-1

I had run into a strange problem, where I used to get an exception at the run-time with the message:
Unhandled exception at 0x0137d451 in I2cTlmTest.exe: 0xC0000005: Access violation reading location 0xcdcdcdd5.

I get this message during creation of object 'C'. I have debugged it but couldn't do much because the values of the (pointer) variables (and the this pointer) are set to (0xcdcdcdcd) as soon as the debug hits the break point in B::getHelper function of class B. I looked around for the possible culprits, thinking that the variables might not be initialized or might be getting deleted before i reach the function. But in the A::getHelper function all the variables are initialized and vanish when I enter the B::getHelper function and are set to magic values.

After a bit of trial debugging I just moved the default-inline A::getHelper to A.cpp. And to my wonder my application started running. I have solved my problem, but don't know why the inline function in A is causing the trouble. Any help is highly appreciated. Sorry for the long question but didn't have any other choice.

I have tried to create a similar scenario with the following code snippet as mine : 

-----classA.h---------
#include"classB.h"

class A
{
public:
      // if the definition of `getHelper()` is moved to A.cpp runs fine. But, why ? 
  Helper* getHelper() const { return ptrB_->getHelper(); } 

  A()
  {
     ptrB_ = new B(this);
  }
private:
    B *ptrB_;
};

----classB.h-----------
class A;

class Helper
{
....
....
};
class B
{

public:
  B(A *ptr):ptrA_(ptr){ helper_ = new Helper(); }
  Helper* getHelper() const { return helper_; }
private:
  A *ptrA_;
  Helper *helper_;
};

-------classC.h--------

#include"classA.h"

class C
{
public:
  C(A &ref): refA_(ref)
  {
     helper_ = refA_.getHelper();
  }

private:
  Helper *helper_;
  A &refA_;
};

------main.cpp---------
#include"classC.h"

int main()
{
  A *ptrA = new A();
  C *ptrC = new C(*ptrA);
  return 0;
}
Uchia Itachi
  • 5,287
  • 2
  • 23
  • 26
  • Can you give the stack trace for the error? – IdeaHat Nov 07 '13 at 17:54
  • 1
    http://stackoverflow.com/questions/370195/when-and-why-will-an-os-initialise-memory-to-0xcd-0xdd-etc-on-malloc-free-new sounds similar to me – willll Nov 07 '13 at 17:54
  • @MadScienceDreams: The call stack is simple, it calls the constructor of `C`, then A::getHelper( where all the values are initialized) and finally B::getHelper ( where it blows off) . – Uchia Itachi Nov 07 '13 at 18:03
  • Why don't you walk through your code with a debugger, line by line, investigating variables as you go? – SigTerm Nov 07 '13 at 18:06
  • @SigTerm: I have done that. The values of the variables are set to magic values when I reach "B::getHelper" function(including the `this` pointer). Until then all the variables have valid values. – Uchia Itachi Nov 07 '13 at 18:09
  • Your code compiles and runs without crashing. (g++ 4.8.1, microsoft compiler from 15.00). On my machine, at least. – SigTerm Nov 07 '13 at 18:11
  • 1
    [Here's the code I used.](http://pastebin.com/yGf8BHqw). The problem is elsewhere. Probably in code you haven't provided. – SigTerm Nov 07 '13 at 18:13
  • @UchiaItachi Do you get the same Access violation on your machine with the code linked to by SigTerm? By the way, what compiler (version and options) are you using to compile? Also, are you using precompiled headers? – gx_ Nov 07 '13 at 18:25

1 Answers1

1

helper_ is not pointing to a valid address.

You can do something like this :

B(A *ptr):ptrA_(ptr){ helper_ = new Helper(); }
willll
  • 1,809
  • 1
  • 16
  • 23