fread, longs and cheap PCs

Hi,

I've been trying to load a file into memory, using the following code:-

Code:

---

long fsize(FILE *fp)
{
  long pos, size;

  pos = ftell(fp);
  fseek(fp, 0, SEEK_END);
  size = ftell(fp);
  fseek(fp, pos, SEEK_SET);
  return size;
}

int main(void)
{
  unsigned char *temp;
  long filesize;
  FILE *fp;

  fp = fopen("data.dat", "rb");
  filesize = fsize(fp);
  temp = (unsigned char *)malloc(filesize);
  if (temp == NULL)
  {
      printf("Memory allocation error\n");
      return -1;
  }
  fread(temp, filesize, 1, fp);
  return 0;
}

---

However, when I compile this with Borland Turbo C 2.0, using a large memory model and then run it, it crashes (and on these cheap Cyrix MX/S3 Virge-based PCs my school uses, the screen messes up completely...).
It seems that temp is allocated correctly (malloc doesn't return NULL), but the fread statement is causing bother. The fact that I'm using a long for filesize when size_t is defined as an unsigned int (max 65,535) is concerning, but as the file I'm testing it on is only 48,200 bytes, it shouldn't overflow upon truncation, should it?

On many systems fread best reads from binary files, doesn't it?

Perhaps shift the opening and closing of this file into the function, use fgets and concat to temp. Have the function return the allocated memory.

Code:

---

char *func(char *filename)
{
//file pointer
//fopen(filename, "r");
//*buf
//return NULL string if error
//size == ...
//buf = malloc(sizeof(char) *size);
//while(!feof(fp))
//fgets()
//strcat()

buf[strlen(buf)- 1] = '0';

return buf;
}


//main

char *contents = func("data.txt");

printf("%s", contents);

free(contents);

//return 0

---

Eh?

The file is binary, so using text read functions to read it is just gonna make things even more complicated (bumping into control codes all over the place, etc.). You ever tried printfing an image? 'Tis well ugly.

Well then I was off the mark then-wasn't I?

Why are you are freading binaries into an unsigned char pointer?

Is this conventional? Why not to a struct or something?

Does it work with the huge memory model?

alex

I think I'd better clear things up a bit...

I'm loading a big chunk of RLE-encoded image data into memory so I can quickly reconstruct the original image and display. I'm currently doing it so that each individual byte is read from the hard drive and dealt with, but that takes ages, so I'm trying to put it temporarily into memory, decode the information into a separate buffer, and deallocate the memory containing the RLE data. Sorry for any confusion.

Have you tried using a 'far' pointer? Since you're using a 16bit compiler, you may be beyond what a 'near' pointer is able to handle. Honestly, I've never had occasion to use them (I'd really suggest getting a new compiler. There are some good free ones!) but I believe you want something like:

unsigned char far *buf;

Note: Far pointers are obsoleted with 32bit compilers.

Quzah.

If you use far pointers, you should also allocate the memory in a separate segment (my guess: farmalloc). You said that reading data from file byte per byte was very slow... Maybe you're better off reading the file in small segments of a few Kb. A second guess :) of mine: you are using the BGI library to draw? This is very slow if you use putpixel to draw one pixel at a time!

alex

{ printf int main (int argc, char **argv)
{
struct stat stat;
void *mem = (void *) NULL;
int fd;

if (stat("fullpath/filename", &stat != -1)
{
if ( mem = malloc ((size_t) stat.st_size) != NULL)
{
if ((fd = open("fullpath/filename", O_RDONLY )) != -1)
{
read(fd, mem, (size_t)stat.st_size);
close(fd);
..... do your own code
free(mem);
}
}
}
else
("File not found\n");
}
return(0);
}

Note : Use Unix and mmap the file, ist easier and much faster,
If Your file is bigger than 64K data segment, use
huge memory model and far pointers.

So how are these RLE-encodings stored as unsigned chars?

And why not decode it while you read it in to storage, bit by bit?

I USED to decode them into storage from disk, but like I said, it takes ages, so I'm trying to stick it in memory first THEN decode it and free the encoded data, which should give me much faster processing. The RLE is PCX-style (> 0xC0 (192) for a run, then the pixel value for it in the next byte).

And I thought compiling in the large memory model meant that data pointers were far by default? :eek:

There might be other problems . . . like some error checking that isn't being done, etc., but one thing that you'll need to do is change:

fread(temp, filesize, 1, fp);

to

fread(temp, (size_t)filesize, (size_t)1, fp);


Hope This Helps,
Joshua Burkholder

as an alternative file length derivation, use this and maybe it'll work out...

Code:

---

#include <io.h>  //  fileno ()
#include <stdio.h>  //  filelength ()
size = filelength ( fileno (fp) );

---