Re: memory leak

[prinzinger]

Thank you Muthiah for your reply!

I was mistaken and looked for the problem at the wrong spot,
but in another replay John Eaton has figured out the problem of my code and now 
it works.

That means, apparantly, that by now the usage of exceptions with Octavecode 
works.

Thanks for helping!

Michael

-----Ursprüngliche Nachricht-----
Von: "Muthiah Annamalai" <address@hidden>
Gesendet: 24.05.07 03:08:00
An: "Michael Prinzinger" <address@hidden>
CC: address@hidden
Betreff: Re: memory leak


On 5/23/07, Michael Prinzinger <address@hidden> wrote:
> Good Day!
>
> I have a problem with Octave potentially causing memory leaks.
> I use liboctave in a C++ project with matrices.
> The problem occurs, when I try to read a matrix, which I have saved 
> previously.
>
> Please have a look on the following code:
>
> FILE* eigenmatrix_file = fopen(filename, "w");
> if(eigenmatrix_file == NULL)
> {
>      throw fileException("Error! in EigenMatrix::save(): can't open file!");
> }
>
> MArray2<Complex>* m_array = new MArray2<Complex>(this->row, this->column, 
> *(new Complex(0.0, 0.0)));
> fread(m_array, sizeof((*m_array)(0)), sizeof(m_array), eigenmatrix_file);
>
>
> This code runs fine with small matrices, and I have no problems whatsoever.
> However after I try it with "real" matrices coming from 80x80 images I want 
> to work with (80x80 = 6400 values per image, that means 6400 x 6400=40960000 
> values or the matrix),
> I get a segmentation fault accessing "eigenmatrix_file" after I created 
> m_array.
> Fidling with the above code, I noticed that even a simple call like
>
> eigenmatrix_file;
> or
> (eigenmatrix_file == NULL);
>
> causes a seg fault after the creation of m_array.
>
> I tried decalring m_array at the very beginning, before I declare and 
> initialize File* eigenmatrix_file, but even in this case, calling the 
> resulting pointer with the statements above (so no dereferencing, just 
> calling the pointer!) causes again a seg fault.
>
> If I outcomment
> // MArray2<Complex>* m_array = new MArray2<Complex>(this->row, this->column, 
> *(new Complex(0.0, 0.0)));
> I don't get this seg fault.
>
> So I have come to the conclusion, that creating
> a new MArray2<Complex>(this->row, this->column, *(new Complex(0.0, 0.0)));
> where row and column are pretty large,
> causes Octave to somehow destroy my C++ stack, at least my previously defined 
> variables get lost.
>
> Note that in general Octave has no problem with large matrices in my programm.
> If the file "filename" does not exist the corresponding matrix is computed 
> and saved without error, only after trying to load a matrix this large, this 
> error occurs.
>
> I hope you can help, I really need to solve this one, to be able to hand in 
> my university project ;)
> Thank You
>
> Michael Prinzinger
> _______________________________________________________________
> SMS schreiben mit WEB.DE FreeMail - einfach, schnell und
> kostenguenstig. Jetzt gleich testen! http://f.web.de/?mc=021192
>
> _______________________________________________
> Help-octave mailing list
> address@hidden
> https://www.cae.wisc.edu/mailman/listinfo/help-octave
>
AFAIK you cannot use exception handling in Octave. If you searched
the mailing lists, JWE has explained many times that Octave was written
at a time C++ didnot support exceptions!

So the code in your file handler must be written to set error_state if required,
and should exit by a statement like:

return octave_value("cannot open file ");

instead of the throw.

I think this could fix the problem. Atleast prevent the segfaults.
-Muthu



_______________________________________________________________
SMS schreiben mit WEB.DE FreeMail - einfach, schnell und
kostenguenstig. Jetzt gleich testen! http://f.web.de/?mc=021192