Day 002: The Type System

Today I opened K&R section 1.2 — Variables and Arithmetic Expressions. A Fahrenheit-to-Celsius conversion table. Fourteen lines of code. It broke my assumptions about how computers do math.

What I Did

I typed in the integer version of the temperature conversion program. Compiled it. Ran it. The output was wrong. Not wrong because of a typo. Wrong because I told the machine to do integer math and integer math throws away remainders.

The formula is celsius = 5/9 * (fahr - 32). When both 5 and 9 are integers, C computes 5/9 as 0. Not 0.555. Zero. Integer division truncates toward zero. Multiply anything by zero and every Celsius value comes back as zero.

The fix: declare the variables as float instead of int. Write 5.0/9.0 instead of 5/9. Now the machine does floating point math and the table comes out right.

This is not a quirk. This is the machine telling me exactly what I asked for. I asked for integer division. I got integer division. C does not protect you from yourself.

The Questions That Came Up

I spent the session writing down every question the reading triggered before looking anything up. Here is what I found.

What are the basic data types in C?

C gives you types that map to what the hardware provides. char is one byte. short is usually two. int is usually four. long is four or eight depending on the platform. float is four bytes. double is eight.

The important part: C does not guarantee sizes. int is defined as “at least 16 bits.” On my machine it is 32 bits. On the PDP-11 where C was born it was 16. The language was designed to run on any hardware. If you need exact sizes — and in security work you almost always do — you use types like int32_t and uint8_t from stdint.h.

Rust learned from this. Every integer type has its size in the name: i32, u8, u64. No ambiguity. Python went the other direction and hid the machine entirely. Python integers grow as large as your memory allows. No overflow. No awareness of what is happening underneath.

Both are valid choices. But if you want to understand the machine — and especially if you want to understand why software breaks — you need to know what C is doing.

Why did my compiler reject code without int main?

Yesterday my hello program compiled fine. Today I wrote a program without int in front of main and the compiler told me “ISO C99 and later do not allow implicit int.”

In original K&R C from 1978, if you left off the return type, the compiler assumed int. This was called implicit int. The C99 standard killed it. You now have to say what you mean.

K&R the book was written during the transition from the old style to ANSI C. Some examples still use the old conventions. When my compiler rejects something from the book, I am watching forty years of language evolution happen in real time. That is not an obstacle. That is the education.

Where does printf come from?

printf is not part of the C language. It is part of the C Standard Library — libc — included through the stdio.h header.

Here is the actual chain of events when I call printf("hello"):

My program calls printf. The printf function in libc does all the formatting work — handling %d, %f, padding, precision — in user space. Then it calls write(), which is a system call. The kernel receives that call and sends the bytes to my terminal.

printf is a convenience layer. I could skip it entirely and call write(1, "hello\n", 6) directly. That day is coming.

Every language seems to have printf-style formatting because every language that came after C was either written in C, ran on an OS written in C, or was designed by people who grew up with C. Python’s % formatting. Rust’s println! macro. Java’s printf. All descendants.

Format specifiers

The % constructions in printf are called format specifiers. The main ones: %d for integers, %f for floats, %c for characters, %s for strings, %x for hexadecimal, %p for pointer addresses.

You can add width and precision: %6d prints an integer in a field at least six characters wide. %6.1f prints a float in a field six wide with one decimal place. This is how you make columns line up.

There is also %n, which writes the number of characters printed so far into a variable. It is a legendary source of security vulnerabilities. Format string attacks exploit %n when user input gets passed directly to printf without sanitization. I will study that in detail later.

Why are variables declared at the top?

In C89 — the standard K&R second edition targets — declaring all variables at the beginning of a block was a rule, not a suggestion. The compiler needed to know all variable sizes up front to set up the stack frame.

C99 relaxed this. You can now declare variables anywhere, including inside a for loop: for (int i = 0; i < 10; i++). But the book follows C89 conventions, so everything is at the top.

Variables declared inside a function are local to that function. They live on the stack. When the function returns, that memory is reclaimed. If you return a pointer to a local variable, that pointer now points at memory that no longer belongs to you. That is a dangling pointer — one of the most dangerous bugs in C. We will get there.

The Feynman Test

How does C decide whether to do integer math or real math?

C looks at the operands. If both are integers, it does integer math and throws away any remainder. If either operand is a float or double, it promotes the other to match and does floating point math. The programmer decides which kind of math happens by choosing the types. The machine does exactly what you tell it. Nothing more.

Why should a security engineer care? Because integer truncation and overflow are real vulnerability classes. A calculation that looks correct in your head can produce zero — or worse, a negative number that wraps around to a massive positive — when the machine does integer math. Bounds checks fail. Buffer sizes go wrong. Exploits follow.

Understanding the type system is not academic. It is the first layer of understanding how software breaks.

What Is Next

Sections 1.3 and 1.4 — the for loop and symbolic constants. The temperature table rewritten three ways, each one teaching something new about how C thinks.

Also: twenty minutes of vimtutor. The editor is part of the craft.

Day 2 of 365. The type system is not a convenience. It is a contract between you and the machine.