size_t or long for the size of a string containing a file?

Question

Suppose I want to read an entire file in memory. I'd open it in binary mode, use fseek to get to the end of the file, then do ftell to get its size. Then I would allocate a string with the same size as the file, and just read it in, right?

The problem is that ftell returns a long int, and malloc should receive a size_t argument. Now, size_t can be larger than long int as far as I can tell (I've just checked C: a reference manual by Harbison and Steele, and Chapter 11 mentions that size_t can be defined as unsigned long long, depending on the compiler). I suppose the opposite could be true (and that would be a real problem, since I'd be casting a long long into a long, or something like that)

So my question is how do I deal with that?

Thanks!

edit: thanks guys, that was really quick! I'll use mmap (and posix_madvise too, which I've found after diggin up info about mmap)!

Answer 1

I think you want to use mmap() instead.

Answer 2

I think the issue of casting is moot. Since the smaller of the two, long, has a maximum of +2^31 (if signed), if you read all of that at once you'd be occupying 2GB of memory. This is usually a bad idea.

Answer 3

If your files are large enough for this to be an issue, they are also large enough for the memory allocation to almost certainly fail.

Either don't worry about it, or switch to an OS specific memory mapping technique.