Resource deadlock on program termination

[BenBE]

Okay, here's another one:

terminate called after throwing an instance of 'std::system_error'
  what():  Verklemmung beim Zugriff auf eine Ressource vermieden
==46880== 
==46880== Process terminating with default action of signal 6 (SIGABRT)
==46880==    at 0x568315B: raise (raise.c:51)
==46880==    by 0x5662858: abort (abort.c:79)
==46880==    by 0x536C910: ??? (in /usr/lib/x86_64-linux-gnu/libstdc++.so.6.0.28)
==46880==    by 0x537838B: ??? (in /usr/lib/x86_64-linux-gnu/libstdc++.so.6.0.28)
==46880==    by 0x5377368: ??? (in /usr/lib/x86_64-linux-gnu/libstdc++.so.6.0.28)
==46880==    by 0x5377D20: __gxx_personality_v0 (in /usr/lib/x86_64-linux-gnu/libstdc++.so.6.0.28)
==46880==    by 0x560FBEE: ??? (in /lib/x86_64-linux-gnu/libgcc_s.so.1)
==46880==    by 0x5610280: _Unwind_RaiseException (in /lib/x86_64-linux-gnu/libgcc_s.so.1)
==46880==    by 0x537869B: __cxa_throw (in /usr/lib/x86_64-linux-gnu/libstdc++.so.6.0.28)
==46880==    by 0x536F73E: std::__throw_system_error(int) (in /usr/lib/x86_64-linux-gnu/libstdc++.so.6.0.28)
==46880==    by 0x53A505F: std::thread::join() (in /usr/lib/x86_64-linux-gnu/libstdc++.so.6.0.28)
==46880==    by 0x12B0DB: drop_background(background_t*) (background.cc:133)
==46880==    by 0x12E2BB: std::_Sp_counted_deleter<background_t*, void (*)(background_t*), std::allocator<void>, (__gnu_cxx::_Lock_policy)2>::_M_dispose() (shared_ptr_base.h:471)
==46880==    by 0x129729: std::_Sp_counted_base<(__gnu_cxx::_Lock_policy)2>::_M_release() (shared_ptr_base.h:155)
==46880==    by 0x128FA2: std::__shared_count<(__gnu_cxx::_Lock_policy)2>::~__shared_count() (shared_ptr_base.h:730)
==46880==    by 0x128C79: std::__shared_ptr<background_t, (__gnu_cxx::_Lock_policy)2>::~__shared_ptr() (shared_ptr_base.h:1169)
==46880==    by 0x128C99: std::shared_ptr<background_t>::~shared_ptr() (shared_ptr.h:103)
==46880==    by 0x12AEA7: read_thread(std::weak_ptr<background_t>) (background.cc:116)
==46880==    by 0x12E5FD: void std::__invoke_impl<void, void (*)(std::weak_ptr<background_t>), std::weak_ptr<background_t> >(std::__invoke_other, void (*&&)(std::weak_ptr<background_t>), std::weak_ptr<background_t>&&) (invoke.h:60)
==46880==    by 0x12E53B: std::__invoke_result<void (*)(std::weak_ptr<background_t>), std::weak_ptr<background_t> >::type std::__invoke<void (*)(std::weak_ptr<background_t>), std::weak_ptr<background_t> >(void (*&&)(std::weak_ptr<background_t>), std::weak_ptr<background_t>&&) (invoke.h:95)
==46880==    by 0x12E49A: void std::thread::_Invoker<std::tuple<void (*)(std::weak_ptr<background_t>), std::weak_ptr<background_t> > >::_M_invoke<0ul, 1ul>(std::_Index_tuple<0ul, 1ul>) (thread:244)
==46880==    by 0x12E42D: std::thread::_Invoker<std::tuple<void (*)(std::weak_ptr<background_t>), std::weak_ptr<background_t> > >::operator()() (thread:251)
==46880==    by 0x12E3BF: std::thread::_State_impl<std::thread::_Invoker<std::tuple<void (*)(std::weak_ptr<background_t>), std::weak_ptr<background_t> > > >::_M_run() (thread:195)
==46880==    by 0x53A4DE3: ??? (in /usr/lib/x86_64-linux-gnu/libstdc++.so.6.0.28)
==46880==    by 0x5623608: start_thread (pthread_create.c:477)
==46880==    by 0x575F262: clone (clone.S:95)

Stack trace originates from running inside callgrind (for performance measurement), but I saw the same issue also running backscrub plain …

[phlash]

Hmmn, that seems to be the read_thread trying to join itself because it held the last reference to the shared_ptr.. might want to review the use of shared_ptr (again!)

[BenBE]

Good hint. Will have a look at this once I pushed the un-related commits from the multipass branch to experimental.