KCachegrind Profiler Frontend / Bugs / #9 Callgrind: threads.c

Logged In: YES
user_id=1238690

Hello,

we have seen the same:

Software verions used:
----------------------
Valgrind: valgrind-2.4.0.rc2.tar.bz2
Callgrind: CVS HEAD version of 11 Mars 2005

Platform
--------
tench:/net/ganet/disc1/home/mea/tmp/callgrind> uname -a
Linux tench.sup.cfmu.eurocontrol.be 2.4.9-e.27CFMU-sg #1 SMP
Thu Jun 10 12:25:19 METDST 2004 i686 unknown

tench:/etc> cat /etc/redhat-release
Red Hat Linux Advanced Server release 2.1AS (Pensacola)

Compiler used:
--------------
tench:/net/ganet/disc1/home/mea> gcc --version
gcc (GCC) 3.4.4 20041122 (prerelease)
Copyright (C) 2004 Free Software Foundation, Inc.
This is free software; see the source for copying
conditions. There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A
PARTICULAR PURPOSE.

mkdir work
cd work
tar -xvjf ../sources/valgrind-2.4.0.rc2.tar.bz2
cd valgrind-2.4.0.rc2
./configure --prefix=/net/ganet/disc1/home/mea/tmp/callgrind
make
make install
cd ..
export PATH=/net/ganet/disc1/home/mea/tmp/callgrind/bin:$PATH
export
PKG_CONFIG_PATH=/net/ganet/disc1/home/mea/tmp/callgrind/lib/pkgconfig

tar -xvjf ../sources/callgrind-20050311.tar.bz2
cd callgrind
./autogen.sh
./configure --prefix=/net/ganet/disc1/home/mea/tmp/callgrind
make
make install

cd /net/ganet/disc1/home/mea/tmp/
mkdir run
cd run

callgrind ls -l /usr/bin
-> OK, is working

Testprog
-------- test.c (got it somewhere on the internet)
/* pthread.c by detour@metalshell.com
*
* Create multiple POSIX thread processes and wait for
* each one to complete.
*
* http://www.metalshell.com/
*
*/

#include <unistd.h>
#include <pthread.h>
#include <stdio.h>
#include <time.h>

#define NUM_THREADS 10

/* This becomes the main() of the new thread. It
is requred to return void * and take void * as a
parameter. */
void *thread_function(void *arg) {
fprintf(stdout, "Thread: %d running.\n", (int)arg);

// Sleep for arg+1 seconds to verify pthread_join.

sleep((int)arg+1);

fprintf(stdout, "Thread: %d done.\n", (int)arg);

pthread_exit(0);
}

int main() {
int cnt;
pthread_t p_thread[NUM_THREADS];

for(cnt = 0; cnt < NUM_THREADS; cnt++)
/* Returns 0 on successful creation of the thread. The
second
parameter, left NULL for this example, allows for
options
like CPU scheduling to be set. */
if(pthread_create(&p_thread[cnt], NULL, thread_function,
(void *)cnt) != 0)
fprintf(stderr, "Error creating the thread");

// Cycle through each thread and wait until it is completed.

for(cnt = 0; cnt < NUM_THREADS; cnt++) {
// Waits for p_thread[cnt] to finish.

pthread_join(p_thread[cnt], NULL);
}

fprintf(stdout, "All threads completed.\n");

return 0;
}

Compilation and test
--------------------

cd test
gcc -g test.c -o test -lpthread

tench:/net/ganet/disc1/home/mea/tmp/callgrind/test> whence test
test
tench:/net/ganet/disc1/home/mea/tmp/callgrind/test>
callgrind ./test
==6868== Callgrind-0.9.11, a call-graph generating cache
profiler for x86-linux.
==6868== Copyright (C) 2002-2004, and GNU GPL'd, by
J.Weidendorfer, N.Nethercote et al.
==6868== Using valgrind-2.4.0.rc2, a program supervision
framework for x86-linux.
==6868== Copyright (C) 2000-2005, and GNU GPL'd, by Julian
Seward et al.
--6868-- warning: L3 cache detected but ignored

--6868-- warning: Pentium with 12 K micro-op instruction
trace cache
--6868-- Simulating a 16 KB cache with 32 B lines
==6868== For more details, rerun with: -v
==6868==
Thread: 0 running.
0x0002EA28
BB# 81558

Callgrind: threads.c:251 (vgSkin_post_signal): Assertion
`*(vgSkin_current_fn_stack.top) == 0' failed.
==6868== at 0xB003514B: vgPlain_skin_assert_fail
(vg_mylibc.c:1170)
==6868== by 0xB003514A: assert_fail (vg_mylibc.c:1166)
==6868== by 0xB0035182: vgPlain_skin_assert_fail
(vg_mylibc.c:1171)
==6868== by 0xB118DE1C: vgSkin_post_signal (threads.c:267)
==6868== by 0xB00785E6:
vgSkinInternal_post_deliver_signal (vg_toolint.c:500)
==6868== by 0xB008FC87: vgArch_signal_return (signal.c:640)
==6868== by 0xB008E6E6: x86_linux_sys_sigreturn_before
(syscalls.c:584)
==6868== by 0xB005F4CA: vgPlain_client_syscall
(vg_syscalls.c:6138)
==6868== by 0xB001754C: handle_syscall (vg_scheduler.c:632)
==6868== by 0xB001786F: vgPlain_scheduler
(vg_scheduler.c:733)
==6868== by 0xB00905FD: vgArch_thread_wrapper (core_os.c:69)

sched status:
running_tid=1

Thread 1: status = VgTs_Runnable
==6868== at 0x3A9D4C85: sigsuspend (in /lib/libc-2.2.4.so)
==6868== by 0x3A997C18: (within /lib/libpthread-0.9.so)
==6868== by 0x3A997CFB: pthread_create (in
/lib/libpthread-0.9.so)
==6868== by 0x80485DF: main (test.c:40)

Thread 2: status = VgTs_WaitSys
==6868== at 0x3AA8DB97: poll (in /lib/libc-2.2.4.so)
==6868== by 0x3A994CEF: (within /lib/libpthread-0.9.so)
==6868== by 0x3AA95829: clone (in /lib/libc-2.2.4.so)

Thread 3: status = VgTs_WaitSys
==6868== at 0x3AA61CB1: nanosleep (in /lib/libc-2.2.4.so)
==6868== by 0x804856D: thread_function (test.c:25)
==6868== by 0x3A994FAE: (within /lib/libpthread-0.9.so)
==6868== by 0x3AA95829: clone (in /lib/libc-2.2.4.so)

Thread 4: status = VgTs_Runnable
==6868== at 0x3AA9581E: clone (in /lib/libc-2.2.4.so)

Note: see also the FAQ.txt in the source distribution.
It contains workarounds to several common problems.

If that doesn't help, please report this bug to:
Josef.Weidendorfer@gmx.de

In the bug report, send all the above text, the valgrind
version, and what Linux distro you are using. Thanks.

tench:/net/ganet/disc1/home/mea/tmp/callgrind/test>

Callgrind: threads.c

Group

Searches

Help

#9 Callgrind: threads.c

Discussion