Auto-forwarding now waits for a result when using blocking mode before returning

[FletcherJ1]

I've added the wait method to the cmd_result class because calling get won't work due to how the presult variable is instantiated.

NuRaft/src/handle_user_cmd.cxx

Lines 168 to 170 in 17aab8d

	ptr< cmd_result< ptr<buffer> > >
	presult( cs_new< cmd_result< ptr<buffer> > >() );

When presult is made, it calls a constructor that sets has_result_ to true. Then, if get() was used it would see that has_result_ is true and return immediately and not wait for the set_result call. Therefore, the explicit wait method will guarantee the current thread to wait until a call to set_result happens.

[greensky00]

@FletcherJ1
Thanks for submitting the new PR.

I get your point regarding has_result_ value. But my understanding is that the code that you are referring to will call the default constructor (i.e., cmd_result()) as there is no parameter given for cmd_result: cs_new< cmd_result< ptr<buffer> > >()

I just checked with GDB and saw that has_result_ value is false.

Thread 1 "asio_service_te" hit Breakpoint 1, nuraft::raft_server::send_msg_to_leader (this=0x555555c65bf0,
    req=std::shared_ptr<nuraft::req_msg> (use count 1, weak count 0) = {...})
    at /home/junahn/sandbox/nuraft/src/handle_user_cmd.cxx:187
187	    rpc_cli->send(req, handler);
(gdb) p presult->has_result_
$1 = false
(gdb) p presult->result_
$2 = std::shared_ptr<nuraft::buffer> (empty) = {get() = 0x0}
(gdb)

If that's the case, we can just use the existing get() function. Could you please double-check? Thanks!

[FletcherJ1]

Ah yes, you're correct. I've swapped to using get and it works!

[greensky00]

Merged, thanks!