Boost exception "bad function call"

[ferchor2003]

When my application starts receiving and handling a few hundred calls the ios.run() method will throw a "bad function call" exception and refuse to reconnect the socket. This of course means the flow of asterisk events to the application stop and with that all my calls.
How can I locate which is the "bad function"? I understand this is boost telling me that execution has reached a function object which is empty.
Dev Environment:
Visual Studio 2015 with Boost 1_75_0. amd64
Deployment Environment:
Windows Server 2019 Datacenter, 64 bit
16 MB Ram on AWS m5.xlarge

This is the main loop of the std::thread running the ios.run():

boost::asio::io_service ios;

do {
	try {
		client = make_unique<aricpp::Client>(ios, astHost, astPort, astUser, astPassword,
			pPlugin->getAstAppName());
		Subscribe(*client.get());

		client->Connect([this, &astHost, &astPort](boost::system::error_code e) {
			if (e)
			{
				pPlugin->LogError(	"Asterisk Connection at %s:%s Failed. Error: %s ", astHost.c_str(), astPort.c_str(), e.message().c_str());
				calls.clear();
				this->ios.restart();
				std::this_thread::sleep_for(1s);
			}
			else
			{
				pPlugin->LogInfo("Asterisk Connection at %s:%s succeeded", astHost.c_str(), astPort.c_str());
				bIsInitialized = true;
			}
		}, 15);	// Ping the web socket every 15 seconds

		ios.run();	

		pPlugin->Loginfo("Asterisk ios connection has terminated");
		if (bIsInitialized)
		{
			// We should try re-connecting to asterisk box
			ios.restart();
			this_thread::sleep_for(1s);
		}
	}
	catch (exception& e)
	{
               // This is where the "bad function call" comes from
		pPlugin->LogError("Asterisk ios connection exception: %s.", e.what());
		client.release();
		std::this_thread::sleep_for(1s);
		ios.restart();
	}
} while (bIsInitialized);

[daniele77]

Hi @ferchor2003 ,

first of all, thank you for using aricpp library.
If you want to support the work on this project, you can consider sponsoring me using github sponsor.

Your application is catching a std::bad_function_call. It's thrown when you call an empty std::function.
I suggest you use a debugger to find which std::function is empty (if you cannot use a debugger you can add a trace before every call to std::function::operator() to find the issue.

Let me know and -- if it's a library issue -- I will fix it.

[ferchor2003]

Thanks @daniele77
I have collected some more detailed logs of what's going on when the error happens. As I said this is only after a hard load, with more than 100 simultaneous calls to the servers sustained for at least 5 minutes.
Apparently when the websocket goes down and the code tries to restart the ios object and recreate the Client it is not enough. After that the next call to ios.run() simply passes (like ios has nothing more to do) and the websocket cannot be re-created to get messages again.

+++ [5-10-2021 23:47:18:489Z] [WARNING] 
webSocket::Close exception: A request to send or receive data was disallowed because the socket had already been shut down in that direction with a previous shutdown call

+++ [5-10-2021 23:47:18:489Z] [AUDIT] [EVENTLOG]
Exception processing Asterisk events: bad function call.
Unable to Initialize Asterisk Plugin

+++ [5-10-2021 23:47:18:489Z] [WARNING] 
webSocket::Close exception: The I/O operation has been aborted because of either a thread exit or an application request

+++ [5-10-2021 23:47:18:489Z] [WARNING] 
webSocket::Close exception: The I/O operation has been aborted because of either a thread exit or an application request

+++ [5-10-2021 23:47:18:489Z] [WARNING] 
webSocket::Close exception: cancel: The file handle supplied is not valid

[daniele77]

I tested the library using SIPP, with an aricpp application that:

1. answer
2. play a sound file
3. hangup

(I will commit the example in the repository on the next release).

With 200 simultaneous calls, the application works fine. With > 200 calls, asterisk starts to complain because of the limited file descriptors.

So, I cannot understand why your application got a "bad function call" in the first place, and after that, why you're not able to restart asio::io_context. I guess it's a problem with your application, not the library. On the other hand, it's nearly impossible to guess what's the problem without the complete application sources (or at least a minimal example that causes the same issue).

[ferchor2003]

Thanks for the time spent on this issue Danielle.
I have found in my application a problem with the state machine that causes some calls to not hangup the call. Fixing that makes the whole thing much more stable and does not cause the "bad call" problem. The asterisk server I'm using is upped to 10k file descriptors so no issue with asterisk on that respect.