[asio-users] bug in asio?

[Vlad <vl...@gm...>]

Hello All!

Few days ago i've developed simple single-threaded socks proxy using asio. It 
worked fine so i tried to make stress-test to test how many connections it 
will serve at once. At this moment i started to receive segfault-s.
On high loads (few thousand clients per second) my server makes a segfault 
with following backtrace:

*** glibc detected *** double free or corruption (fasttop): 0x09bb4330 ***

Program received signal SIGABRT, Aborted.
0x080e42a1 in kill ()
(gdb) bt 
#0  0x080e42a1 in kill ()
#1  0x080da90b in pthread_kill (thread=16384, signo=6) at signals.c:69
#2  0x080da93b in __pthread_raise (sig=6) at signals.c:200
#3  0x080e455b in abort ()
#4  0x080f14cf in __libc_message ()
#5  0x080f56f9 in malloc_printerr ()
#6  0x080f6b96 in free ()
#7  0x080ca601 in operator delete ()
#8  0x080538eb in __gnu_cxx::new_allocator<std::_List_node<std::pair<int 
const, asio::detail::reactor_op_queue<int>::op_base*> > >::deallocate 
(this=0x9643b28, 
    __p=0x9bb4330)
    
at /usr/lib/gcc/i486-linux-gnu/4.0.4/../../../../include/c++/4.0.4/ext/new_allocator.h:94
#9  0x08053910 in std::_List_base<std::pair<int const, 
asio::detail::reactor_op_queue<int>::op_base*>, std::allocator<std::pair<int 
const, asio::detail::reactor_op_queue<int>::op_base*> > >::_M_put_node 
(this=0x9643b28, __p=0x9bb4330)
    
at /usr/lib/gcc/i486-linux-gnu/4.0.4/../../../../include/c++/4.0.4/bits/stl_list.h:317
#10 0x08053971 in std::list<std::pair<int const, 
asio::detail::reactor_op_queue<int>::op_base*>, std::allocator<std::pair<int 
const, asio::detail::reactor_op_queue<int>::op_base*> > >::_M_erase 
(this=0x9643b28, __position={_M_node = 0x9bb4330})
    
at /usr/lib/gcc/i486-linux-gnu/4.0.4/../../../../include/c++/4.0.4/bits/stl_list.h:1163
#11 0x080539a3 in std::list<std::pair<int const, 
asio::detail::reactor_op_queue<int>::op_base*>, std::allocator<std::pair<int 
const, asio::detail::reactor_op_queue<int>::op_base*> > >::erase 
(this=0x9643b28, __position={_M_node = 0x9bb4330})
    
at /usr/lib/gcc/i486-linux-gnu/4.0.4/../../../../include/c++/4.0.4/bits/list.tcc:98
#12 0x08053b30 in asio::detail::hash_map<int, 
asio::detail::reactor_op_queue<int>::op_base*>::erase (this=0x9643b28, 
it={_M_node = 0x9bb4330})
    at /usr/local/include/asio/detail/hash_map.hpp:168
#13 0x08053c50 in asio::detail::reactor_op_queue<int>::dispatch_all_operations 
(
    this=0x9643b24, descriptor=319, result=@0xbfe704ec)
    at /usr/local/include/asio/detail/reactor_op_queue.hpp:163
#14 0x08054a79 in asio::detail::epoll_reactor<false>::run (this=0x9641af0, 
    block=false) at /usr/local/include/asio/detail/epoll_reactor.hpp:405
#15 0x0806270d in 
asio::detail::task_io_service<asio::detail::epoll_reactor<false> >::do_one 
(this=0x954a7a8, lock=@0xbfe70600, this_idle_thread=0xbfe705ac)
    at /usr/local/include/asio/detail/task_io_service.hpp:245
#16 0x0806291e in 
asio::detail::task_io_service<asio::detail::epoll_reactor<false> >::run 
(this=0x954a7a8) at /usr/local/include/asio/detail/task_io_service.hpp:87
#17 0x080629c4 in asio::io_service::run (this=0xbfe706c8)
    at /usr/local/include/asio/impl/io_service.ipp:48
#18 0x0805f712 in main (argc=3, argv=0xbfe70934) at main/main.cpp:38
(gdb) 

-----------------------------------------------------------------------

As you can see, asio tries to free memory twice.

About program: i'm using timers to make timeouts on client/server requests.

I'm profiled my proggie with valgrind, but it didn't found any errors except 
this.

btw. i've maked my software by modifying example from asio distribution 
(http_server).

Please help! thanks!