Thread: Dynamic memory allocation and runtime errors.

Hi,

I'm trying to write a program which involves storing a fair amount of data in arrays a therefore repeated calls of malloc (On the scale of a few MB, i'm nowhere near the 2GB/4GB limit). This works perfectly up until I try and build my third array (a little bigger than the others) at which point the program crashes. If I build the array with about a factor of ten less entries it works (until i get a crash later which i guess is because i'm no massively overrunning the array). Interestingly if i try to free any arrays at this point i also get crashes. All i can do in terms of memory allocation is build small arrays.

Here's a code snippet to give the idea of what i'm doing, i've checked that all of the array sizes are reasonable and aren't being screwed up by something before any malloc calls:

Code:

int main ()
{
...


    double* convlvData;
    
    double *x_input, *y_input, *x_inputPad, *y_inputPad, *y_input_cmplx, *y_input_fft;
    int len_input, len_input_cmplx, len_inputPad;
    
    double *x_resp, *x_respW, *y_resp, *y_respW, *y_resp_cmplx, *y_resp_fft;
    int len_resp, len_resp_cmplx; //len resp must be less than len_data, someone validate for this please.

...



            if (statusFlag != 0) statusFlag = wrap(x_resp, &x_respW, y_resp, &y_respW, len_resp, len_input, &DataPadRef);


            showMessage(" Padding Data... ", 1);


            if (statusFlag != 0) statusFlag = convlvDataPad(x_input, y_input, &x_inputPad, &y_inputPad, DataPadRef, len_input);

//The crash happens within convlvDataPad
...
}

int wrap(double* x_resp, double** x_respW, double* y_resp, double** y_respW, int len_resp, int len_input, int *DataPadRef)
{
...

    statusFlag = buildArray ( x_respW , (len_input) );
    statusFlag = statusFlag && buildArray ( y_respW , (len_input) );
...
}

//I.e Wrap contains some array creation.

int convlvDataPad(double* data_x, double* data_y, double** paddata_x, double** paddata_y, int DataPadRef, int len_input)
{
    ...
            
    statusFlag = buildArray ( paddata_x , padLen);
        
       //CRASH OCCURS IN ABOVE FUNCTION


    statusFlag = statusFlag && buildArray ( paddata_y , padLen);


...
}

int buildArray ( double** array, int size )
{
    int statusFlag = 1;
    showMessage(" (buildArray) Enter BuildArray... ", 1);
    *array = ( double* ) malloc ( sizeof ( double ) * size );
      
 //CRASH HAPPENS HERE IN CALL FROM CONVLV DATA DURING MALLOC LINE

    showMessage(" (buildArray) Malloc finished... ", 1);
    statusFlag = ( *array != NULL );
    checkError ( statusFlag , "Memory allocation has failed (buildArray)." , 0 );
    return statusFlag;
}

Thanks in advance for any help that can be offered.

Might have fixed it, turns out i was overrunning the arrays built in wrap so when new arrays started to be built they were trying to build on a memory section which was already occupied (I think) which i guess caused the crash.

I'm using visual studio express for coding right now, is there any functionality in this to check that i'm not overrunning arrays?

Cheers!

Originally Posted by Talus

Might have fixed it, turns out i was overrunning the arrays built in wrap so when new arrays started to be built they were trying to build on a memory section which was already occupied (I think) which i guess caused the crash.

I'm using visual studio express for coding right now, is there any functionality in this to check that i'm not overrunning arrays?

Cheers!

Nope.

C has absolutely no run time error checking that is not provided by the operating system.

It is on you as a programmer to write code that doesn't overrun your array boundaries or whatever other nasties might happen.

@Salem

Thanks, I'll give that a look. It could be incredibly helpful when i'm allocating and freeing everywhere.

@CommonTater

So you're saying that the program would never crash due to the fact that I was overrunning arrays and then making new ones where they may have overran into? I didn't realise this. Must have been something else causing the crash, which is a little worrying. I'll look more into it.

Thanks for your help anyhow

Originally Posted by Talus

@Salem

Thanks, I'll give that a look. It could be incredibly helpful when i'm allocating and freeing everywhere.

@CommonTater

So you're saying that the program would never crash due to the fact that I was overrunning arrays and then making new ones where they may have overran into? I didn't realise this. Must have been something else causing the crash, which is a little worrying. I'll look more into it.

Thanks for your help anyhow

Nope, I didn't say it wouldn't crash... In fact Windows would catch that in many cases and produce an allocation error.

What it absolutely will do is behave in an unexpected and/or unpredictable manner... You may get variables that change "like magic", you may get stack corruption and/or lockups, you might go for months and see nothing, then one day (usually when demonstrating it for either your boss or a big client) it's going to screw up royally... One prime example was something that happened to me in one of my programs. I was accidentally overrunning an array and clobbered a variable next to it (a struct with an int at the top)... one day I gets a call from my customer... "How come it says we have 4 million and some torgue wrenches in stock?"... Took me a while to figure that one out, I can tell you.

Just about any screwy thing can happen.