comp.lang.ada
 help / color / mirror / Atom feed
* alternative for buggy accept_socket to create a non-blocking version
@ 2017-10-25 17:23 Matt Borchers
  2017-10-25 19:14 ` Matt Borchers
                   ` (3 more replies)
  0 siblings, 4 replies; 7+ messages in thread
From: Matt Borchers @ 2017-10-25 17:23 UTC (permalink / raw)


I have no choice but to use an old GNAT compiler -- version 6.3.0.  In this version there is a confirmed bug (fixed in 6.3.2) that prevents the non-blocking version of accept_socket from working.  I want to call the version of accept_socket with the 'timeout' parameter, but it doesn't do what it should.

Basically, I've built a way to read data from a socket and all goes well when a response is made.  However, I need a way to abandon the accept call if a response is never sent within a small time window.

This is my foray into sockets programming and I don't fully understand all of the nuances yet.  I tried setting the Receive_Timeout socket option but it had no effect -- I don't think it does what I need anyway.  This all seems pretty straight-forward assuming there is no bug in the non-blocking version of accept_socket, but I don't know how to work-around this bug.

I can imagine a pair of tasks in which one does the blocking accept (A) and the other task (B) does the time-out.  If the time-out occurs first, B will abort the task A otherwise after A succeeds in accepting the message it will abort B.  This seems like it might not be the best work-around solution.

Does anybody know of a good way to create a type of non-blocking accept_socket?

My code basically is:

with GNAT.Sockets; use GNAT.Sockets;
...
declare
  sock,
  socket   : SOCKET_TYPE;
  address  : SOCK_ADDR_TYPE;
  data     : STREAM_ELEMENT_ARRAY(1..256);
  last     : STREAM_ELEMENT_OFFSET;
  sel_stat : SELECTOR_STATUS;
begin
  create_socket( socket );
  bind_socket( socket, address );
  listen_socket( socket );
  accept_socket( socket, sock, address );
  --I wish I could do this
  --accept_socket( socket, sock, address, timeout => 2.0, status => sel_stat );
  ...
  loop
    receive_socket( sock, data, last );
    ...
  end loop;
  close_socket( socket );
end;

Thanks,
Matt

^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: alternative for buggy accept_socket to create a non-blocking version
  2017-10-25 17:23 alternative for buggy accept_socket to create a non-blocking version Matt Borchers
@ 2017-10-25 19:14 ` Matt Borchers
  2017-10-26 13:43   ` Shark8
  2017-10-25 19:18 ` Dmitry A. Kazakov
                   ` (2 subsequent siblings)
  3 siblings, 1 reply; 7+ messages in thread
From: Matt Borchers @ 2017-10-25 19:14 UTC (permalink / raw)


Well, I should have thought of this solution sooner.  I downloaded the GNAT 6.3.2 version (where the bug is fixed) and I copied all of the execution path of the non-blocking accept_socket subprogram into a single subroutine in a new child package of GNAT.Sockets.

My first few tests are successful.

^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: alternative for buggy accept_socket to create a non-blocking version
  2017-10-25 17:23 alternative for buggy accept_socket to create a non-blocking version Matt Borchers
  2017-10-25 19:14 ` Matt Borchers
@ 2017-10-25 19:18 ` Dmitry A. Kazakov
  2017-10-27  9:14 ` Björn Lundin
  2017-10-27 16:26 ` Daniel Norte Moraes
  3 siblings, 0 replies; 7+ messages in thread
From: Dmitry A. Kazakov @ 2017-10-25 19:18 UTC (permalink / raw)


On 2017-10-25 19:23, Matt Borchers wrote:

> I have no choice but to use an old GNAT compiler -- version 6.3.0.
> In  this version there is a confirmed bug (fixed in 6.3.2) that prevents the
> non-blocking version of accept_socket from working.

Never had problems with that.

> Does anybody know of a good way to create a type of non-blocking
> accept_socket?
Set Receive_Timeout?

-- 
Regards,
Dmitry A. Kazakov
http://www.dmitry-kazakov.de

^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: alternative for buggy accept_socket to create a non-blocking version
  2017-10-25 19:14 ` Matt Borchers
@ 2017-10-26 13:43   ` Shark8
  0 siblings, 0 replies; 7+ messages in thread
From: Shark8 @ 2017-10-26 13:43 UTC (permalink / raw)


On Wednesday, October 25, 2017 at 1:14:47 PM UTC-6, Matt Borchers wrote:
> Well, I should have thought of this solution sooner.  I downloaded the GNAT 6.3.2 version (where the bug is fixed) and I copied all of the execution path of the non-blocking accept_socket subprogram into a single subroutine in a new child package of GNAT.Sockets.
> 
> My first few tests are successful.

You could try using a Generic Package taking a TIMEOUT formal parameter (and if you're using task-interfaces, an interface to your socket-task) then simply have one task do your processing and the other send an abort to the other task if/when the timeout occurs.

^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: alternative for buggy accept_socket to create a non-blocking version
  2017-10-25 17:23 alternative for buggy accept_socket to create a non-blocking version Matt Borchers
  2017-10-25 19:14 ` Matt Borchers
  2017-10-25 19:18 ` Dmitry A. Kazakov
@ 2017-10-27  9:14 ` Björn Lundin
  2017-10-27 23:46   ` Matt Borchers
  2017-10-27 16:26 ` Daniel Norte Moraes
  3 siblings, 1 reply; 7+ messages in thread
From: Björn Lundin @ 2017-10-27  9:14 UTC (permalink / raw)


On 2017-10-25 19:23, Matt Borchers wrote:

> Does anybody know of a good way to create a type of non-blocking accept_socket?

put the socket in select (the socket function - not tasking) with a
timeout of 2 secs (in the readable set)
if select times out, you get 0 as retrun value,
 if select returns > 0 you have someone trying to connect.
You can then call accept and know it will not be blocking.

You may have to pragma import select
(And I hope it works with gnat.sockets)

looking into
<https://www2.adacore.com/gap-static/GNAT_Book/html/rts/g-socket__ads.htm>

  procedure Check_Selector

seem to be exactly this



> 
> My code basically is:
> 
> with GNAT.Sockets; use GNAT.Sockets;
> ...
> declare
>   sock,
>   socket   : SOCKET_TYPE;
>   address  : SOCK_ADDR_TYPE;
>   data     : STREAM_ELEMENT_ARRAY(1..256);
>   last     : STREAM_ELEMENT_OFFSET;
>   sel_stat : SELECTOR_STATUS;
> begin
>   create_socket( socket );
>   bind_socket( socket, address );
>   listen_socket( socket );
>   accept_socket( socket, sock, address );
>   --I wish I could do this
>   --accept_socket( socket, sock, address, timeout => 2.0, status => sel_stat );
>   ...
>   loop
>     receive_socket( sock, data, last );
>     ...
>   end loop;
>   close_socket( socket );
> end;
> 
> Thanks,
> Matt
> 


-- 
--
Björn

^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: alternative for buggy accept_socket to create a non-blocking version
  2017-10-25 17:23 alternative for buggy accept_socket to create a non-blocking version Matt Borchers
                   ` (2 preceding siblings ...)
  2017-10-27  9:14 ` Björn Lundin
@ 2017-10-27 16:26 ` Daniel Norte Moraes
  3 siblings, 0 replies; 7+ messages in thread
From: Daniel Norte Moraes @ 2017-10-27 16:26 UTC (permalink / raw)


Em quarta-feira, 25 de outubro de 2017 15:23:52 UTC-2, Matt Borchers  escreveu:
> I have no choice but to use an old GNAT compiler -- version 6.3.0.  In this version there is a confirmed bug (fixed in 6.3.2) that prevents the non-blocking version of accept_socket from working.  I want to call the version of accept_socket with the 'timeout' parameter, but it doesn't do what it should.
> 
> Basically, I've built a way to read data from a socket and all goes well when a response is made.  However, I need a way to abandon the accept call if a response is never sent within a small time window.
> 
> This is my foray into sockets programming and I don't fully understand all of the nuances yet.  I tried setting the Receive_Timeout socket option but it had no effect -- I don't think it does what I need anyway.  This all seems pretty straight-forward assuming there is no bug in the non-blocking version of accept_socket, but I don't know how to work-around this bug.
> 
> I can imagine a pair of tasks in which one does the blocking accept (A) and the other task (B) does the time-out.  If the time-out occurs first, B will abort the task A otherwise after A succeeds in accepting the message it will abort B.  This seems like it might not be the best work-around solution.
> 
> Does anybody know of a good way to create a type of non-blocking accept_socket?
> 
> My code basically is:
> 
> with GNAT.Sockets; use GNAT.Sockets;
> ...
> declare
>   sock,
>   socket   : SOCKET_TYPE;
>   address  : SOCK_ADDR_TYPE;
>   data     : STREAM_ELEMENT_ARRAY(1..256);
>   last     : STREAM_ELEMENT_OFFSET;
>   sel_stat : SELECTOR_STATUS;
> begin
>   create_socket( socket );
>   bind_socket( socket, address );
>   listen_socket( socket );
>   accept_socket( socket, sock, address );
>   --I wish I could do this
>   --accept_socket( socket, sock, address, timeout => 2.0, status => sel_stat );
>   ...
>   loop
>     receive_socket( sock, data, last );
>     ...
>   end loop;
>   close_socket( socket );
> end;
> 
> Thanks,
> Matt

Or you can use lib Anet.

[]'s Dani.

^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: alternative for buggy accept_socket to create a non-blocking version
  2017-10-27  9:14 ` Björn Lundin
@ 2017-10-27 23:46   ` Matt Borchers
  0 siblings, 0 replies; 7+ messages in thread
From: Matt Borchers @ 2017-10-27 23:46 UTC (permalink / raw)


The bug I needed to work around was actually in the Check_Selector subprogram. That is in the call path of Accept_Socket with the timeout parameter. 

So, I actually ended up doing what you suggest, but needed a fixed version of Check_Selector.

^ permalink raw reply	[flat|nested] 7+ messages in thread

end of thread, other threads:[~2017-10-27 23:46 UTC | newest]

Thread overview: 7+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2017-10-25 17:23 alternative for buggy accept_socket to create a non-blocking version Matt Borchers
2017-10-25 19:14 ` Matt Borchers
2017-10-26 13:43   ` Shark8
2017-10-25 19:18 ` Dmitry A. Kazakov
2017-10-27  9:14 ` Björn Lundin
2017-10-27 23:46   ` Matt Borchers
2017-10-27 16:26 ` Daniel Norte Moraes

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox