QSharedPointer crash in Linux

aliks-os

The code at below works fine in Windows, but crash in Linux. Please advice what I do wrong and how the problem can be fixed.
I get the following error message
free(): invalid pointer

class Process {

public:
   Process();
   MyStruct m_data;
};

Process::Process()
{
   QSharedPointer<MyStruct>  dataPtr = QSharedPointer<MyStruct>(&m_data);   <---Here crash
}

jeremy_k

Why are you creating a shared pointer to a class member variable?

From the documentation:
QSharedPointer will delete the pointer it is holding when it goes out of scope, provided no other QSharedPointer objects are referencing it.

When the two QSharedPointer instances in Process::Process() go out of scope at the end of the function, delete will be called on &m_data, which wasn't allocated with new.

aliks-os

@jeremy_k said in QSharedPointer crash in Linux:

Why are you creating a shared pointer to a class member variable?

So the correct approach will be as below, rigth?:

class Process {

public:
   Process();
   MyStruct *m_data;
};

Process::Process()
{
   m_data = new MyStruct();
   QSharedPointer<MyStruct>  dataPtr = QSharedPointer<MyStruct>(m_data);
}

jeremy_k

@aliks-os said in QSharedPointer crash in Linux:

@jeremy_k said in QSharedPointer crash in Linux:

Why are you creating a shared pointer to a class member variable?

So the correct approach will be as below, rigth?:

class Process {

public:
   Process();
   MyStruct *m_data;
};

Process::Process()
{
   m_data = new MyStruct();
   QSharedPointer<MyStruct>  dataPtr = QSharedPointer<MyStruct>(m_data);
}

That depends on the goal, but I would guess not. As is, the instance of MyStruct will be created and deleted in the Process constructor. At that point, m_data will be a dangling pointer. It might be easier to understand a simplified form:

MyStruct *dangling;
void f()
{
  QSharedPointer<MyStruct> dataPtr(new MyStruct);
  dangling = dataPtr.data();
  // MyStruct instance deleted here
  // dangling doesn't point to a valid object
}

aliks-os

@jeremy_k
yes you are right. Many thanks for your help. I have understood my mistake. In my case better to restruct the workflow of my code.
But I still wondering why the first version of my code works in windows but not in linux. )))

jeremy_k

Chance plus the choices made by the compiler and runtime sometimes appear to produce working programs for undefined behavior.