Another common mistake

int process_event (Event * evt)

{

Packet *p = evt->getPacket(1001);

if (s)

{

cout << setw(6) << s->iValue(0)

<< setw(6) << s->iValue(4)

<< setw(6) << s->iValue(5) << endl;

delete s;

}

return 0;

}

We have seen something like this...

No problem here.

int process_event (Event * evt)

{

Packet *s1 = evt->getPacket(1001);

Packet *s2 = evt->getPacket(1002);

if (s1 && s2)

{

cout << setw(6) << s1->iValue(0)

<< setw(6) << s2->iValue(0)

<< endl;

delete s1;

delete s2;

}

return 0;

}

A problem here?

Indeed. On the right, the “if” becomes true only if both the s1 and s2 objects are created. They usually are, so usually there is no problem. However, we might have a corrupt event and only one object gets created, not the other one. Then the “delete” is never reached, and as we go on, we leave the object behind. This is a “bad” memory leak. It is bad because

• Those memory leaks occur only very infrequently, and are extremely hard to recognize (not to mention to find) in the development phase of a project.

• It takes a corrupt or otherwise non-standard events to occur, and in the test phase you often work with simulated data which are less likely to show this problem;

• the problem will show up only when there is a significant amount of data to process, and that’s usually when the production has already started.

int process_event (Event * evt)

{

Packet *s1 = evt->getPacket(1001);

Packet *s2 = evt->getPacket(1002);

if (s1 && s2)

{

cout << setw(6) << s1->iValue(0)

<< setw(6) << s2->iValue(0)

<< endl;

}

if (s1) delete s1;

if (s2) delete s2;

return 0;

}

the solution

Previous slide

Next slide

Back to first slide

View graphic version