🔥 LET’S GO. TIME TO BUILD A 🧠⚔️ LEGENDARY PIPE RACE CONDITION SIMULATOR.

---

🧬 12_pipe_zero_byte_race

⚠️ Concept:

Simulate edge-case behavior when:

• A writer writes 0 bytes to a pipe

• A reader polls or blocks

• You create race windows with usleep, poll, and pipe() behavior

---

📄 pipe_zero_byte_race.c

#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <fcntl.h>
#include <pthread.h>
#include <poll.h>
#include <string.h>
#include <errno.h>
 
#define PIPE_READ  0
#define PIPE_WRITE 1
 
int pipe_fds[2];
 
void	*x_writer(void *arg)
{
	(void)arg;
	usleep(50000); // let reader hit poll first
	printf("[writer] 🧃 Writing 0 bytes to pipe\n");
	ssize_t	wrote = write(pipe_fds[PIPE_WRITE], "", 0);
	printf("[writer] write() returned %zd (%s)\n", wrote, strerror(errno));
	return (NULL);
}
 
void	*x_reader(void *arg)
{
	(void)arg;
	struct pollfd pfd = {
		.fd = pipe_fds[PIPE_READ],
		.events = POLLIN
	};
 
	printf("[reader] ⏳ Waiting for pipe data via poll()...\n");
	int	ret = poll(&pfd, 1, 2000);
	if (ret == -1)
	{
		perror("poll");
		return (NULL);
	}
	if (ret == 0)
	{
		printf("[reader] ⏱️ poll() timed out\n");
		return (NULL);
	}
	if (pfd.revents & POLLIN)
	{
		char buf[8];
		ssize_t r = read(pipe_fds[PIPE_READ], buf, sizeof(buf));
		printf("[reader] 🧠 Read %zd bytes\n", r);
	}
	else
		printf("[reader] 🔒 poll() woke up without POLLIN\n");
 
	return (NULL);
}
 
int	main(void)
{
	if (pipe(pipe_fds) == -1)
	{
		perror("pipe");
		exit(1);
	}
 
	pthread_t writer, reader;
	pthread_create(&reader, NULL, x_reader, NULL);
	pthread_create(&writer, NULL, x_writer, NULL);
 
	pthread_join(reader, NULL);
	pthread_join(writer, NULL);
 
	close(pipe_fds[PIPE_READ]);
	close(pipe_fds[PIPE_WRITE]);
 
	return (0);
}

---

🧪 Expected Behavior

• Reader blocks in poll()

• Writer writes 0 bytes

• poll() might not wake up

• May see:

  • poll() timed out

  • OR nothing is read

• But kernel wakes poll() anyway on some systems

---

title: Is this undefined behavior?
```ad-note
title: Kernel Specific
No — `write(fd, NULL, 0)` is defined and **always returns 0**  
But triggering side effects on `poll()` **is implementation-dependent**  
→ This is what makes this edge behavior interesting to test.

---

✅ Compile & Run

gcc -Wall -Wextra -Werror -pthread pipe_zero_byte_race.c -o pipe_zero_byte_race
./pipe_zero_byte_race

---

🧠 180 IQ Notes

• Try reversing order of threads (reader second) 🔁

• Try writing 1 byte instead of 0 📦

• Try poll with POLLHUP, POLLERR, etc.

• Try using select() instead of poll()

• Try replacing pipe with socketpair for bidirectional async

---

🧠 Status

✅ [[asymmetric_c/12_pipe_zero_byte_race/]] ready to lock
⏳ Would benefit from:

• ➕ 12b_pipe_write_1byte.c for control

• ➕ 12c_socketpair_poll_test.c to compare

Say “continue” or “next variant” to expand the rabbit hole 🕳️👇