Gnat.Sockets.Stream memory "leak"

Gnat.Sockets.Stream memory "leak"

[Jano]

Okay, here I go.

I've spent today quite a time tracking down a memory leak. It has 
finally resulted it was exactly not a leak, only a misassumption on my 
part.

It happens that each call to Gnat.Sockets.Stream allocates a new stream 
object to return its access. 

My assumption was that, like the Ada.Streams.Stream_IO.Stream call, that 
Stream_access need not to be freed. Now, I see not any recommendation or 
note about the necessity of freeing these resources neither in Stream_IO 
nor in Gnat.Sockets. I'm mistaken about Stream_IO not needing the free? 
I've been naive about Gnat.Sockets silence about it?

And, as a side note, why doesn't exist a Stream_access type in 
Ada.Streams? I find somewhat inconvenient defining it in my packages and 
converting here and there.

Thanks!

-- 
-------------------------
Jano
402450.at.cepsz.unizar.es
-------------------------

Re: Gnat.Sockets.Stream memory "leak"

[@ 2003-07-02 9:50 UTC (permalink / raw)]

Jano wrote:

> And, as a side note, why doesn't exist a Stream_access type in 
> Ada.Streams? I find somewhat inconvenient defining it in my packages and 
> converting here and there.

My opinion is that they left this for each concrete implementation of 
streams. For instance, Ada.Streams.Streams_IO has the Stream_Access type 
and GNAT.Sockets has it as well. You do not have to define it in your 
own packages.

Rodrigo

Re: Gnat.Sockets.Stream memory "leak"

[Jano]

=?UTF-8?B?Um9kcmlnbyBHYXJjw61h?= <rodrigo.garcia.ARROBA.epfl.ch> dice...
> Jano wrote:
> 
> > And, as a side note, why doesn't exist a Stream_access type in 
> > Ada.Streams? I find somewhat inconvenient defining it in my packages and 
> > converting here and there.
> 
> My opinion is that they left this for each concrete implementation of 
> streams. For instance, Ada.Streams.Streams_IO has the Stream_Access type 
> and GNAT.Sockets has it as well. You do not have to define it in your 
> own packages.

That's okay, I'm thinking more about when I need to store an access 
somewhere (not merely pass it as an access parameter), being it 
completely irrelevant what is the precise accessed stream. Then I need 
to roll out my own stream_access to Root'Class.

-- 
-------------------------
Jano
402450.at.cepsz.unizar.es
-------------------------

Re: Gnat.Sockets.Stream memory "leak"

[@ 2003-07-04 10:23 UTC (permalink / raw)]

Jano wrote:
> =?UTF-8?B?Um9kcmlnbyBHYXJjw61h?= <rodrigo.garcia.ARROBA.epfl.ch> dice...
> 
>>Jano wrote:
>>
>>
>>>And, as a side note, why doesn't exist a Stream_access type in 
>>>Ada.Streams? I find somewhat inconvenient defining it in my packages and 
>>>converting here and there.
>>
>>My opinion is that they left this for each concrete implementation of 
>>streams. For instance, Ada.Streams.Streams_IO has the Stream_Access type 
>>and GNAT.Sockets has it as well. You do not have to define it in your 
>>own packages.
> 
> 
> That's okay, I'm thinking more about when I need to store an access 
> somewhere (not merely pass it as an access parameter), being it 
> completely irrelevant what is the precise accessed stream. Then I need 
> to roll out my own stream_access to Root'Class.
> 

I do not understand you here... Why do you want to store an access to a 
stream? And where (in memory or in disk)? To store it in memory you only 
have to create a variable of the Stream_Access type...

Can you give an example of what you mean?

Rodrigo

PS: I would not worry about having to free the memory taken for the 
stream access as long as you close the socket.

Re: Gnat.Sockets.Stream memory "leak"

[Jano]

=?UTF-8?B?Um9kcmlnbyBHYXJjw61h?= <rodrigo.garcia.ARROBA.epfl.ch> dice...
> I do not understand you here... Why do you want to store an access to a 
> stream? And where (in memory or in disk)? To store it in memory you only 
> have to create a variable of the Stream_Access type...

In memory, of course. I have some objects that operate over streams. I 
pass the stream in an access parameter, but to store it I don't care 
(nor know) if they are Stream_io.Stream_access or Sockets.Stream_access, 
etc so I can't use any of the preexistent Stream_access types. I need to 
define my own access to Root_stream_type'Class. I simply would find 
convenient that these root'Class access where defined where it all 
starts: in Ada.Streams. Since I must "with" that package to have access 
parameters to streams, it would spare me the definition.

I hope it's clearer now.

-- 
-------------------------
Jano
402450.at.cepsz.unizar.es
-------------------------

Re: Gnat.Sockets.Stream memory "leak"

[@ 2003-07-07 8:39 UTC (permalink / raw)]

Jano wrote:
> =?UTF-8?B?Um9kcmlnbyBHYXJjw61h?= <rodrigo.garcia.ARROBA.epfl.ch> dice...
> 
>>I do not understand you here... Why do you want to store an access to a 
>>stream? And where (in memory or in disk)? To store it in memory you only 
>>have to create a variable of the Stream_Access type...
> 
> 
> In memory, of course. I have some objects that operate over streams. I 
> pass the stream in an access parameter, but to store it I don't care 
> (nor know) if they are Stream_io.Stream_access or Sockets.Stream_access, 
> etc so I can't use any of the preexistent Stream_access types. I need to 
> define my own access to Root_stream_type'Class. I simply would find 
> convenient that these root'Class access where defined where it all 
> starts: in Ada.Streams. Since I must "with" that package to have access 
> parameters to streams, it would spare me the definition.
> 
> I hope it's clearer now.

Yes, I fully understand your problem now and, frankly, I do not really 
know why they did not define an access to streams type in Ada.Streams. 
Anyway, you can still use the one defined in Stream_IO, as it is an 
access to the Root_Stream_Type'Class (you can also assign "socket 
streams" or any other kind of stream to it).

Rodrigo