Re: [Boost-users] boost::asio - reason for asio_handler_invoke

From: mikpol (mike.polyakov_at_[hidden])
Date: 2008-01-08 09:44:33

mikpol wrote:
> 
> I've been trying to understand the reasons to have asio_handler_invoke
> function, and so far I am unable to understand why it is needed. It is
> used to override invocation policy for a function-like object based on the
> handler. boost::asio gives one example with using io_service::strand...
> 

I have finally understood the reason for asio_handler_invoke after reading
asio TR2 proposal carefully. It is due to the bound handlers invoked by the
library for functions such as async_read and async_read_until. They need to
call read function on the socket multiple times, and a handler created by
the library is needed to be called in the same way as the user provided
handler. I knew about this before, but I couldn't understand what was shared
between those handler's that needs to be protected from simultaneous
invocation by multiple threads. It turns out this is the socket object
itself, which is referenced from all these handlers and needs protection
from potential concurrent invocation.

If anyone finds a problem in this explanation or would like to comment on
other reasons why it is needed, please post something, I would be interested
in hearing your opinion.

Thanks.

-- 
View this message in context: http://www.nabble.com/boost%3A%3Aasio---reason-for-asio_handler_invoke-tp14653663p14691075.html
Sent from the Boost - Users mailing list archive at Nabble.com.