc/libstdc++
Comment 6 Jakub Jelinek 2006-09-23 06:42:00 EDT
That's just user error, you must have used --nodeps to force installing of
FC6 package (libstdc++-4.1.1-13) on FC5.  That of course can't work,
FC6 packages need DT_GNU_HASH support in the dynamic linker, which is only
available in FC6 glibc.  libstdc++-4.1.1-13 has rtld(GNU_HASH) among Requires
and only FC6+ glibc provides rtld(GNU_HASH).
Comment 7 Ben Konrath 2006-09-23 14:54:36 EDT
I don't actually remember messing with the system in this way, but I guess I
must have. Thanks Caolan and Jakub for tracking this down. 

Note You need to log in before you can comment on or make changes to this bug.