Asynchronous I/O on a serial device

Asynchronous I/O on a serial device

[Kim Rostgaard Christensen]

Hi there

I am desperately trying to implement a protocol on top of a serial 
transport.

I am running Debian, and for the moment using /dev/ttyACM0 for reading. 
This works fine using standard the Text_IO package and In_File.
Only, when I need to write to the device It fails because It is in 
In_File mode - Obviously.
Reset on the file handle gives me a USE_ERROR exception, probably due to 
it being a special file - please correct me if I am wrong.

Do I need to implement this more system-near/specific, or is there some 
way I can do bi-directional communication like, for instance two file 
handles. (I tried Form => "shared=yes", and it also failed).

It is a usb device that emulates UART, and it seems to be contempt with 
just about any Baud setting.

/Kim

Re: Asynchronous I/O on a serial device

[Randy Brukardt]

"Kim Rostgaard Christensen" <krc@retrospekt.dk> wrote in message 
news:it0noa$ce3$1@munin.nbi.dk...
> Hi there
>
> I am desperately trying to implement a protocol on top of a serial 
> transport.
>
> I am running Debian, and for the moment using /dev/ttyACM0 for reading. 
> This works fine using standard the Text_IO package and In_File.
> Only, when I need to write to the device It fails because It is in In_File 
> mode - Obviously.
> Reset on the file handle gives me a USE_ERROR exception, probably due to 
> it being a special file - please correct me if I am wrong.

Stream_IO allows changing the mode of an open stream (specifically to handle 
this sort of problem). But whether that is allowed in your specific compiler 
I can't say -- you'll have to experiment.

                        Randy.

Re: Asynchronous I/O on a serial device

[Yannick Duchêne (Hibou57)]

Le Sat, 11 Jun 2011 23:46:47 +0200, Kim Rostgaard Christensen  
<krc@retrospekt.dk> a écrit:

> Hi there
Hi you :)

> I am desperately trying to implement a protocol on top of a serial  
> transport.
>
> I am running Debian, and for the moment using /dev/ttyACM0 for reading.  
> This works fine using standard the Text_IO package
Not good in my opinion (see later)

> and In_File.
> Only, when I need to write to the device It fails because It is in  
> In_File mode - Obviously.
> Reset on the file handle gives me a USE_ERROR exception, probably due to  
> it being a special file - please correct me if I am wrong.
I do not know about your device node, so I could not say if its writable  
or not, but note you also have Inout_File in Ada (do not know what it  
named Inout_File instead of In_Out_File, but that's another story) beside  
of In_File, so could perfectly have input and output on the file.

> Do I need to implement this more system-near/specific, or is there some  
> way I can do bi-directional communication like, for instance two file  
> handles. (I tried Form => "shared=yes", and it also failed).
I would not say more system specific, but rather more close to some  
implicit specifications here: I am not sure you have good reasons [*] to  
use Text_IO, and you may better have a file of some basic type, like bytes  
or characters (well, a file of the more atomic type your protocol defines).

[*] If something is missing we do not know, tell us.

-- 
“Syntactic sugar causes cancer of the semi-colons.”  [Epigrams on  
Programming — Alan J. — P. Yale University]
“Structured Programming supports the law of the excluded muddle.” [Idem]
“c++; /* this makes c bigger but returns the old value */” [Anonymous]

Re: Asynchronous I/O on a serial device

[Kim Rostgaard Christensen]

On 2011-06-12 07:23, Randy Brukardt wrote:
> "Kim Rostgaard Christensen"<krc@retrospekt.dk>  wrote in message
> news:it0noa$ce3$1@munin.nbi.dk...
>> Hi there
>>
>> I am desperately trying to implement a protocol on top of a serial
>> transport.
>>
>> I am running Debian, and for the moment using /dev/ttyACM0 for reading.
>> This works fine using standard the Text_IO package and In_File.
>> Only, when I need to write to the device It fails because It is in In_File
>> mode - Obviously.
>> Reset on the file handle gives me a USE_ERROR exception, probably due to
>> it being a special file - please correct me if I am wrong.
>
> Stream_IO allows changing the mode of an open stream (specifically to handle
> this sort of problem). But whether that is allowed in your specific compiler
> I can't say -- you'll have to experiment.
>
>                          Randy.
>
>
I'm pretty sure I tried Stream_IO and it gave me the same result - 
USE_ERROR, or something similar. The problem is that it is a special 
file. Reset works like a charm on regular files, both in Text and Stream_IO

/Kim

Re: Asynchronous I/O on a serial device

[Kim Rostgaard Christensen]

On 2011-06-12 07:42, Yannick Duch�ne (Hibou57) wrote:
> Le Sat, 11 Jun 2011 23:46:47 +0200, Kim Rostgaard Christensen
> <krc@retrospekt.dk> a �crit:
>
>> Hi there
> Hi you :)
Ca va ? ;-)

...
> I do not know about your device node, so I could not say if its writable
> or not, but note you also have Inout_File in Ada (do not know what it
> named Inout_File instead of In_Out_File, but that's another story)
> beside of In_File, so could perfectly have input and output on the file.
The node is writeable, everything works via minicom. A googling reveals, 
at lest to me, that Inout_File is placed solely in the Direct_IO 
package. I will try it out later today.
I will look forward to the Inout_File/In_Out_File tale, someday.

>> Do I need to implement this more system-near/specific, or is there
>> some way I can do bi-directional communication like, for instance two
>> file handles. (I tried Form => "shared=yes", and it also failed).
> I would not say more system specific, but rather more close to some
> implicit specifications here: I am not sure you have good reasons [*] to
> use Text_IO, and you may better have a file of some basic type, like
> bytes or characters (well, a file of the more atomic type your protocol
> defines).
My reasons for using Text_IO is that is a QnD way of getting the input 
protocol logic implemented, while I was fiddling around with the 
bi-directional part.
The protocol is line based ascii encoded base16 values, but the parser 
is not pretty as of this writing (the QnD part)
I agree that it should be changed to something more application specific.

I was suggested to use POSIX.IO, but this is, from my point of view too 
system specific for my primitive application.

/Kim

Re: Asynchronous I/O on a serial device

[Kim Rostgaard Christensen]

...
> The node is writeable, everything works via minicom. A googling reveals,
> at lest to me, that Inout_File is placed solely in the Direct_IO
> package. I will try it out later today.
No luck. Direct_IO apparently only works on seekable files and gives a 
use error as well.

/Kim

Re: Asynchronous I/O on a serial device

[Yannick Duchêne (Hibou57)]

Le Sun, 12 Jun 2011 09:49:04 +0200, Kim Rostgaard Christensen  
<krc@retrospekt.dk> a écrit:

> On 2011-06-12 07:42, Yannick Duchêne (Hibou57) wrote:
>> Le Sat, 11 Jun 2011 23:46:47 +0200, Kim Rostgaard Christensen
>> <krc@retrospekt.dk> a écrit:
>>
>>> Hi there
>> Hi you :)
> Ca va ? ;-)

Ouwaouaw, une grenouille (a froggy ;) ). Tu me connais d'où, dis moi, je  
veux tout savoir :) Sur un forum ? Sur l'autre c.l.a (celui des frenchy) ?

Le Sun, 12 Jun 2011 12:35:58 +0200, Kim Rostgaard Christensen  
<krc@retrospekt.dk> a écrit:

> ...
>> The node is writeable, everything works via minicom. A googling reveals,
>> at lest to me, that Inout_File is placed solely in the Direct_IO
>> package. I will try it out later today.
> No luck. Direct_IO apparently only works on seekable files and gives a  
> use error as well.
>
> /Kim
I had a quick first idea I did not tell about, because I've never tried  
such a thing so far with Ada : could you open it twice, so as to get one  
stream for input and another stream for output ?

I may have a try on a tty (should be close enough) later today or tomorrow  
(but can't promise anything).

-- 
“Syntactic sugar causes cancer of the semi-colons.”  [Epigrams on  
Programming — Alan J. — P. Yale University]
“Structured Programming supports the law of the excluded muddle.” [Idem]
“c++; /* this makes c bigger but returns the old value */” [Anonymous]

Re: Asynchronous I/O on a serial device

[Kim Rostgaard Christensen]

On 2011-06-12 13:04, Yannick Duchêne (Hibou57) wrote:
>> Ca va ? ;-)
>
> Ouwaouaw, une grenouille (a froggy ;) ). Tu me connais d'où, dis moi, je
> veux tout savoir :) Sur un forum ? Sur l'autre c.l.a (celui des frenchy) ?

Je regrette de dire qui n'ont jamais eu le plaisir (or something like that)
No, I'm just a Dane using every excuse he gets, for using his limited 
(and very rusty) French.

> I had a quick first idea I did not tell about, because I've never tried
> such a thing so far with Ada : could you open it twice, so as to get one
> stream for input and another stream for output ?
>
> I may have a try on a tty (should be close enough) later today or
> tomorrow (but can't promise anything).

I'm pretty sure I also went down that road. It raised a USE_ERROR and I 
tracked it down to a paragraph in s-fileio.adb where it raises an error 
if the shared=yes is not set on the file.
As previously mentioned this did not help :-\

/Kim

Re: Asynchronous I/O on a serial device

[anon]

To connect to the UART device directly you should use to use the 
Gnat.Sockets.* or AdaSockets packages.

Or you could use PolyORB packages which contains a number of serial
I/O packages. Like:  GNAT.Serial_Communications;


Download => Libre.Adacore.com


In <it0noa$ce3$1@munin.nbi.dk>, Kim Rostgaard Christensen <krc@retrospekt.dk> writes:
>Hi there
>
>I am desperately trying to implement a protocol on top of a serial 
>transport.
>
>I am running Debian, and for the moment using /dev/ttyACM0 for reading. 
>This works fine using standard the Text_IO package and In_File.
>Only, when I need to write to the device It fails because It is in 
>In_File mode - Obviously.
>Reset on the file handle gives me a USE_ERROR exception, probably due to 
>it being a special file - please correct me if I am wrong.
>
>Do I need to implement this more system-near/specific, or is there some 
>way I can do bi-directional communication like, for instance two file 
>handles. (I tried Form => "shared=yes", and it also failed).
>
>It is a usb device that emulates UART, and it seems to be contempt with 
>just about any Baud setting.
>
>/Kim

Re: Asynchronous I/O on a serial device

[Kim Rostgaard Christensen]

On 2011-06-12 23:01, anon@att.net wrote:
> To connect to the UART device directly you should use to use the
> Gnat.Sockets.* or AdaSockets packages.
>
> Or you could use PolyORB packages which contains a number of serial
> I/O packages. Like:  GNAT.Serial_Communications;

The GNAT.Serial_Communications is just what I needed, although It gives 
some extra conversion overhead.

Can anyone give an example on how to use Gnat.Sockets/AdaSockets for 
accessing a UART device? I have only used it for IPv4 networking and 
Bluetooth.

Best
Kim

Re: Asynchronous I/O on a serial device

[Simon Clubley]

On 2011-06-14, Kim Rostgaard Christensen <krc@retrospekt.dk> wrote:
> On 2011-06-12 23:01, anon@att.net wrote:
>> To connect to the UART device directly you should use to use the
>> Gnat.Sockets.* or AdaSockets packages.
>>
>> Or you could use PolyORB packages which contains a number of serial
>> I/O packages. Like:  GNAT.Serial_Communications;
>
> The GNAT.Serial_Communications is just what I needed, although It gives 
> some extra conversion overhead.
>

You may wish to read through a thread from about 2 months ago about this
package in which it was discovered that it is forcing RTS/CTS flow control.

This will cause you problems unless your device is using RTS/CTS flow
control.

Simon.

-- 
Simon Clubley, clubley@remove_me.eisner.decus.org-Earth.UFP
Microsoft: Bringing you 1980s technology to a 21st century world