Segfault in intrusive_ptr destructor

[mavam]

I haven't managed to write a small test case, but here is already the stacktrace. Ideally you know what's going on, but I believe this requires more context. The troublemaker is a cow_tuple<T> stored outside a libcppa handler. I obtain the cow_tuple as follows:

auto opt = tuple_cast<T>(last_dequeued());
assert(opt.valid());
f(*opt); // stores *opt.

As the actors terminate, this segfault occurs:

#0  0x0000000000000035 in ?? ()
#1  0x00007ffff7a94bd5 in cppa::intrusive_ptr<cppa::process_information>::~intrusive_ptr (this=0x7fffac000b80, __in_chrg=<optimized out>) at /home/matthias/opt/gcc/include/cppa/intrusive_ptr.hpp:88
#2  0x00007ffff671fe0b in cppa::detail::tdata<cppa::intrusive_ptr<cppa::process_information>, cppa::detail::addressed_message>::~tdata (this=0x7fffac000b58, __in_chrg=<optimized out>) at /home/matthias/libcppa/./cppa/detail/tdata.hpp:227
#3  0x00007ffff67226c1 in cppa::detail::tuple_vals<cppa::intrusive_ptr<cppa::process_information>, cppa::detail::addressed_message>::~tuple_vals (this=0x7fffac000b40, __in_chrg=<optimized out>) at /home/matthias/libcppa/./cppa/detail/tuple_vals.hpp:46
#4  0x00007ffff6722720 in cppa::detail::tuple_vals<cppa::intrusive_ptr<cppa::process_information>, cppa::detail::addressed_message>::~tuple_vals (this=0x7fffac000b40, __in_chrg=<optimized out>) at /home/matthias/libcppa/./cppa/detail/tuple_vals.hpp:46
#5  0x00007ffff7b6ab53 in cppa::intrusive_ptr<cppa::detail::abstract_tuple>::reset (this=0x61b818, new_value=0x617fc0) at /home/matthias/opt/gcc/include/cppa/intrusive_ptr.hpp:113
#6  0x00007ffff7b697e1 in cppa::cow_ptr<cppa::detail::abstract_tuple>::reset (this=0x61b818, value=0x617fc0) at /home/matthias/opt/gcc/include/cppa/cow_ptr.hpp:88
#7  0x00007ffff6729af9 in cppa::any_tuple::reset (this=0x61b818) at /home/matthias/libcppa/src/any_tuple.cpp:59
#8  0x00007ffff6736c01 in cppa::detail::abstract_actor<cppa::local_actor>::release_node (this=0x61b540, node=0x61b800) at /home/matthias/libcppa/./cppa/detail/abstract_actor.hpp:210
#9  0x00007ffff6735f8d in cppa::detail::receive_policy::invoke<cppa::thread_mapped_actor, cppa::behavior> (this=0x61b738, client=0x61b540, node=0x61b800, fun=..., awaited_response=...) at /home/matthias/libcppa/./cppa/detail/receive_policy.hpp:113
#10 0x00007ffff6735d77 in cppa::detail::receive_policy::receive_wo_timeout<cppa::thread_mapped_actor, cppa::behavior> (this=0x61b738, client=0x61b540, fun=...) at /home/matthias/libcppa/./cppa/detail/receive_policy.hpp:137
#11 0x00007ffff6735276 in cppa::detail::receive_policy::receive<cppa::thread_mapped_actor> (this=0x61b738, client=0x61b540, bhvr=...) at /home/matthias/libcppa/./cppa/detail/receive_policy.hpp:149
#12 0x00007ffff6734b24 in cppa::detail::stacked_actor_mixin<cppa::thread_mapped_actor, cppa::detail::abstract_actor<cppa::local_actor> >::dequeue (this=0x61b540, bhvr=...) at /home/matthias/libcppa/./cppa/detail/stacked_actor_mixin.hpp:64
#13 0x00007ffff674b712 in cppa::detail::do_receive_helper::until<cppa::ge_reference_wrapper<bool> >(cppa::ge_reference_wrapper<bool>&&) (this=0x7ffff270f8d0, stmt=...) at /home/matthias/libcppa/./cppa/detail/receive_loop_helper.hpp:122
#14 0x00007ffff6745bfb in cppa::detail::mailman_loop () at /home/matthias/libcppa/src/mailman.cpp:114
#15 0x00007ffff674dd2a in operator() (__closure=0x61ba50) at /home/matthias/libcppa/src/network_manager.cpp:82
#16 0x00007ffff675038e in std::_Bind_simple<{anonymous}::network_manager_impl::start()::<lambda()>()>::_M_invoke<>(std::_Index_tuple<>) (this=0x61ba50) at /home/matthias/opt/gcc/lib/gcc/x86_64-redhat-linux/4.7.0/../../../../include/c++/4.7.0/functional:1598
#17 0x00007ffff6750269 in std::_Bind_simple<{anonymous}::network_manager_impl::start()::<lambda()>()>::operator()(void) (this=0x61ba50) at /home/matthias/opt/gcc/lib/gcc/x86_64-redhat-linux/4.7.0/../../../../include/c++/4.7.0/functional:1586
#18 0x00007ffff675019a in std::thread::_Impl<std::_Bind_simple<{anonymous}::network_manager_impl::start()::<lambda()>()> >::_M_run(void) (this=0x61ba38) at /home/matthias/opt/gcc/lib/gcc/x86_64-redhat-linux/4.7.0/../../../../include/c++/4.7.0/thread:115
#19 0x00007ffff55f76c0 in std::(anonymous namespace)::execute_native_thread_routine (__p=<optimized out>) at ../../../../../libstdc++-v3/src/c++11/thread.cc:73
#20 0x0000003191207b31 in start_thread () from /lib64/libpthread.so.0
#21 0x0000003190edfd2d in clone () from /lib64/libc.so.6

System: 64-bit Linux with GCC 4.7.

[Neverlord]

I don't think tuple_cast is the problem here, since the following runs just fine.

int main() {
    cow_tuple<int> itup;
    cow_tuple<int> itup2;

    send(self, 42);
    send(self, 1, 2.0f);
    int i = 0;
    receive_for(i, 2) (
        on<int>() >> [&] {
            auto opt = tuple_cast<int>(self->last_dequeued());
            if (opt) itup = *opt;
            else cerr << "WTF? tuple_cast failed in line " << __LINE__ << endl;
        },
        on<int, float>() >> [&] {
            auto opt = tuple_cast<int, anything>(self->last_dequeued());
            if (opt) itup2 = *opt;
            else cerr << "WTF? tuple_cast failed in line " << __LINE__ << endl;
        }
    );
    cout << get<0>(itup) << endl;
    cout << get<0>(itup2) << endl;
    return 0;
}

However, the stack trace points to mailman_loop, which is part of libcppa's network layer. So it might be serialization issue.

[mavam]

This one does not seem to crop up anymore after pulling from unstable. Let's keep the issue open after a bit more testing.

[mavam]

I am closing this issue after not being able to reproduce it anymore.