deadlock when working with 2 event loops, asio-grpc loop never yeilds to foreign loop about asio-grpc

[Tradias]

I cannot see anything wrong with the code snippet. The deadline is too high indeed but I do not see any potential for overflow in the code. I suspect that:

• There is some UB or ODR violation elsewhere in your code. These errors are difficult to debug, I recommend commenting out as much code as possible until the error disappears, then slowly enable it again.
• There might be some compiler optimization that skews the value. I recommend looking at ioc.stopped() which should be false. If it is not then somewhere in the code you do ioc.stop() or ioc.get_executor().on_work_finished().
• Run the contexts on separate threads, see example.

[RavikumarTulugu]

ok let me disable the optimization and see if the value is still huge.

[RavikumarTulugu]

wanted to wait till i am sure the issue is resolved. This works better since grpc doesn't give out epoll fd. poll () comes to rescue.

static void
grpcPollTimerCallback (
const boost::system::error_code &error,
boost::asio::deadline_timer *timer,
boost::asio::io_context &ioc,
agrpc::GrpcContext &grpc_context
) {
if ( ! error ) {
timer -> expires_from_now ( boost::posix_time::milliseconds ( 1 ) );
timer -> async_wait ( boost::bind ( grpcPollTimerCallback,
boost::asio::placeholders::error,
timer,
std::ref ( ioc ),
std::ref ( grpc_context )));
grpc_context.poll ();
}
return;
}

boost::asio::deadline_timer grpcPollTimer ( ioc );
grpcPollTimer.expires_from_now ( boost::posix_time::milliseconds ( 1 ) );
grpcPollTimer.async_wait ( boost::bind ( grpcPollTimerCallback,
boost::asio::placeholders::error,
&grpcPollTimer,
std::ref ( ioc ),
std::ref ( grpc_context_ )));

[RavikumarTulugu]

There is a bug for sure in case of single thread with multiple event loops , but we are able to get around with the poll method , the poll approach works for now. may be in future , some one else will stumble up on the same in a different scenario. closing this for now.