I whipped up a quick little program to test some ideas I have been thinking of, and was rather dismayed when it segfaults when closing.

The source:

So all I essentially do is load two tiny sprites into an array (which works apparently). I do not draw them (yet), I just have them there. As soon as I press a key to leave, it segfaults.

I notice that if I change 'readkey()' to while(!key[KEY_ESC]){} the segfault seems to miraculously disappear.

Any reasons why?

Yeah, basically framesArray[] is an array of 'Frame', which has a BITMAP *bmpFrame; inside it.

edit:
So what the CAnimation::LoadFrames() does is load the frames with the required bitmaps. It's hardcoded and whatnot, but I was just testing out some ideas.

edit2:

I am using Windows XP Pro SP2 btw. Not sure if the OS would make a difference here...

Well, return 1; was changed from return 0; I was kind of flabergasted for a bit and started changing things to try and nail down the problem. That was actually why the whole 'clean()' methods were developed. CFrame has a clean method which basically checks to see if the bmpFrame is empty or not. If it is not empty then the bitmap is destroyed. CAnimation has a method of the same name which goes through each frame in it's frameArray calling the clean() method of each frame. The segfault was appearing before I added this, so I don't think it was relevant.

void CFrame::clean()
{
    if(bmpFrame) destroy_bitmap(bmpFrame);
}

void CAnimation::clean()
{
    int i=0;
    for(i;i<intNumFrames;i++)
    {
        frameArray<i>.clean();
    }
}

What I am doing is making an OOP animation system. There was a thread here about the whole thing.

Well, no. The clean() methods are supposed to mimic one. I remember using destroy_bitmap(blah) in destructors before and it caused segfaults when exiting as well.

I'll make a note to do that. I still would like to know what caused that segfault...

Well, when I commented out the loading of the images it did not crash. This confuses me because the images loaded fine as far as I can tell. I didn't actually draw them or even use them in any way.

I find it weird that when I uncomment the image loading and comment out the readkey() and replace with the conditional loop the problem disappears.

edit:
Perhaps these are the code demons my teacher was talking/joking about

hmm after doing what you said to do Kitty Cat an 'Application Error' occurs.

I'll upload a picture of my error. It is different then before, because before I would get the whole send/don't send error report. I would also have to manually kill the process from task manager before I could recompile before. Now the process seems to be taken out automatically.

frameArray is an array of Frames

class CAnimation
{
    private:
       CFrame frameArray[2];
    //...

}

It's actually been initialized with a value of two... I do so in the LoadFrames() method.

When will people realise that destroy_bitmap() deals fine with NULL?
And do set pointers to NULL after freeing them. It will prevent crashes from double frees or invalid pointer dereferences.

About your problem: make sure you don't call Allegro functions from global destructors (I didn't read the whole thread, so maybe this was proposed before).

Maybe it crashed freeing an unitialized BITMAP * that doesn't = NULL.