Skip to contentSkip to conversation

what happened with thread, from 2.2 to 2.4 ?!

DISCUSSION Blindauer Emmanuel 3 participants 3 messages Dec 2, 2001 – Dec 4, 2001

lore
Share & tools
lore
#0 lore raw

Hi
what happenned with thread from 2.2 to 2.4?
I have some problems with threaded programs, working under 2.2 and no more under 2.4
The test program is short:
----
#include <stdlib.h>
void main() {
char *t="1.0";
double d=0;
d=strtod(t,(char **)NULL);
}
---
this program compiled with "gcc -g -lpthread test.c" has strange behaviour. The problem only appear using gdb to see the value of d
In most case, the value returned by strtod under a 2.4 kernel is nan. I say most, because some 2.4 kernels don't fail other this line. I have done some test with differents distributions, and so differents version of kernel, gcc, gdb and libc.
http://manu.agat.net/bug.html

When it was possible, for the computer with the bug, under an 2.4 kernel, I've recompiled a 2.2.20, and the bug has disappeared!

Does someone have an idea about that ?

#1 lore raw
Quoted 21 lines · you

Hi
what happenned with thread from 2.2 to 2.4?
I have some problems with threaded programs, working under 2.2 and no more under 2.4
The test program is short:
----
#include <stdlib.h>
void main() {
char *t="1.0";
double d=0;
d=strtod(t,(char **)NULL);
}
---
this program compiled with "gcc -g -lpthread test.c" has strange behaviour. The problem only appear using gdb to see the value of d
In most case, the value returned by strtod under a 2.4 kernel is nan. I say most, because some 2.4 kernels don't fail other this line. I have done some test with differents distributions, and so differents version of kernel, gcc, gdb and libc.
http://manu.agat.net/bug.html

Quoted 4 lines

When it was possible, for the computer with the bug, under an 2.4 kernel, I've recompiled a 2.2.20, and the bug has disappeared!

Does someone have an idea about that ?

I verified your problem using 2.4.16 (with gdb-20010813(MI_OUT)) and 2.2.15pre18 (with gdb-4.18). Thinking about it, I remembered that in the past there have been some problems with newer versions of gdb with threaded programs. So, I tried using gdb-4.18 on 2.4.16 and it worked fine.

Methinks that this looks like a gdb issue.

As further testing I added a couple of statements after the strtod() call:

        printf( "%f\n", d);
        return;

If I step through the program, I see the nan value displayed by gdb and also printed by printf(). If I put a breakpoint at the printf() and run the program, the proper value is printed out. This looks a whole lot like gdb is having a problem when stepping over the strtod() call.

I will forward a copy of this message to the gdb mailing list.

David

#2 lore raw
In reply to Blindauer Emmanuel

The kernel version may not be the reason for this.
glibc 2.2.1 or 2.2.2 had a problem which some uninitialised variables when linked to pthread. (i think i can remember the strfmon call) What is your glibc version ?

Gerold

Quoted 4 lines · Blindauer Emmanuel

Hi
what happenned with thread from 2.2 to 2.4?
I have some problems with threaded programs, working under 2.2 and no more under 2.4