Stack smashing detected during fread on binary file in C - c

Question

UPDATE IN BOTTOM====
So a while ago I made the following function, which I successfully used to get the grey values from images (w x h dimension) that were converted to .bin-files. It just gives an array of all pixel values.
It was, however, not as a function like this but put in the main() immediately.
// read the BIN-file as grayscale image
void decodeBIN(const char* filename, short image[], int w, int h){
int i = 0;
unsigned char buffer[16]; // no specific size attributed
FILE *ptr;
ptr = fopen(filename, "rb");
if (!ptr){
printf("\nUnable to open file!\n"); // error
}
while (!feof(ptr)){
fread(buffer,2,1,ptr); // read w bytes to buffer
image[i] = buffer[1];
//printf("%u ", image[i]); // DEBUG
i++;
}
fclose(ptr);
printf("\nBinary image read (npixels: %i).\n", i-1); // DEBUG
}
I decided to expand the code, so I rewrote it to the previous function and put it in a separate file for functions and also made a header file. The extra file for functions and the header file work 100% so that's not the issue. Now, this code does not work anymore and I get a stack smashing error. Some variables called after this function have also jumped to another value, so I figured the problem was with the buffer (I didn't know about the correct size for the buffer, but it worked...). After some experimentation and testing, I came up with the following function. I replaced the buffer with a char array named image2 to simply try and test it:
void decodeBIN(const char* filename, short image[], int w, int h){
int i = 0, res;
char image2[];
FILE *ptr;
ptr = fopen(filename, "rb"); //"MySnap_20180327-2239-010.bin"
if (!ptr){
printf("\nUnable to open file!\n"); // error
}
res = fread(image2,1,w*h,ptr) // need to read w*h pixels
while (i < w*h){ // DEBUG
printf("%i ", (int)image2[i]); // DEBUG
i++;
}
printf("\nRead %u bytes\n", res); // DEBUG
fclose(ptr);
printf("Binary image read (npixels: %i).\n", i); // DEBUG
}
I'm a bit lost in how it used to work and all of a sudden when I move the code from main() to a function it stops working, so any help is welcome!
Thanks in advance.
Disclaimer: I'm aiming to write this with the help of as few libraries as possible
===== UPDATE:
After the answer of #alainmerigot I got this code, which helped with getting the correct values:
void decodeBIN(const char* filename, unsigned char image[], int w, int h){
int i = 0, res;
FILE *ptr;
res = fread(image,sizeof(char),w*h,ptr) // need to read w*h pixels
fclose(ptr);
}
The segmentation fault and jumped variables are still in place though, so here a more upper-level oversight of what I'm doing:
char filenamePathed["[path of file]/file.bin"];
short img1[npixels]; // npixels = w*h
printf("i_file: %i\n", i_file); // correct value
decodeBIN(filenamePathed, img_curr, w, h); // decode bin
printf("i_file: %i\n", i_file); // value jumped
while (i < npixels){
img1[i] = (short)img_curr[i];
i++;
}
Perhaps it is good to know that I'm doing this iteratively for multiple files (time series)? I also need it to end up in a (short) format (or integer, but eventually needs to be memory-efficient and pixels have a range of 0-255 so int is a bit abundant imo).

Answer 1

The problem with your second function is that you write in array image2 while no space has been reserved for it. Declaring char image2[]; only says that an array exists and that the address of this array can be found in var image2, but no space is associated with it, hence the problem.
You can associate space with this array by several means.
Using permanent storage in the heap
image2 = malloc(x*y); // but do not forget to free(image2) at the end of the function
Using temporary storage in the stack (space is automatically freed when leaving the function).
image2 = alloca(x*y); // slightly faster than malloc and do not require to free() the image
But the best is to use a array with parametrized size (since C99). Your array should be declared as
char image2[w*h]; // will use the value of w and h to define array size
If you want to do other things than printing the image values in your function, you should store the image in permanent memory and have a mean to know the address of the array in your program. This is probably what you intended and is the reason why you have short image[] in your parameter list.
The solution is just simply to use image instead of image2 in fread().
But, the declaration of image should be coherent and image should be an array of char not short.
Beware also of declarations. In your first function, the image is an array of unsigned char and in the second an array of char. While the storage size is identical and fread() will store the same values, they are not equivalent. If used in an arithmetic context, image[i] will be interpreted differently and the results will likely be different. In general, images are unsigned.

Answer 2

Apparently, the problem was with the allocation of image, although I'm not sure why it was wrong.
I used to allocate it with unsigned char image[npixels]; and the solution to the error appeared to be unsigned char image[npixels*7];
Somehow it works, but if anyone has an explanation, please do so :)