How C Programs Are Built

"Compiling" a C program is really four steps: the preprocessor expands #include and #define, the compiler translates C into assembly, the assembler produces machine code, and the linker joins it with libraries into an executable. Knowing the steps makes error messages — and multi-file projects later — far less mysterious.

The four stages

 main.c ──▶ preprocessor ──▶ compiler ──▶ assembler ──▶ main.o ──▶ linker ──▶ main (executable)
           (#include,        (C → assembly) (assembly →             (+ C library,
            #define)                         machine code)            other .o files)

Normally one command runs all four. You can stop after any stage to look at the intermediate result.

The preprocessor

The preprocessor is a text-substitution step. #include <stdio.h> is literally replaced by the contents of the stdio.h header file, and #define names are replaced by their values:

#include <stdio.h>
#define GREETING "Hi there"
#define REPEAT 3

int main(void) {
    for (int i = 0; i < REPEAT; i++) {
        printf("%s\n", GREETING);
    }
    return 0;
}

Hi there
Hi there
Hi there

After preprocessing, the compiler never sees GREETING or REPEAT, only "Hi there" and 3. Lesson 51 covers the preprocessor in depth.

Headers and libraries

A header (.h file) contains declarations: it tells the compiler that a function such as printf exists and what arguments it takes. The function's actual machine code lives in a library — for printf, the C standard library, which the linker adds automatically. A few libraries need asking for, like the maths library on Linux (-lm).

That split explains two different kinds of error:

• Compiler errors — your C is invalid, or you used a function without its header.
• Linker errors ("undefined reference to …", or "undefined symbol" on macOS) — the code compiled, but the linker can't find the function's definition: a missing library or a missing .c file.

Errors and warnings

An error stops compilation. A warning means the code is valid C but probably wrong — and in C, "valid but wrong" is the dangerous kind. Here, the comparison uses = (assignment) instead of ==:

#include <stdio.h>

int main(void) {
    int stock = 0;
    if (stock = 5) {
        printf("In stock\n");
    }
    return 0;
}

main.c:5:15: warning: using the result of an assignment as a condition without parentheses [-Wparentheses]

This program would compile and run — and always print "In stock". Treat every warning as a bug to fix. Many projects add -Werror, which turns warnings into errors.

Compiled vs. interpreted

Python or JavaScript programs are run by an interpreter that reads the source every time. A C program is translated once, ahead of time, into machine code for one kind of processor and operating system. That's why C is fast and needs no runtime installed — and also why you must recompile after every change, and compile separately for each platform.

Try it yourself

Compile this with gcc -E main.c | tail -5 to see the preprocessed end of the file, then with gcc -S main.c and open main.s to see the assembly the compiler generated.

With tax: 60