2002-02-07 20:12:04

by Felipe Contreras

[permalink] [raw]
Subject: Weird bug in linux, glibc, gcc or what?

Hi,

I've found a weird problem in linuxthreads. When I get out of a thread it
happends one of three, the new thread get's defuct and the proccess never
ends, it segfaults, or it works.

The most weird is that it depends on the kernel, and also when I run the
test trought gdb there is no problem.

Here is the test:

#include <pthread.h>

void *test(void *arg) {
puts("Thread2");
return 0;
}

int main() {
pthread_t tt;
puts("Before Thread2");
pthread_create(&tt,NULL,test,NULL);
puts("After Thread2");
return 0;
}

The output:

1:src# ./test
Before Thread2
After Thread2
Thread2

This time it just kept waiting:

8957 vc/1 00:00:00 test
8958 ? 00:00:00 test <defunct>

I run it again:

1:src# ./test
Before Thread2
After Thread2

And again:

1:src# ./test
Before Thread2
Thread2
Segmentation fault

Now with gdb-5.1.1:

Starting program: /usr/src/./test
(no debugging symbols found)...[New Thread 1024 (LWP 9168)]
Before Thread2
[New Thread 2049 (LWP 9169)]
[New Thread 1026 (LWP 9170)]
Thread2
After Thread2

Program exited normally.

As I said the results vary from system to system, here are some
convinations:

* linux-2.4.10+glibc+2.2.4+gcc-2.95.3: Runs fine, but once in a while it
keeps waiting for the defunct thread.

* linux-2.4.17+glibc+2.2.4+gcc-2.95.3: The same, but the defunct problem
happends much more often.

* linux-2.4.10+glibc+2.2.5+gcc-3.0.3: Segfaults or waits for defuncts.

* linux-2.4.17+glibc+2.2.5+gcc-3.0.3: The same.

I'm lost on this, I supose it's a convination of glibc-linux problem.

Any ideas?

--
Felipe Contreras


2002-02-08 09:49:00

by Catalin Marinas

[permalink] [raw]
Subject: Re: Weird bug in linux, glibc, gcc or what?

On Thu, 7 Feb 2002, Felipe Contreras wrote:

> I've found a weird problem in linuxthreads. When I get out of a thread it
> happends one of three, the new thread get's defuct and the proccess never
> ends, it segfaults, or it works.
[snip]
> #include <pthread.h>
>
> void *test(void *arg) {
> puts("Thread2");
> return 0;
> }
>
> int main() {
> pthread_t tt;
> puts("Before Thread2");
> pthread_create(&tt,NULL,test,NULL);
> puts("After Thread2");
> return 0;
> }

Try this patch and see if it works:

--- thr_test.c Fri Feb 8 09:45:35 2002
+++ thr_test1.c Fri Feb 8 09:45:36 2002
@@ -10,5 +10,6 @@ int main() {
puts("Before Thread2");
pthread_create(&tt,NULL,test,NULL);
puts("After Thread2");
+ pthread_join(tt, NULL);
return 0;
}

--
Catalin


________________________________________________________________________
This email has been scanned for all viruses by the MessageLabs SkyScan
service. For more information on a proactive anti-virus service working
around the clock, around the globe, visit http://www.messagelabs.com
________________________________________________________________________

2002-02-08 14:44:10

by J.A. Magallon

[permalink] [raw]
Subject: Re: Weird bug in linux, glibc, gcc or what?


On 20020207 Felipe Contreras wrote:
>Hi,
>
>I've found a weird problem in linuxthreads. When I get out of a thread it
>happends one of three, the new thread get's defuct and the proccess never
>ends, it segfaults, or it works.
>
>The most weird is that it depends on the kernel, and also when I run the
>test trought gdb there is no problem.
>
>Here is the test:
>
>#include <pthread.h>
>
>void *test(void *arg) {
> puts("Thread2");
> return 0;
>}
>
>int main() {
> pthread_t tt;
> puts("Before Thread2");
> pthread_create(&tt,NULL,test,NULL);
> puts("After Thread2");
> return 0;
>}
>

Buggy program that could give unspecified behaviour, unless pthread
standard talks about orphaned threads...

Your main program can die (exit) before child thread ends, so it has
nobody to notify its dying or return to.

Try with a sleep(1) before main return, or better, do a
pthread_join().

--
J.A. Magallon # Let the source be with you...
mailto:[email protected]
Mandrake Linux release 8.2 (Cooker) for i586
Linux werewolf 2.4.18-pre9-slb #3 SMP Fri Feb 8 01:33:12 CET 2002 i686