Hello guys,

I need some help regarding a (very) basic Event handling framework I
just started on today.
Please note that I'm still fairly new to programming itself and that
I might not know things that seem obvious to others.

The attached file is my work so far. I have a class called Backbone which houses 2 pointers for a linked list. Every linked list member contains 2 private pointers (one forward and one backward), the associated public functions to acces them, a string for the Event_ID and a function pointer to the function that shall be called, if an event with the corresponding Event_ID is invoked.

I also have a simple test class with exactly one function to test whether the function is indeed called.



I can compile without problems, but I get a runtime error "Access Violation: Read at address 0x4E4". This runtime error gets thrown once I call the function getNext() through a pointer, but I honestly have no idea why.

I am using D1 and Tango, though Tango is only for the line printing and can easily be converted to Phobos.

I hope you can help me!

You escape pointer to stack allocated object.

Why can't I do that?

What can I do to circumvent this?

On 2011-11-19 21:51, Tano Sakaria wrote:
> Why can't I do that?
>
> What can I do to circumvent this?

Allocate the object on the heap using "new".

-- 
/Jacob Carlborg

I'm sorry, but I don't understand.

How can I use "new" in conjunction with a pointer? "new" creates an object, but don't I need a pointer to said object instead for my linked list?

On 11/19/2011 01:22 PM, Tano Sakaria wrote:
> I'm sorry, but I don't understand.
>
> How can I use "new" in conjunction with a pointer? "new" creates an
> object, but don't I need a pointer to said object instead for my
> linked list?

Without seeing the original code, foo() below returns a pointer to an object allocated on the heap:

struct S
{}

S * foo()
{
    S * p = new S;
    return p;
}

void main()
{
    S * p = foo();
}

That object belongs to the D runtime and may be destroyed at an unspecified time when there are no more references to it.

Ali

When I do that, I get an error "Cannot implicitly convert type HandlerLink to HandlerLink*".

By the way, how can I split up my classes into seperate source files?
Just copying the code into one file per class and changing the imports
doesn't work. I tried compiling the classes seperately, into libraries
and what not, but all I get is Linker-errors as soon as compiling the
Backbone class.
Setting the -I flag for the import path does not work, either.

I could not figure out how to see Backbone.d, so I cannot see your code.

On 11/20/2011 01:52 AM, Tano Sakaria wrote:
> When I do that, I get an error "Cannot implicitly convert type
> HandlerLink to HandlerLink*".

That error must have been changed since D1. I get a similar one when I return a pointer to a 'class' variable as opposed to a 'struct' variable.

If you are really using a class, then you don't need a pointer:

class S
{}

S foo()
{
    /*
     * The left hand side is a class variable, while the
     * anonymous object that has been created with new on the
     * right hand side is a clas object. Class objects are
     * accessed by class variables.
     *
     * They are different things. They are confusing because
     * the object is defined as 'class S{}' earlier and
     * its reference variable is defined as 'S var'. (I think
     * this is the same in Java and perhaps C#.)
     *
     * Returning the class variable from the function is fine
     * because the actual object lives in the heap.
     */
    S var = new S;
    return var;
}

void main()
{
    S var = foo();
}

Please provide a smaller code in plain text. (Some people put their code on a clipboard-like code site. You can do that too if the code is too long.)

> By the way, how can I split up my classes into seperate source files?

(Note: It would be better if this question had its separate thread.)

The following is on Linux. I think the object file extensions are .obj on Windows.

> Just copying the code into one file per class and changing the imports
> doesn't work.

You must give those files to the compiler as well. Assuming the program has the main module (deneme.d) and two others:

$ dmd deneme.d hayvan/kedi.d hayvan/kopek.d

> I tried compiling the classes seperately,

That works too. Assuming that the two modules have been compiled with the -c flag to generate their .o files:

$ dmd deneme.d hayvan/kedi.o hayvan/kopek.o

> into libraries

That works too. First make a library that consists of the modules:

$ dmd hayvan/kedi.d hayvan/kopek.d -lib -ofhayvan

Then use that library when building the program:

$ dmd deneme.d hayvan.a

Most of the above are from one of the not-yet-translated chapters of D.ershane:

  http://ddili.org/ders/d/moduller.html

> and what not, but all I get is Linker-errors as soon as compiling the
> Backbone class.
> Setting the -I flag for the import path does not work, either.

-I is for the location of the import files; it doesn't involve compilation.

Ali

Thanks a lot, it works now!