Re: [Boost-users] [asio] reopening of serial port fails - why?

Subject: Re: [Boost-users] [asio] reopening of serial port fails - why?
From: Markus Werle (numerical.simulation_at_[hidden])
Date: 2009-03-06 07:57:06

Bill Somerville <bill <at> classdesign.com> writes:

> I think you should simplify your close semantics, passing an error code 
> to your do_close() method is making it do more work than necessary.

I am an absolute beginner with asio.
I read through the docs and the examples, but obviously missed 
some points.

The example I presented is stolen from the web from 
http://www.nabble.com/Simple-serial-port-demonstration-with-boost-asio-
asynchronous-I-O-td19849520.html

> I don't understand the recreation of the serial_port object although I 
> have not used the serial_port functionality myself. Also the close() 
> method posts a do_close() call with a default constructed error object 
> which is not handled correctly.

In my GUI application I have 2 buttons: connect/disconnect.
I simply want them to work and I probably want to switch ports during runtime.

So here is my nitty-gritty task:
Write a class with the following interface

class SerialLineCommunicator
{
public:
Connect (std::string const & PortName);
Disconnect();
WriteToSerialPort(std::string const & Text);
};

This class should write all received characters to std::cout.

The problem I have is: The Disconnect() method seems to have some problems
which I was trying to track down, but I got lost. 

Neither a close() nor a delete of thread or communicator resolves
the issue.

I found that the demo code I had stolen from the internet 
obviously has the same problem if adopted such that
it terminates the connection and then retakes it.
I have not found any other working demo code and the docs
are not helpful to me either.

> I'm not sure why you need to use a loop in the main routine, if you want 
> a server then use pending async calls to keep the io_service alive, if 
> you want a client then perhaps errors should simply cause graceful 
> termination.

Can you probably provide some code that *should* work? 
I have no idea anymore.

Thanks,

Markus