assert failed in processLimit

[penfree]

I got an exception and the program crashed, i don't really know how to reproduce it, it is not triggered by a special parameter. would you please help me to figure it out. @gelldur

• The error log: /thirdparty/EventBus-3.0.4/lib/src/dexode/EventBus.cpp:57: std::size_t dexode::EventBus::processLimit(std::size_t): Assertion `removeFrom != _eventStreams.end()' failed.

• The source code is here:

  {
      std::lock_guard writeGuard{_mutexStreams};
      if(!_eventStreams.empty())
      {
          // If anything was added then we need to add those elements
          std::move(_eventStreams.begin(), _eventStreams.end(), std::back_inserter(eventStreams));
      }
      std::swap(eventStreams, _eventStreams); // move data TO member
  
      // Check do we need remove something
      if(_eventStreams.size() != _eventToStream.size())
      {
          auto removeFrom = std::remove_if(
              _eventStreams.begin(), _eventStreams.end(), [this](const auto& eventStream) {
                  for(const auto& element : _eventToStream)
                  {
                      // Don't remove if we point to the same place (is it UB ?)
                      if(element.second == eventStream.get())
                      {
                          return false;
                      }
                  }
                  return true;
              });
          assert(removeFrom != _eventStreams.end());
          _eventStreams.erase(removeFrom, _eventStreams.end());
      }
  }

[penfree]

i have gdb into the core.

(gdb) p _eventStreams
$1 = std::vector of length 0, capacity 0
(gdb) p _eventToStream
$2 = std::map with 3 elements = {
  [0x55fad188f928 <dexode::eventbus::internal::type_id_ptr<dolphin::StreamAddEvent>::id>] = 0x55fad26f2510,
  [0x55fad18907f8 <dexode::eventbus::internal::type_id_ptr<dolphin::StreamRemoveEvent>::id>] = 0x55fad26af030,
  [0x55fad1891a48 <dexode::eventbus::internal::type_id_ptr<dolphin::StreamStatusChangeEvent>::id>] = 0x7f4644005fa0}
(gdb) quit

[gelldur]

Hello, could you send more ? I mean parts of your code? I will try to check again this part of code. Thanks. Some questions:

1. Do you use EventBus in multithreaded environment ? (Send events in one thread, process in another)
2. Do you call process() in multiple threads?

[penfree]

Hello, could you send more ? I mean parts of your code? I will try to check again this part of code. Thanks. Some questions:

1. Do you use EventBus in multithreaded environment ? (Send events in one thread, process in another)
2. Do you call process() in multiple threads?

1. yes, i use it in multithreaded environment.
2. yes, every threads postpone and process immediately。

the code is almost the same as readme, except it is called in multithread environment.

[gelldur]

Overall there should be only 1 thread that consumes ( calls process() ). You can send events from any thread ( calling postpone(Event) ). I will try write test for your use-case. I think I didn't tested it like that.

Usually you have 1 consumer thread for example main thread that will process events and call your listeners. Thanks to that you always know on what thread you listeners would be executed.

[penfree]

Overall there should be only 1 thread that consumes ( calls process() ). You can send events from any thread ( calling postpone(Event) ). I will try write test for your use-case. I think I didn't tested it like that.

oh, that is not easy to do. so i should use a cycle thread to consume the events?

[gelldur]

Ok nvm, I will create test for this. Maybe there is a bug.

In source code I see there is already mutex for "single process"

std::lock_guard writeGuardProcess{_mutexProcess}; // Only one process at the time

[penfree]

Ok nvm, I will create test for this. Maybe there is a bug.

In source code I see there is already mutex for "single process"

std::lock_guard writeGuardProcess{_mutexProcess}; // Only one process at the time

emm, i saw that too, so maybe mutex is missed somewhere. i found that when call frequency is slow, it works well

[gelldur]

Thanks, I will find issue and fix it soon (I will try this weekend). My best guess for now that there is some data race between process() and postpone()

[penfree]

@gelldur is there any progress?

[gelldur]

Sorry didn't have time. I will try soon. If you find solution It would be nice to merge your PR :)