Prototyping rules/recommendations sought

R

rs

Dear All,


I have a question regarding proptypes for functions. What is the
recommended practice? The way I do it is to put all my external functions
in a header file, while protyping internal (file scope) functions at the
start of the source file. I've seen many people (especially using gcc
under linux) don't botehr prototyping internal functions but just declare
them inline so to speak.

Is there any recommendations advantages/disdvantages of the approach? It
seems to be that there is a tradeoff here:

1. If the prototypes are going to help me by catching incorrect parameter
types,numbers etc.. they are worth having

2. However its a hassle to maintain them.

So, the question is, are most modern compilers able to catch type/number
errors so that I can avoid prototyping internal functions? What are your
suggestions?
 
H

Hallvard B Furuseth

rs said:
So, the question is, are most modern compilers able to catch type/number
errors so that I can avoid prototyping internal functions? What are your
suggestions?

If a function has no declaration, only a definition, and is used before
it is defined, it gets used without a prototype. Some compilers may
give a warning if the expected prototype is wrong, but they may not
infer the prototype from the later definition. So you should prototype
such static functions. Other static functions need no declarations.

Of course, you might forget that you are using a static function before
declaring it unless the compiler warns about implicitly declared
functions (e.g. with the gcc -Wimplicit option), so you may prefer to
declare all static functions anyway.
 
R

Robert Stankowic

rs said:
Dear All,


I have a question regarding proptypes for functions. What is the
recommended practice? The way I do it is to put all my external functions
in a header file, while protyping internal (file scope) functions at the
start of the source file. I've seen many people (especially using gcc
under linux) don't botehr prototyping internal functions but just declare
them inline so to speak.

Is there any recommendations advantages/disdvantages of the approach? It
seems to be that there is a tradeoff here:

1. If the prototypes are going to help me by catching incorrect parameter
types,numbers etc.. they are worth having

2. However its a hassle to maintain them.

So, the question is, are most modern compilers able to catch type/number
errors so that I can avoid prototyping internal functions? What are your
suggestions?

Not a suggestion (who am I to give suggestions)..
I write the functions in a compilation unit "upside down" and put protoypes
of only those functions which are called from a different compilation unit
into a header:

in somecode.c:

int foo(void)
{
return 1;
}

double bar(double some_val)
{
return some_val / 1.234;
}

int main(void)
{
int int_val = foo();
double my_dbl = bar(2.55);
return 0;
}

in somecode.h i have only

double bar(double);
because foo() is used only in somecode.c, but bar() is called from
someothercode.c as well.

just my <whatever currency you like> 0.02
Robert
 
I

Irrwahn Grausewitz

Robert Stankowic said:
"rs" <[email protected]> schrieb:

Not a suggestion (who am I to give suggestions)..
I write the functions in a compilation unit "upside down" and put protoypes
of only those functions which are called from a different compilation unit
into a header:

in somecode.c:

int foo(void)
{
return 1;
}

You may want to change this to:

static int foo(void)
...

just my <whatever currency you like> 0.02

I added mine, that makes 0.04. ;-)

Regards
 
M

Mark F. Haigh

rs said:
Dear All,
1. If the prototypes are going to help me by catching incorrect parameter
types,numbers etc.. they are worth having
So, the question is, are most modern compilers able to catch type/number
errors so that I can avoid prototyping internal functions? What are your
suggestions?

Quoting the C90 & C99 standards: A function prototype is a declaration
of a function that declares the types of its parameters.

We end up with the interesting effect that the definition of a function
is also a prototype:

/* Both definition and prototype! */
static int foo(char *p)
{
/* ... */
return 42;
}

int main(void)
{
char *p;
foo(p); /* Prototype in scope, full type checking */
return 0;
}

Many experienced programmers do this whenever possible for static
functions. As you noted, why maintain function prototypes for static
functions when you don't have to?


Mark F. Haigh
(e-mail address removed)
 
R

rs

Does this apply only to functions that are defined BEFORE they are used or
also to those defined after they are used?
 
C

CBFalconer

Robert said:
Not a suggestion (who am I to give suggestions)..
I write the functions in a compilation unit "upside down" and put
protoypes of only those functions which are called from a different
compilation unit into a header:

in somecode.c:

int foo(void)
{
return 1;
}

double bar(double some_val)
{
return some_val / 1.234;
}

int main(void)
{
int int_val = foo();
double my_dbl = bar(2.55);
return 0;
}

in somecode.h i have only

double bar(double);
because foo() is used only in somecode.c, but bar() is called from
someothercode.c as well.

In this case you should also declare foo() as being static, to
avoid polluting the external name space. The general rule is that
the header contains only things meant to be visible to other
modules. Because of the context sensitive meaning of static, you
should use it for all declarations visible over file scope that
are NOT intended to be visible externally.
 
I

Irrwahn Grausewitz

[Please don't top-post; fixed.]
"Mark F. Haigh" <[email protected]> wrote:

Does this apply only to functions that are defined BEFORE they are used or
also to those defined after they are used?

To have a prototype in scope when a function is called, the function
has to be either defined (as a function definition serves as prototype
as well) or explicitly prototyped /before/ used.

Mark's example above is of the first kind: definition before use.

The same example, using a separate prototype:

/* Prototype: */
static int foo(char *);

int main(void)
{
char *p;
foo(p); /* Prototype in scope, full type checking */
return 0;
}

/* Note that the definition of foo can now savely be
moved to the end of the file: */

static int foo(char *p)
{
/* ... */
return 42;
}

If you omit the prototype in this example, the compiler cannot
perform full type checking.

HTH
Regards
 
M

Mark Gordon

[Please don't top-post; fixed.]
"Mark F. Haigh" <[email protected]> wrote:

Does this apply only to functions that are defined BEFORE they are
used or also to those defined after they are used?

To have a prototype in scope when a function is called, the function
has to be either defined (as a function definition serves as prototype
as well) or explicitly prototyped /before/ used.

<snip>

<mode=awkward sod>
Not all function definitions provide prototypes.
static int foo(*p)
char *p
{
/* ... */
return 42;
}

does not provide a prototype. This, of course, is a very good reason for
never using this type of function definition.
</mode>
 
I

Irrwahn Grausewitz

Mark Gordon said:
Irrwahn Grausewitz <[email protected]> wrote:

<snip>

<mode=awkward sod>
Not all function definitions provide prototypes.
static int foo(*p)
char *p
{
/* ... */
return 42;
}

does not provide a prototype. This, of course, is a very good reason for
never using this type of function definition.
</mode>

Yes, you are right. I stopped using "old-style" function definitions
years ago, and almost forgot it ever existed.

Regards
 
M

Mark F. Haigh

Mark said:
Not all function definitions provide prototypes.
static int foo(*p)
char *p
{
/* ... */
return 42;
}

does not provide a prototype. This, of course, is a very good reason for
never using this type of function definition.

Very good point.

<OT>
Some compilers have switches to warn you when this occurs. For gcc,
it's -Wstrict-prototypes, which will produce something like:

foo.c:13: warning: function declaration isn't a prototype
</OT>


Mark F. Haigh
(e-mail address removed)
 
M

Mark Gordon

Not all function compile...

I put in an extra * and missed a ;. It should have been

static int foo(p)
char *p;
{
/* ... */
return 42;
}

That'll teach me to compile first.
 

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments. After that, you can post your question and our members will help you out.

Ask a Question

Members online

Forum statistics

Threads
473,744
Messages
2,569,483
Members
44,902
Latest member
Elena68X5

Latest Threads

Top