Thread: Memory allocation without assignment?

First, the relevant source code can be downloaded from here (it was not written by me).

And the section of code that has me scratching my head can be found in line 350 of hmm.cpp:

Code:

349:	if (possibleSrc && possibleSrc->find((*it)->state())!=possibleSrc->end())
350:	  new Transition(*it, node, o);

The above portion of code is part of a function definition within a class definition. At first I thought it was just code that the author somehow forgot about, so I commented it out. When I did that, the results did not come out right. So it is doing something. I just have no idea what.

To give it a vague outline, it looks like this.

Code:

class A
{
public:
A(){}
};
class B
{
public:
B(){}
some_memfunc(){new A();} // HUH?????
};

I've been tracking the thing in gdb for hours now, and there is no light at the end of the tunnel.

Originally Posted by iMalc

2. It may not be a bug at all and may not leak. The constructor of the object might store the pointer to 'this' in some global list, so that the pointer can be freed later.

Yup, you were (kind of) right. I looked more closely (I guess my brain is fried right now) at the constructor for Transition and it's doing something quite funky:

Code:

Transition::Transition(HmmNode* from, HmmNode* to, unsigned long obs)
{
  _from = from;
  _to = to;
  _obs = obs;
  if (_from && _to) {
    _from->outs().push_back(this);
    _to->ins().push_back(this);
  }
}

It's getting pointers to some other data structure and then appending itself to them. Don't know why I didn't notice that before. Thanks.