Question

I’m writing some code for a statistics package, and I start off by reading in data to a pointer array. I initialize the pointer and allocate sufficient memory using malloc; however, I sometimes get an error in the memory allocation at the end of the code below.

#include <stddef.h>
#include <stdio.h>
#include <stdlib.h>
#include "stats.h"

int main(int argc, char **argv) {
FILE *fp, *outFile, *outBin;    // create a file identifier
size_t n, nbins; // # of data points
double *data; // pointer to hold data
double average, variance, *med; // stat returns
int medianComplete, histComplete, i; // return 1 on success
hist_t *Histogram;

// read in the number of bins from exe arguments
nbins = atoi(argv[1]);
nbins = (size_t)nbins;

// open the binary datafile and read in first value
// which is number of data points
// use exe input for filename
fp = fopen(argv[2],"rb");
fread(&n, sizeof(size_t),1,fp);

// allocate enough memory to hold all data
data = (double*)malloc(sizeof(double)*n);
if (!data)
    printf("Memory allocation error");
fread(data,sizeof(double),n,fp);

This program compiles and runs well on my personal machine (MacOSX), but fails due to a segmentation error when I try to run it on a Linux server. I used Valgrind to see if I could track down the error and I receive the following result.

==8641== Warning: silly arg (-501426814648844128) to malloc()
==8641== Invalid write of size 1
==8641==    at 0x4C2B20D: mempcpy (mc_replace_strmem.c:956)
==8641==    by 0x4EA2F15: _IO_file_xsgetn (fileops.c:1423)
==8641==    by 0x4E971D2: fread (iofread.c:44)
==8641==    by 0x40086D: main (runstats.c:28)
==8641==  Address 0x0 is not stack'd, malloc'd or (recently) free'd

This is the first program I’ve ever written that uses pointers, and I’m at a loss for why it would work on one system but not the other.

Answer 1

The sizeof(size_t) is platform dependent. On OSX with the default 32 bit code, size_t is 4 bytes. On 64 bit Linux, size_t is 8 bytes. If you’re running on 64 bit Linux, you’re reading a different n that you’re reading on 32 bit OSX.

If you must use a binary data format, don’t use the size of platform specific types as the field size. Decide whether the file header is 4 or 8 bytes, little or big endian, and use that consistently.

See:
What's sizeof(size_t) on 32-bit vs the various 64-bit data models?