[SOLVED] multiple declaration error

piobair · 09-12-2014, 08:49 AM

I have a case of interlocking subroutines.

Code:

int mouse[2];
void doit();
void submouse(int btn, int state, int x, int y){
  mouse[0] = x;
  mouse[1] = y;
  doit();
}
void newwindow(){
  subWindow[1][0] = glutCreateSubWindow(mainWindow, 10, 10, 616, 274);
  windowlevel = 1;
  glutMouseFunc(submouse);
}

void doit(){
  // if subwindow has not been created, create subwindow
  if(windowlevel == 0)newwindow();
  else{ //process whatever the mouse clicked
  }
}

In other words, doit() calls newwindow() which creates submouse() which calls doit().
However, that coding produces the following error message:

Quote:

program.c:4418:6: error: static declaration of ‘doit’ follows non-static declaration
program.c:3745:6: note: previous declaration of ‘doit’ was here

What am I doing wrong?

smallpond · 09-12-2014, 09:32 AM

It's saying that your actual code has the "static" keyword in one place but not the other:

Code:

static 
void doit() {

piobair · 09-12-2014, 12:21 PM

I do not use the "static" keyword in either place.
The latest version of gcc for my i686 machine (Debian stable) is 4:4.7.2-1. The current for my AMD64 machine is 4:4.8.2-4
This program has compiled without error on the AMD64 for the past year.

Why is the 4:4.7.2-1 giving me static (sic), and how do I get it to compile?

Quote:

Originally Posted by smallpond

It's saying that your actual code has the "static" keyword in one place but not the other:

Code:

static 
void doit() {

metaschima · 09-12-2014, 01:27 PM

Are you sure 'doit()' is only declared once in your program (check other files as well).

P.S. doit is not the best name for a function.

NevemTeve · 09-12-2014, 03:43 PM

Just check the lines the compiler mentioned.
(btw this program is bigger than good -- the right size is a few hundred lines, not five thousands)

piobair · 09-12-2014, 05:25 PM

doit() is declared twice: once as a placeholder, and the second time with its 230 lines of code. The placeholder is necessary since I have interlocking subroutines.

doit() gets called whenever a mouse gets clicked. It sorts out which window (of currently about a dozen) was under the mouse when it got clicked, and then calls the appropriate subroutine for execution of the mouse-picked function. Note: doit() is not static and not recursive: It is called the first time per mouse click on the main window, whereupon it creates an appropriate subwindow. It then goes away until the mouse is clicked on that subwindow. Think of doit() as a universal state monitor. Including header files, the program is currently up to ~7000 lines of C & OpenGL, and growing.

Quote:

Originally Posted by metaschima

Are you sure 'doit()' is only declared once in your program (check other files as well).

P.S. doit is not the best name for a function.

metaschima · 09-12-2014, 05:43 PM

Either rename the doit function or make it static (prototype and declaration). I suspect it may be declared in an include file somewhere. If you make it static it will override the previous declaration. Not the best solution, renaming it would be better.

piobair · 09-12-2014, 10:18 PM

See bug report http://gcc.gnu.org/ml/gcc-bugs/2004-12/msg00415.html
The declaration of a function should match the prototype of that declaration. I am seeing that bug error in gcc 4:4.7.2-1 but not 4:4.8.2-4
In other words, it is a compiler error, not a coding error.

NevemTeve · 09-13-2014, 12:29 AM

So, this function was declared as 'static' even if you stated otherwise in #3.
The definition and the declaration should be identical, even if some compilers (or compiler versions) don't give error if they aren't.

(Also you should do a little reorganization, eg:

old:
static void doit ();

new:
static int dispatch_mouse_event ();
)

piobair · 09-13-2014, 07:39 AM

1. either "static void doit()" or "void static doit()" results in an error message, "invalid storage class for function doit". Apparently this is a bug: http://www.geeksforgeeks.org/what-ar...unctions-in-c/
2. Upon research, my use of the prototype of doit is not only acceptable practice, but is actually recommended. The bug appeared when gcc started requiring that the arguments of the prototype exactly match the arguments of the coded function. Gcc version 3.* allowed an empty argument field, which could be promoted by the coded version of the function.

Quote:

Originally Posted by NevemTeve

So, this function was declared as 'static' even if you stated otherwise in #3.
The definition and the declaration should be identical, even if some compilers (or compiler versions) don't give error if they aren't.

(Also you should do a little reorganization, eg:

old:
static void doit ();

new:
static int dispatch_mouse_event ();
)

NevemTeve · 09-13-2014, 09:47 AM

Sure, the complete code should include the parameters, too:

Code:

static void doit (void);
...
static void doit (void)
{...}

but, as I of said, it should be redesigned, eg:

Code:

typedef struct MouseEvent {
    int button;
    int state;
    int x;
    int y;
} MouseEvent;

static void dispatch_mouse_event (const MouseEvent *me);

static void submouse(int btn, int state, int x, int y)
{
  MouseEvent me;

  me.button= button;
  me.state= state;
  me.x= x;
  me.y= y;
  dispatch_mouse_event (&me);
}

static void dispatch_mouse_event (const MouseEvent *me)
{
  // if subwindow has not been created, create subwindow
  if(windowlevel == 0)newwindow();
  else{ //process whatever the mouse clicked
  }
}

metaschima · 09-13-2014, 10:45 AM

The '-pedantic' did it, that's why I never use it.

ntubski · 09-13-2014, 03:22 PM

Quote:

Originally Posted by piobair

See bug report http://gcc.gnu.org/ml/gcc-bugs/2004-12/msg00415.html
The declaration of a function should match the prototype of that declaration. I am seeing that bug error in gcc 4:4.7.2-1 but not 4:4.8.2-4
In other words, it is a compiler error, not a coding error.

The compiler error in that bug report is that the compiler fails to reject or at least warn about the code.

Quote:

https://gcc.gnu.org/ml/gcc-bugs/2004-12/msg00418.html

The C standard says this is invalid code which is why 4.0.0 errors out.

NevemTeve · 09-13-2014, 11:09 PM

Standards come and go, compilers may or may not follow their changes; eventually, it's the programmer who is responsible for his code.