Package: libc6-dev Version: 2.2.5-11.2
Description: a system function call within a pthread (when statically linked) ends the program but leaves the processus zombified
The same problem does not occur on unstable systems (with libc6 2.3.2-4)
Here is a test program that demonstrates the problem, it consists of a main function which calls ThreadFunction which itself calls system(const char*).
/* test.c */ #include <stdlib.h> #include <pthread.h> #include <stdio.h>
void* ThreadFunction(void* arg)
{
printf("Thread Started\n");system("ls -l");
printf("Thread Stopping\n");
return NULL; }
int main(int argc, char** argv)
{
pthread_t thread;
void* result;
pthread_create(&thread, NULL, ThreadFunction, NULL);
pthread_join(thread, &result);
return 0; } /* END test.c
example of dynamic linking: # gcc test.c -lpthread -o test # ldd test libpthread.so.0 => /lib/lipthread.so (0x40017000) libc.so.6 => /lib/libc.so.6 (0x4002b000) /lib/ld-linux.so.2 => (0x40000000) # ./test Thread Started ... (output of ls -l) Thread Stopping // CORRECT BEHAVIOUR
example of static linkinkg: # gcc test.c -static -lpthread -o test # ldd test not a dynamic executable # ./test Thread Started # ps aux | grep test root 20027 0.0 0.1 2548 160 pts/1 S 13:26 ./test root 20028 0.0 0.1 2548 160 pts/1 S 13:26 ./test
// after launching ./test, Thread Started is printed out, but the command prompt shows up immediately and ps sees the processus still running.
if i do this on a machine running 2.3.2-4 or gentoo using 2.3.2 as well, the problem does not occur...
i used diiferent compilers with no noticeable change:
# dpkg -l gcc*
ii gcc-2.95 2.95.4-7
ii gcc-3.0 3.0.4-7
is this a bug? is this backportable from the 2.3 series?
oh by the way, uname -a shows 2.4.18-586tsc but fails as well on 2.4.21 (homemade) and 2.4.22 (gentoo).
Severity: unknown (annoying to me)
have a nice day,
Gabriel Biberian
-- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
