C question

From: Sean 'Captain Napalm' Conner <spc_at_conman.org>
Date: Thu Aug 1 17:25:01 2002

It was thus said that the Great Eric J. Korpela once stated:
>
> > So here's where I'm missing it. Under all three fread scenarios, the
> > compiler doesn't throw a warning...they all compile cleanly. But only the
> > first works.
>
> Lack of warnings or code don't have anything to do with whether the code
> is correct...

  Oh, that reminds me, if you can, crank up the warning level on your
compiler if you can, and track down every single one of them. I run my code
through

                          gcc -Wall -pedantic -ansi

  (although if the supplied system headers are messed up, I might drop the
-ansi flag). Also, be aware that a unspecified char declaration

        char c;

  can cause c to be treated as a signed *or* unsigned quantity, depending
upon the compiler, so the following code *may* not work on some systems:

        void foo(void)
        {
          char c;
          char buffer[256];

          for (c = 0 ; c < 256 ; c++)
          {
            buffer[c] = c;
          }
        }

  Internally, C will treat the assignment to buffer[c] as:

        *(buffer + c) = c;

  And if c is treated as a signed quantity, then once you hit the value 128,
what you get is (effectively):

        *(buffer - 128) = -128;

  which is *not* what you want. So beware of using chars as indicies.

> Call fread with the following parameters:
> 1. byte value at address (ucMem+roms[i].iROMStart+MEM_ROM) converted into
> a pointer to type char.
> 2. The value 1 (assuming byte is typedefed to "unsigned char")
> 3. The value stored in roms[i].iROMLen
> 4. The value stored in pFH.
>
> It's fairly easy to see why the first parameter is wrong... No warning is
> issued because the type of the first parameter is correct. Without the
> typecast, a warning should be issued.

  Actually, it's correct---it's a pointer. The declaration of fread() takes
a void * as the first parameter, and *any* pointer type can be cast into a
void * without problem, no typecast needed.

  -spc (Been there, ported the code, have the core file ... )
Received on Thu Aug 01 2002 - 17:25:01 BST