Two declarations of the same function in c

Question

I had this question in a test, and I still don't understand the answer I was given:

Let's say I wrote the following code:

#include <math.h>
#include <stdio.h>

float cos(float x){
    return 1-x*x/4;
}

int main()
{
    printf("%0f",cos(0.05f)+sin(0.05f));
}

Let's assume cos and sin are declared and defined in the math library (receiving and returning double), and I'm trying to link my code with the math library.

Another assumption is that cos is defined in math.c.

The question was:

"Will the code compile/link successfully? if so, which cos function will be called?"

The answer was:

"Yes, the code will compile and my cos will be called".

How could this behavior be explained? Aren't these multiple definitions of the same function?

Answer 1

Your teacher may have made a mistake and intended to use double cos(double x). In this case, many C implementations will accept the program, and it will link and run because the linker takes every module from the object modules it is supplied but only takes the needed modules from the libraries it is supplied. Thus, because cos is already defined in the program, the linker will not take it from the math library. However, although this works in many C implementations, it violates the rules of standard C, which reserves the library identifiers; normal programs may not define them.

Another possibility is that your teacher did not intend to include math.h. This would make the declaration of cos not be an error, since it would not be conflicting with another declaration, but it would mean that sin should also be declared by the program, since it is used.

Answer 2

It will not work at all.

You will see an error like this one:

conflicting types for 'cos'

I got it with codeblocks with gcc compiler ...

There is another solution you can use, please check this post: Override a function call in C

Answer 3

It will not compile.

I added a return 0; at the end of main() to remove a second problem with -Wall -Werror. If you do this you will see:

$ gcc -Wall -Werror costest1.c -o costest -lm
costest1.c:5:1: error: conflicting types for ‘cos’

This fails at the compile stage because math.h also defines a function called cos. Note the prototype for cos is:

double cos(double x);

not

float cos(float x);

If you did not include math.h, you would be able to compile, but would get:

$ gcc -Wall -Werror costest1.c -o costest -lm
costest1.c:5:1: error: conflicting types for built-in function ‘cos’ [-Werror]
costest1.c: In function ‘main’:
costest1.c:13:3: error: implicit declaration of function ‘sin’ [-Werror=implicit-function-declaration]
costest1.c:13:32: error: incompatible implicit declaration of built-in function ‘sin’ [-Werror]
cc1: all warnings being treated as errors

This is because cos is not a normal function, but handled as a builtin. As you can see it's defined in terms of sin. If cos were a normal function, you would see a duplicate symbol error of some sort.

In C you cannot have two functions with the same name even if they have different arguments. In C++ you can, in that identically named methods may differ by the calling parameters (but not just by the return type).

Answer 4

EDIT:

I assume the question was about C++ and not C since compiling your code as a C program will generate conflicting types error: https://eval.in/93380.

This behaviour is caused by the Function Overloading. The cos() function has the assembly name _cos@double and redeclaring the cos() function to be accepting a float argument will have the assembly name_cos@floatand it will not conflict with the cos() defined in the math library. And calling cos() with a float argument will be translated to a call to the assembly function _cos@float which is your own cos().

Note that Function Overloading is allowed only in C++ and not C.

In C you can only do it without changing the arguments and the return types of the function (https://eval.in/93381) otherwise the previous error will be generated.

As the assumption clarifies (cos is defined in math.c), the cos() function in the case is other than the one defined in the math.h in which case the the answer would be true if it was defined to be accepting an argument of type float and returning a value of type float. In this case the program will compile without any issues.

Answer 5

Online C2011 standard:

6.7 Declarations
...
Constraints
...
4 All declarations in the same scope that refer to the same object or function shall specify compatible types

The code you posted violates the above constraint; math.h declares cos as

double cos(double x);    

This behavior cannot be explained as C; it can be explained as C++, which allows for name overloading.

Make sure you're really talking about C and not C++, otherwise you are going to wind up being very confused.

Answer 6

No it does not compile.

The header file math.h declares double cos(double). Attempting to overload is not allowed in C.

error C2371: 'cos' : redefinition; different basic types