Segmentation fault using boost::thread

Question

I've written an application using threads from boost::thread. It compiles and works fine on my local machine. Problem occurs on the one of the servers. I've send main.cpp file and compiled it the same way I did on my local machine:

g++ -g main.cpp -o rdzen -lboost_thread
ulimit -c unlimited

I'm executing it with:

./rdzen input.txt dictionary.txt output.txt

then I got:

Segmentation fault (core dumped)

I used gdb to find out the reason:

gdb rdzen core

The backtrace is:

#0  0x0804c039 in boost::detail::atomic_exchange_and_add (pw=0x53006d76, dv=-1)
at /usr/local/include/boost/smart_ptr/detail/sp_counted_base_gcc_x86.hpp:50
#1  0x0804c11a in boost::detail::sp_counted_base::release (this=0x53006d72)
at /usr/local/include/boost/smart_ptr/detail/sp_counted_base_gcc_x86.hpp:143
#2  0x0804c17c in ~shared_count (this=0xbd928a8c, __in_chrg=<value optimized out>)
at /usr/local/include/boost/smart_ptr/detail/shared_count.hpp:305
#3  0xb2b388e1 in ~shared_ptr (this=0xbd928b3c) at ./boost/smart_ptr/shared_ptr.hpp:169
#4  boost::shared_ptr<boost::detail::thread_data_base>::operator= (this=0xbd928b3c)
at ./boost/smart_ptr/shared_ptr.hpp:305
#5  boost::thread::start_thread (this=0xbd928b3c) at libs/thread/src/pthread/thread.cpp:184
#6  0x0805022c in thread<boost::_bi::bind_t<void, void (*)(int, char*, char*, char*, int), boost::_bi::list5<boost::_bi::value<int>, boost::_bi::value<char*>, boost::_bi::value<char*>, boost::_bi::value<char*>, boost::_bi::value<int> > > > (this=0xbd928b3c, f=...)
at /usr/local/include/boost/thread/detail/thread.hpp:205
#7  0x0804a88a in main (argc=4, argv=0xbd928cb4) at main.cpp:542

main.cpp:542 is:

boost::thread watek1(boost::bind(&watek, 0, argv[1], argv[2], argv[3], 0));

What is the reason and why the same code works on my local machine and not on the server? Thanks in advance for help.

`boost::thread` does its own binding, so your line should read `boost::thread watek1(&watek, 0, argv[1], argv[2], argv[3], 0);`. — ildjarn, Jul 02 '12 at 18:02
@Niall C.: void watek(int numer_watku, char *plik_modelu, char *plik_slownika, char *plik_wyjsciowy, int przesuniecie) But I think function doesn't matter. When I try to use just: void foo() { cout << "Thread X"; } segfault occurs too. — Daniel Gadawski, Jul 02 '12 at 18:14
@ildjarn: I changed it the way you've shown, but the result is the same. — Daniel Gadawski, Jul 02 '12 at 18:16
@rodrigo: No, I'm compiling just like in the content of my question. — Daniel Gadawski, Jul 02 '12 at 18:17
@DanielGadawski: Then you should try compiling with `-pthread` and see what happens. Using threads in GCC without this option is risky at best. — rodrigo, Jul 02 '12 at 18:28
Does the back trace always looks like this? Does it reproduceable happen every time you start your program? For it looks like it might be a vanilla heap corruption. Are the arguments (argv) still valid after the thread was constructed? Are you starting more that one thread like this? — Torsten Robitzki, Jul 04 '12 at 20:21
It was a server-side issue. After kernel update it works fine. — Daniel Gadawski, Jul 17 '12 at 06:35

Segmentation fault using boost::thread

0 Answers0