This is a quick refresher on Go concurrency, not a beginner's guide. If you want to learn concurrency from the ground up with practical exercises, check out my other book — Gist of Go: Concurrency.

The book is AI-free.

Goroutines • Channels • Select • Pipelines • Time • Context • Wait groups • Data races • Race conditions • Mutexes • Semaphores • Signaling • Run once • Object pool • Atomics • Testing • Scheduling • Diagnostics • Final thoughts

# Goroutines

The foundation of concurrency in Go is goroutines – functions started with the go keyword:

func main() {
    var wg sync.WaitGroup
    wg.Add(2)
    go func() {
        defer wg.Done()
        fmt.Println("worker 1")
    }()
    go func() {
        defer wg.Done()
        fmt.Println("worker 2")
    }()
    wg.Wait()
}

worker 2
worker 1

The Go runtime juggles these goroutines and distributes them among operating system threads running on CPU cores. Compared to OS threads, goroutines are lightweight, so you can create hundreds or thousands of them.

Goroutines are completely independent. The main function is also a goroutine, but it starts implicitly when the program starts. When main ends, other goroutines also shut down.

We use a wait group (sync.WaitGroup) to wait for goroutines to finish in the example above. A wait group has a counter inside. Calling Add(n) increments it by n, while Done() decrements it by one. Wait() blocks the calling goroutine (in this case, main) until the counter reaches zero. This way, main waits for both workers to finish before it exits.

WaitGroup.Go automatically increments the wait group counter, runs a function in a goroutine, and decrements the counter when it's done:

func main() {
    var wg sync.WaitGroup
    wg.Go(func() {
        fmt.Println("worker 1")
    })
    wg.Go(func() {
        fmt.Println("worker 2")
    })
    wg.Wait()
}

worker 2
worker 1

# Channels

Goroutines can pass values to each other through channels. A channel is like a window where one goroutine can throw something and another can catch it:

func main() {
    messages := make(chan string)

    go func() { messages <- "ping" }()

    msg := <-messages
    fmt.Println(msg)
}

ping

Sending a value through a channel is a synchronous operation. When the sending goroutine writes a value to the channel (ch <- val), it blocks and waits for someone to receive that value (<-ch). Only then does it continue.

Output channel

Returning an output channel from a function and filling it within an internal goroutine is a common pattern in Go. This allows the caller to receive values through the channel while the owning function retains control of it:

func generate(start, stop int) chan int {
    out := make(chan int)
    go func() {
        for i := start; i < stop; i++ {
            out <- i
        }
    }()
    return out
}

Closing a channel

To signal readers that all data has been sent, the writer goroutine closes the channel with close():

func generate(start, stop int) chan int {
    out := make(chan int)
    go func() {
        defer close(out)
        for i := start; i < stop; i++ {
            out <- i
        }
    }()
    return out
}

The reader checks the channel's status with a second value ("comma OK") when reading:

func main() {
    in := generate(5, 10)
    for {
        num, ok := <-in
        if !ok {
            break
        }
        fmt.Print(num, " ")
    }
}

5 6 7 8 9

While the channel is open, the reader receives the next value and a true status. If the channel is closed, the reader gets a zero value and a false status.

A channel can only be closed once. Closing it again or writing to a closed channel causes a panic.

The only reason to close a channel is to signal to its readers that all data has been sent. If this isn't important to the readers, then you don't need to close it. When a channel is no longer used, Go's garbage collector will free its resources, whether it's closed or not.

Channel iteration

range automatically reads the next value from the channel and checks if it's closed. If the channel is closed, it exits the loop:

func main() {
    nums := generate(5, 10)
    for n := range nums {
        fmt.Print(n, " ")
    }
}

5 6 7 8 9

Range over a channel returns a single value, not a pair, unlike range over a slice.

Directional channels

You can protect yourself from accidental write/close errors by setting the channel direction. Channels can be:

• chan (bidirectional): for reading and writing (default);
• chan<- (send-only): for writing only;
• <-chan (receive-only): for reading only.

You can't read from a send-only channel or write to a receive-only channel (nor can you close it).

Channels are usually initialized for both reading and writing, and specified as directional in function parameters. Go automatically converts a regular channel to a directional one:

stream := make(chan int)

go func(in chan<- int) {
    in <- 42
}(stream)

func(out <-chan int) {
    fmt.Println(<-out)
}(stream)

42

Buffered channels

Buffered channels work like a FIFO queue with a fixed-size buffer for storing values.

As long as the buffer has free space, writing to the channel doesn't block the goroutine. Similarly, as long as the buffer contains values, reading from the channel doesn't block the goroutine:

stream := make(chan int, 3)
stream <- 11
stream <- 12
stream <- 13

fmt.Println(<-stream)
fmt.Println(<-stream)

11
13

By default, if you don't specify a buffer size, a channel is unbuffered (buffer size equals zero).

Buffered channels work with the built-in len() and cap() functions:

stream := make(chan int, 3)
stream <- 11
fmt.Println(cap(stream), len(stream))

3 1

Reading from a closed buffered channel returns values from the buffer and a true status. Once all values are taken, it returns a zero value and a false status, like a regular channel:

stream := make(chan int, 1)
stream <- 11
close(stream)

val, ok := <-stream
fmt.Println(val, ok)
// 11 true

val, ok = <-stream
fmt.Println(val, ok)
// 0 false

11 true
0 false

nil channel

Like any type in Go, channels have a zero value, which is nil.

Writing to or reading from a nil channel blocks the goroutine indefinitely:

var stream chan int

go func() {
    // blocks forever
    stream <- 1
}()

// blocks forever
<-stream

Closing a nil channel causes a panic:

var stream chan int
close(stream)
// panic: close of nil channel

# Select

The select statement is somewhat like switch, but specifically designed for channels. Here's what it does:

• Checks which cases are not blocked.
• If multiple cases are ready, randomly selects one to execute.
• If all cases are blocked and there is a default case, executes it.
• If all cases are blocked and there is no default case, waits until one is ready.

Select is used to manage data flow in pipelines:

// merge sends values from in1 and in2 to the output channel.
func merge(in1, in2 <-chan int) <-chan int {
    out := make(chan int)
    go func() {
        defer close(out)
        for in1 != nil || in2 != nil {
            select {
            case val1, ok := <-in1:
                if ok { out <- val1 } else { in1 = nil }
            case val2, ok := <-in2:
                if ok { out <- val2 } else { in2 = nil }
            }
        }
    }()
    return out
}

// Suppose we send 10..12 to in1, 20..22 to in2,
// and call merge(in1, in2)

10 11 20 12 21 22

To cancel goroutines:

// process modifies values from in and send them to out
// until in is exhausted or cancel is closed.
func process(cancel chan struct{}, in <-chan int) <-chan int {
    out := make(chan int)
    go func() {
        for val := range in {
            select {
            case out <- val*10:
            case <-cancel:
                fmt.Println("canceled")
                return
            }
        }
    }()
    return out
}

// Suppose we send values 11 and 12 to in
// and then call close(cancel)

110
120
canceled

For non-blocking operations:

// multiplier returns a function that multiplies
// the input by 10 and sends it to the channel
// or returns an error if the channel is busy.
func multiplier(ch chan<- int) func(n int) error {
    return func(n int) error {
        select {
        case ch <- n*10:
            return nil
        default:
            return errors.New("busy")
        }
    }
}

func main() {
    nums := make(chan int, 1)
    multiply := multiplier(nums)

    err := multiply(11)
    fmt.Println(<-nums, err)
    // 110 <nil>

    err = multiply(12)
    fmt.Println(<-nums, err)
    // 120 <nil>

    err = multiply(13)
    err = multiply(14)
    fmt.Println(err)
    // busy
}

110 <nil>
120 <nil>
busy

And for much more.

# Pipelines

A pipeline is a sequence of operations where each step takes input data, processes it in a specific way, and outputs it. The input and output of each operation is a channel.

A typical pipeline looks like this:

• Reader: Reads input data from a file, database, or network.
• N processors: Transform, filter, aggregate, or enrich data using external sources.
• Writer: Writes the processed data to a file, database, or network.

func read[T any]() <-chan T {
    out := make(chan T)
    go func() {
        defer close(out)
        for {
            // read data from somewere
            data := // ...
            out <- data
        }
    }()
    return out
}

func process[T any](in <-chan T) <-chan T {
    out := make(chan T)
    go func() {
        defer close(out)
        for inData := range in {
            // process the data
            outData = // ...
            out <- outData
        }
    }()
    return out
}

func write[T any](in <-chan T) <-chan struct{} {
    done := make(chan struct{})
    go func() {
        defer close(done)
        for data := range in {
            // write the data
        }
    }()
    return done
}

Output channel

A goroutine can signal other goroutines that it has finished its work using an output channel:

func generate(start, stop int) <-chan int {
    out := make(chan int)
    go func() {
        defer close(out)
        for i := start; i < stop; i++ {
            out <- i
        }
    }()
    return out
}

func main() {
    nums := generate(5, 10)
    for n := range nums {
        fmt.Print(n, " ")
    }
}

5 6 7 8 9

Done channel

If a goroutine doesn't need to return results, it can signal completion using a done channel:

func work() <-chan struct{} {
    done := make(chan struct{})
    go func() {
        defer close(done)
        fmt.Println("work done")
    }()
    return done
}

func main() {
    done := work()
    <-done
}

work done

Cancel channel

To terminate a goroutine early, a calling goroutine can use a cancel channel:

func generate(cancel chan struct{}, n int) <-chan int {
    out := make(chan int)
    go func() {
        defer close(out)
        for i := 1; i <= n; i++ {
            select {
            case out <- i:
            case <-cancel:
                return
            }
        }
    }()
    return out
}

func main() {
    cancel := make(chan struct{})
    defer close(cancel)

    nums := generate(cancel, 10)
    fmt.Println(<-nums)
    fmt.Println(<-nums)
    fmt.Println(<-nums)
}

1
2
3

Error handling

There are three approaches to error handling in concurrent pipelines.

➊ Return on the first error:

// calculate produces answers for the given numbers.
func process(in <-chan int) (<-chan int, <-chan error) {
	out := make(chan Answer)
	errc := make(chan error, 1)
	go func() {
		defer close(out)
		for n := range in {
			ans, err := fetchAnswer(n)
			if err != nil {
				errc <- err  // return with error
				return
			}
			out <- ans
		}
		errc <- nil          // return with nil
	}()
	return out, errc
}

➋ Use a result type:

// Result contains an answer or an error.
type Result struct {
	answer int
	err    error
}

// calculate produces answers for the given numbers.
func calculate(in <-chan int) <-chan Result {
	out := make(chan Result)
	go func() {
		defer close(out)
		for n := range in {
			ans, err := fetchAnswer(n)
			out <- Result{ans, err}  // return answer + error
		}
	}()
	return out
}

➌ Collect errors separately:

// calculate produces answers for the given numbers.
func calculate(in <-chan int, errc chan<- error) <-chan int {
	out := make(chan Answer)
	go func() {
		defer close(out)
		for n := range in {
			ans, err := fetchAnswer(n)
			if err == nil {
				out <- ans   // send answer
			} else {
				errc <- err  // or error
			}
		}
	}()
	return out
}

# Time

Besides handling date and time, the time package offers tools for managing time-sensitive operations in concurrent programs.

After

time.After() returns a channel that is initially empty, but receives a value after the timeout period. It's useful for timing out operations:

// withTimeout executes a function with a given timeout.
func withTimeout(timeout time.Duration, fn func()) error {
    done := make(chan struct{})
    go func() {
        defer close(done)
        fn()
    }()

    // blocks until fn completes or the timer expires,
    // whichever happens first
    select {
    case <-done:
        return nil
    case <-time.After(timeout):
        return errors.New("timeout")
    }
}

withTimeout() waits for fn() to complete, but thanks to time.After(), it won't wait longer than the timeout duration:

func main() {
    var err error

    // completes in time
    err = withTimeout(
        50*time.Millisecond,
        func() { fmt.Println("work done") },
    )
    fmt.Println("err =", err)

    // gets canceled on timeout
    err = withTimeout(
        50*time.Millisecond,
        func() {
            time.Sleep(100 * time.Millisecond)
            fmt.Println("work done")
        },
    )
    fmt.Println("err =", err)
}

work done
err = <nil>
err = timeout

Timer

A timer (time.Timer) is a structure with a C channel to which it sends the current time when it triggers (expires). Timers are useful for planning future executions:

done := make(chan struct{})

timer := time.NewTimer(50 * time.Millisecond)
go func() {
    eventTime := <-timer.C  // blocks for 50ms
    fmt.Println("work done at", eventTime)
    close(done)
}()

<-done

work done at 2009-11-10 23:00:00.05

Stop() stops the timer and returns true if it hasn't expired yet, and false otherwise:

// timer expires after 50ms
timer := time.NewTimer(50 * time.Millisecond)
go func() {
    eventTime := <-timer.C
    fmt.Println("work done at", eventTime)
}()

// after 10ms, the timer hasn't expired yet
time.Sleep(10 * time.Millisecond)

if timer.Stop() {
    fmt.Println("execution canceled")
} else {
    fmt.Println("too late to cancel")
}

execution canceled

It's often more convenient to use the time.AfterFunc() wrapper function. It waits for duration d and then executes function f:

done := make(chan struct{})
work := func() {
    fmt.Println("work done")
    close(done)
}

// executes work after 50ms
time.AfterFunc(50*time.Millisecond, work)
<-done

work done

time.AfterFunc() returns a timer that you can cancel before execution starts:

// executes the function after 50ms
timer := time.AfterFunc(50*time.Millisecond, func() {})

// after 10ms, the timer hasn't expired yet
time.Sleep(10 * time.Millisecond)

if timer.Stop() {
    fmt.Println("execution canceled")
}

execution canceled

If a timer is used in a loop, it's better to create a single timer and reset it instead of creating a new instance on each iteration:

// consumer reads tokens from the input channel and alerts
// if a value does not appear in a channel after an hour.
func consumer(in <-chan token) {
    const timeout = time.Hour
    timer := time.NewTimer(timeout)
    for {
        timer.Reset(timeout)
        select {
        case <-in:
            // do stuff
        case <-timer.C:
            // log warning
        }
    }
}

// Suppose we send 10,000 values to the in channel
// and measure memory usage.

Memory used: 4 KB, # allocations: 6

Ticker

A ticker is like a timer, but it keeps firing until you stop it. Tickers are useful for executing periodic tasks:

// fires every 50ms
ticker := time.NewTicker(50 * time.Millisecond)
defer ticker.Stop()

go func() {
    for {
        // waits for ticker to fire on each iteration
        at := <-ticker.C
        fmt.Println("work done at", at)
    }
}()

// enough time for the ticker to fire 3 times
time.Sleep(160*time.Millisecond)
ticker.Stop()

work done at 2009-11-10 23:00:00.05
work done at 2009-11-10 23:00:00.10
work done at 2009-11-10 23:00:00.15

NewTicker(d) creates a ticker that sends the current time to the channel C at interval d. You must stop the ticker eventually with Stop() to free up resources.

If the channel reader can't keep up with the ticker, the ticker will skip ticks.

# Context

The main purpose of context is to cancel operations, either manually or by timeout/deadline.

The function accepts a context and uses its Done() channel to listen for cancellation:

// work performs a task for 50 ms unless canceled.
// Returns an error when canceled.
func work(ctx context.Context) error {
    done := make(chan struct{})

    go func() {
        time.Sleep(50 * time.Millisecond)
        fmt.Println("work done")
        close(done)
    }()

    select {
    case <-done:
        return nil
    case <-ctx.Done():
        return ctx.Err()
    }
}

Cancel manually (context.Canceled error):

func main() {
    // empty context
    ctx := context.Background()
    // manual canellation context
    ctx, cancel := context.WithCancel(ctx)
    defer cancel()

    done := make(chan struct{})
    go func() {
        // takes 50 ms unless canceled
        err := work(ctx)
        fmt.Println("err =", err)
        close(done)
    }()

    // cancels after 10 ms
    time.Sleep(10 * time.Millisecond)
    cancel()
    <-done
}

err = context canceled

Cancel by timeout (context.DeadlineExceeded error):

func main() {
    ctx := context.Background()
    // cancels after 10 ms
    ctx, cancel := context.WithTimeout(ctx, 10*time.Millisecond)
    defer cancel()

    done := make(chan struct{})
    go func() {
        // takes 50 ms unless canceled
        err := work(ctx)
        fmt.Println("err =", err)
        close(done)
    }()

    <-done
}

err = context deadline exceeded

Cancel by deadline (context.DeadlineExceeded error):

func main() {
    ctx := context.Background()
    // cancels at now + 10 ms
    deadline := time.Now().Add(10 * time.Millisecond)
    ctx, cancel := context.WithDeadline(ctx, deadline)
    defer cancel()

    done := make(chan struct{})
    go func() {
        // takes 50 ms unless canceled
        err := work(ctx)
        fmt.Println("err =", err)
        close(done)
    }()

    <-done
}

err = context deadline exceeded

Context is layered. A context object is immutable. To add new properties to a context, a new (child) context is created based on the old (parent) context. The shorter timeout between the parent and child contexts always wins. The child context can only shorten the parent's timeout, not extend it:

func main() {
    // parent context with a 100 ms timeout
    const dur100ms = 100 * time.Millisecond
    parentCtx, cancel := context.WithTimeout(context.Background(), dur100ms)
    defer cancel()

    // child context with a 10 ms timeout
    const dur10ms = 10 * time.Millisecond
    childCtx, cancel := context.WithTimeout(parentCtx, dur10ms)
    defer cancel()

    // now the work gets canceled
    err := work(childCtx)
    fmt.Println("err =", err)
}

err = context deadline exceeded

Multiple cancels are safe. You can call cancel() on the context as many times as you want. The first cancel will work, and the rest will be ignored.

You can specify a custom cancellation cause using context.WithCancelCause(), context.WithTimeoutCause() and context.WithDeadlineCause(). This cause is accessible through context.Cause():

ctx, cancel := context.WithCancelCause(context.Background())
cancel(errors.New("the night is dark"))
fmt.Println(context.Cause(ctx))

the night is dark

You can register a function to execute when the context is canceled with context.AfterFunc():

ctx, cancel := context.WithCancel(context.Background())
cleanup := func() { fmt.Println("cleanup") }
context.AfterFunc(ctx, cleanup)
cancel()
time.Sleep(10 * time.Millisecond)

cleanup

Context can pass additional information about a call using context.WithValue(), which creates a context with a value for a specific key. But it's generally better to avoid passing values in context. It's better to use explicit parameters or custom structs instead.

# Wait groups

The sync.WaitGroup type lets you wait for one or more goroutines to finish:

const n = 10
var wg sync.WaitGroup
wg.Add(n)
for range n {
    go func() {
        defer wg.Done()
        fmt.Print(".")
    }()
}
wg.Wait()

..........

A WaitGroup doesn't know anything about the goroutines it manages. It works with an internal counter. Calling wg.Add(1) increments the counter by one, while wg.Done() decrements it. wg.Wait() blocks the calling goroutine until the counter reaches zero.

The Go method combines Add, starting a goroutine, and Done:

var wg sync.WaitGroup
for range 10 {
    wg.Go(func() {
        fmt.Print(".")
    })
}
wg.Wait()

..........

All methods are safe to use from multiple goroutines.

Normally, all Add calls happen before Wait. But technically, there's nothing stopping you from doing some of the Add calls before Wait and some after (from another goroutine).

You can call Wait from multiple goroutines. They will all block until the group's counter reaches zero.

# Data races

A data race happens when multiple goroutines access shared data, and at least one of them modifies it. We need to protect the data from this kind of concurrent access.

A data race doesn't always cause a runtime panic. That's why Go provides a special tool called the race detector. You can turn it on with the race flag, which works with the test, run, build, and install commands.

var total int

// There's a data race on total.
var wg sync.WaitGroup
wg.Go(func() { total++ })
wg.Go(func() { total++ })
wg.Wait()

fmt.Println("total:", total)

total: 2

go run -race main.go

==================
WARNING: DATA RACE
...
2
Found 1 data race(s)

Channels are safe for concurrent reading and writing, and they don't cause data races.

Ways to prevent data races:

• Avoid concurrent data modification (typically by using channels).
• Synchronize access with mutexes.
• Use only atomic operations.

Race conditions

A race condition happens when an unpredictable order of operations from multiple goroutines leads to an incorrect system state:

// There's a race condition when working with balance.
withdraw := func(amount int) {
    if getBalance() < amount {
        return
    }
    time.Sleep(time.Millisecond)
    setBalance(getBalance() - amount)
}

setBalance(50)

var wg sync.WaitGroup
wg.Go(func() { withdraw(40) })
wg.Go(func() { withdraw(40) })
wg.Wait()

fmt.Println("balance:", getBalance())

balance: -30

If individual operations are concurrent-safe, Go's race detector won't find any issues. Because of this, it doesn't catch race conditions:

go run -race main.go

balance: -30

You can't fully eliminate uncertainty in a concurrent environment. Events will happen in an unpredictable order — that's just how concurrency works. However, you can prevent a race condition — often by protecting a composite operation with a mutex:

var mu sync.Mutex
withdraw := func(amount int) {
    mu.Lock()
    defer mu.Unlock()

    if getBalance() < amount {
        return
    }
    time.Sleep(time.Millisecond)
    setBalance(getBalance() - amount)
}

setBalance(50)

var wg sync.WaitGroup
wg.Go(func() { withdraw(40) })
wg.Go(func() { withdraw(40) })
wg.Wait()

fmt.Println("balance:", getBalance())

balance: 10

Compare-and-set

Sometimes you can prevent a race condition without using mutexes by applying an atomic compare-and-set operation or one of its flavors:

// CompareAndSet changes the value to new if the current value equals old.
// Returns true if the value was changed.
CompareAndSet(old, new any) bool

// CompareAndSwap changes the value to new if the current value equals old.
// Returns the old value.
CompareAndSwap(old, new any) any

// CompareAndDelete deletes the value if the current value equals old.
// Returns true if the value was deleted.
CompareAndDelete(old any) bool

// etc

The idea is always the same:

• Check if the assumed (old) state matches reality.
• If it does, change the state to new.
• If not, do nothing.

# Mutexes

The sync.Mutex type protects shared data and parts of your code from being accessed concurrently:

var total int
var mu sync.Mutex

var wg sync.WaitGroup
for range 100 {
    wg.Go(func() {
        mu.Lock()
        time.Sleep(time.Millisecond)
        total++
        mu.Unlock()
    })
}
wg.Wait()

total: 100

The mutex guarantees that only one goroutine can run the code between Lock() and Unlock() at a time.

A mutex is used in these situations:

• When multiple goroutines are modifying the same data.
• When one goroutine is modifying the data and others are reading it.

If all goroutines are only reading the data, you don't need a mutex.

TryLock

The TryLock method tries to lock the mutex, just like a regular Lock. But if it can't, it returns false right away instead of blocking the goroutine:

var total int
var mu sync.Mutex

var wg sync.WaitGroup
for range 100 {
    wg.Go(func() {
        if !mu.TryLock() {
            return
        }
        defer mu.Unlock()
        time.Sleep(time.Millisecond)
        total++
    })
}
wg.Wait()

total: 1

RWMutex

The sync.RWMutex type distinguishes between readers and writers. It provides two sets of methods:

• Lock / Unlock lock and unlock the mutex for both reading and writing.
• RLock / RUnlock lock and unlock the mutex for reading only.

var total int
var mu sync.RWMutex
var wg sync.WaitGroup

// 10 writers.
for range 10 {
    wg.Go(func() {
        mu.Lock()
        defer mu.Unlock()
        time.Sleep(time.Millisecond)
        total++
    })
}

// 10 readers.
for range 10 {
    wg.Go(func() {
        // Try switching from RLock/RUnlock to Lock/Unlock
        //and see how it affects the elapsed time.
        mu.RLock()
        defer mu.RUnlock()
        time.Sleep(time.Millisecond)
        _ = total
    })
}

wg.Wait()

elapsed: 10ms

Here's how it works:

• If a goroutine locks the mutex with Lock(), other goroutines will be blocked if they try to use Lock() or RLock().
• If a goroutine locks the mutex with RLock(), other goroutines can also lock it with RLock() without being blocked.
• If at least one goroutine has locked the mutex with RLock(), other goroutines will be blocked if they try to use Lock().

This creates a "single writer, multiple readers" setup.

Locker

Both sync.Mutex and sync.RWMutex implement the same sync.Locker interface:

type Locker interface {
    Lock()
    Unlock()
}

By using Locker instead of a specific mutex type, you can build components that don't depend on a specific lock implementation. This lets the client decide which lock to use.

Channel as mutex

You can use a channel instead of a mutex to protect shared data:

var total int
lock := make(chan struct{}, 1)

var wg sync.WaitGroup
wg.Go(func() {
    lock <- struct{}{}
    defer func() { <-lock }()
    total++
})
wg.Go(func() {
    lock <- struct{}{}
    defer func() { <-lock }()
    total++
})
wg.Wait()

total: 2

# Semaphores

A semaphore is like a container with N available slots and two operations: acquire to take a slot and release to free a slot. Here are the semaphore rules:

• Calling acquire takes a free slot.
• If there are no free slots, acquire blocks the goroutine that called it.
• Calling release frees up a previously taken slot.
• If there are any goroutines blocked on acquire when release is called, one of them will immediately take the freed slot and unblock.

You can implement a simple semaphore with a buffered channel, where N is the channel's size. To acquire the semaphore, send a value into the channel. To release it, take a value from the channel:

// Try changing nConc and see how the elapsed time changes.
const nConc = 4
const nCalls = 100
sema := make(chan struct{}, nConc)

var wg sync.WaitGroup
for range nCalls {
    sema <- struct{}{} // acquire
    wg.Go(func() {
        defer func() { <-sema }() // release
        time.Sleep(time.Millisecond) // do some work
    })
}
wg.Wait()

elapsed: 25ms

For more complex situations, use the golang.org/x/sync/semaphore package.

Rendezvous

A rendezvous lets two goroutines wait for each other:

• There are two goroutines — G1 and G2 — and each one can signal that it's ready.
• If G1 signals but G2 hasn't yet, G1 blocks and waits.
• If G2 signals but G1 hasn't yet, G2 blocks and waits.
• When both have signaled, they both unblock and continue running.

You can implement a simple rendezvous with a wait group:

var rend sync.WaitGroup
rend.Add(2)

var wg sync.WaitGroup
wg.Go(func() {
    fmt.Println("before rendezvous")
    rend.Done()
    rend.Wait()
    fmt.Println("after rendezvous")
})
wg.Go(func() {
    fmt.Println("before rendezvous")
    rend.Done()
    rend.Wait()
    fmt.Println("after rendezvous")
})
wg.Wait()

before rendezvous
before rendezvous
after rendezvous
after rendezvous

Barrier

A barrier is a general case of a rendezvous. It lets N goroutines wait for each other:

• The barrier has a counter (starting at 0) and a threshold N.
• Each goroutine that reaches the barrier increases the counter by 1.
• The barrier blocks any goroutine that reaches it.
• Once the counter reaches N, the barrier unblocks all waiting goroutines.

You can implement a simple barrier with a wait group:

const n = 4
var bar sync.WaitGroup
bar.Add(n)

var wg sync.WaitGroup
for range n {
    wg.Go(func() {
        fmt.Println("before the barrier")
        bar.Done()
        bar.Wait()
        fmt.Println("after the barrier")
    })
}
wg.Wait()

before the barrier
before the barrier
before the barrier
before the barrier
after the barrier
after the barrier
after the barrier
after the barrier

# Signaling

The sync.Cond (conditional variable) type lets one goroutine signal to another that it's ready, and lets the other goroutine wait for that signal.

A Cond includes a mutex and has two methods — Wait and Signal.

• Wait unlocks the mutex and suspends the goroutine until it receives a signal.
• Signal wakes the goroutine that is waiting on Wait.
• When Wait wakes up, it locks the mutex again.

cond := sync.NewCond(&sync.Mutex{})
done := false

var wg sync.WaitGroup
wg.Go(func() {
    cond.L.Lock()
    fmt.Println("G1 is ready to signal")
    done = true
    cond.Signal()
    cond.L.Unlock()
})
wg.Go(func() {
    cond.L.Lock()
    for !done {
        cond.Wait()
    }
    fmt.Println("G2 received the signal")
    cond.L.Unlock()
})
wg.Wait()

G1 is ready to signal
G2 received the signal

If there are multiple waiting goroutines when Signal is called, only one of them will be resumed. If there are no waiting goroutines, Signal does nothing.

You can also use the Broadcast method. While Signal wakes up only one goroutine waiting on Cond.Wait, the Broadcast method wakes up all such goroutines.

You can signal with a channel:

signal := make(chan struct{}, 1)

go func() {
    // do something
    signal <- struct{}{}
}()

go func() {
    <-signal
    // do something
}()

And broadcast too:

broadcast := make(chan struct{})

go func() {
    // do something
    close(broadcast)
}()

go func() {
    <-broadcast
    // do something
}()

go func() {
    <-broadcast
    // do something
}()

Broadcasting with a condition variable is limited: it only sends a signal, not the actual data, and it only works once. With channels, you can build a publish/subscribe system that doesn't have these limitations:

type Publisher struct {
    sbox []chan int // subscription channels
    mu   sync.Mutex // protects the state
}

func (p *Publisher) Subscribe() <-chan int {
    p.mu.Lock()
    defer p.mu.Unlock()
    sub := make(chan int, 1)
    p.sbox = append(p.sbox, sub)
    return sub
}

func (p *Publisher) Broadcast(v int) {
    p.mu.Lock()
    defer p.mu.Unlock()
    for _, sub := range p.sbox {
        select {
        case sub <- v:
        default:
        }
    }
}

# Run once

The sync.Once type makes sure that the given function runs only once. If multiple goroutines call Once.Do at the same time, only one will run the function, while the others will wait until it returns:

total := 0
initState := func() {
    total += 1
}

var once sync.Once

var wg sync.WaitGroup
wg.Go(func() {
    once.Do(initState)
    // do something
})
wg.Go(func() {
    once.Do(initState)
    // do something
})
wg.Wait()

total: 1

Once is perfect for one-time initialization or cleanup in a concurrent environment.

Besides the Once type, the sync package also includes three convenience once-functions:

// Calls f only once.
func (o *Once) Do(f func())

// Returns a function that calls f only once.
func OnceFunc(f func()) func()

// Returns a function that calls f only once
// and returns the value from that first call.
func OnceValue[T any](f func() T) func() T

// Returns a function that calls f only once
// and returns the pair of values from that first call.
func OnceValues[T1, T2 any](f func() (T1, T2)) func() (T1, T2)

# Object pool

The sync.Pool type helps reuse memory instead of allocating it every time, which reduces the load on the garbage collector:

pool := sync.Pool{
    New: func() any {
        buf := make([]byte, 1024)
        return &buf
    },
}

// Only allocates 4*1024 B, despite 4000 loop iterations.
var wg sync.WaitGroup
for range 4 {
    wg.Go(func() {
        for range 1000 {
            buf := pool.Get().(*[]byte)
            sink = buf
            pool.Put(buf)
        }
    })
}
wg.Wait()

Memory allocated: 4 KB

Get takes an item from the pool. If there are no available items, it creates a new one using New (which we have to define ourselves, since the pool doesn't know anything about the items it creates). Put returns an item back to the pool.

Things to keep in mind:

• New should return a pointer, not a value, to reduce memory copying and avoid extra allocations.
• The pool has no size limit. If you start 1000 more goroutines that all call Get at the same time, 1000 more buffers will be allocated.
• After an item is returned to the pool with Put, you shouldn't use it anymore (since another goroutine might already have taken and started using it).

# Atomics

An operation without synchronization can only be truly atomic if it translates to a single processor instruction. Such operations don't need locks and won't cause issues when called concurrently (even the write operations).

There are only a few atomics, and they're all found in the sync/atomic package:

Int32     Bool
Int64     Value
Uint32    Pointer
Uint64

Each atomic type provides the following methods:

• Load reads the value of a variable.
• Store sets a new value.
• Swap sets a new value (like Store) and returns the old one.
• CompareAndSwap sets a new value only if the current value is still what you expect it to be.

var n atomic.Int32
n.Store(10)
swapped := n.CompareAndSwap(10, 42)
fmt.Println("CompareAndSwap 10 -> 42:", swapped)
fmt.Println("n =", n.Load())

CompareAndSwap 10 -> 42: true
n = 42

Numeric types also provide an Add method that increments the value by the specified amount.

All methods are either translated into a single CPU instruction or are otherwise guaranteed to be atomic, so they are safe to use from multiple goroutines.

The composition of atomics is always non-atomic:

var delta atomic.Int32
var counter atomic.Int32

func increment() {
    // Not atomic; causes a race condition.
    delta.Add(1)
    sleep(10)
    counter.Add(delta.Load())
}

// After 100 concurrent increments,
// the final value is NOT guaranteed.

counter = 9386

A bulletproof way to make a composite operation atomic and prevent race conditions is to use a mutex:

var delta int32
var counter int32
var mu sync.Mutex

func increment() {
    // Atomic; doesn't cause a race condition.
    mu.Lock()
    delta += 1
    sleep(10)
    counter += delta
    mu.Unlock()
}

// After 100 concurrent increments, the final value is guaranteed:
// counter = 1+2+...+100 = 5050

counter = 5050

Sometimes you can use an atomic type instead of a mutex to exit early:

type Gate struct {
    closed atomic.Bool
}

func (g *Gate) Close() {
    if !g.closed.CompareAndSwap(false, true) {
        return // ignore repeated calls
    }
    // The gate is closed.
    // We can free resources now.
}

# Testing

If your concurrent program uses channels or custom types with synchronization methods like Wait, you can use those in your tests. This way, your tests won't be much more complicated than if the code were synchronous:

// Calc calculates something asynchronously.
func Calc() <-chan int {
    out := make(chan int, 1)
    go func() {
        out <- 42
    }()
    return out
}

func Test(t *testing.T) {
    // Wait for the Calc goroutine to finish.
    got := <-Calc()
    if got != 42 {
        t.Errorf("got: %v; want: 42", got)
    }
}

PASS

If there aren't any suitable synchronization "handles" in the code you're testing, you can use the synctest package. It exports two functions:

func Test(t *testing.T, f func(*testing.T))
func Wait()

synctest.Test runs an isolated bubble. The bubble uses a fake clock, and you can manually control goroutine synchronization with synctest.Wait.

synctest.Wait blocks until all goroutines in the bubble — except the one that called Wait — have either finished or are durably blocked. This lets you wait for a specific goroutine to finish or get blocked, so you can check the program's state:

// NewProc starts the calculation.
func NewProc() *Proc {
    p := &Proc{done: make(chan struct{})}
    go func() {
        p.res = 42
        <-p.done // (X)
        p.res = 0
    }()
    return p
}

func Test(t *testing.T) {
    synctest.Test(t, func(t *testing.T) {
        p := NewProc()
        defer p.Stop()

        // Wait for the goroutine to block at point X.
        synctest.Wait()
        if got := p.Res(); got != 42 {
            t.Fatalf("got %v, want 42", got)
        }
    })
}

PASS

The fake clock in synctest.Test move forward only if: ➊ all goroutines in the bubble are durably blocked; ➋ there's a future moment when at least one goroutine will unblock; and ➌ synctest.Wait isn't running. Thanks to this, time-dependent tests run instantly:

// Calc processes a value from the input channel.
// Times out if no input is received after 3 seconds.
func Calc(in chan int) (int, error) {
    select {
    case v := <-in:
        return v * 2, nil
    case <-time.After(3 * time.Second):
        return 0, ErrTimeout
    }
}

func Test(t *testing.T) {
    synctest.Test(t, func(t *testing.T) {
        ch := make(chan int)
        got, err := Calc(ch) // runs instantly

        if err != ErrTimeout {
            t.Errorf("got: %v; want: %v", err, ErrTimeout)
        }
        if got != 0 {
            t.Errorf("got: %v; want: 0", got)
        }
    })
}

PASS

The following operations durably block a goroutine:

• A blocking send or receive on a channel created within the bubble.
• A blocking select statement where every case is a channel created within the bubble.
• Calling Cond.Wait.
• Calling WaitGroup.Wait if all WaitGroup.Add calls were made inside the bubble.
• Calling time.Sleep.

Blocking on mutexes, I/O, or system calls is not considered durable, and the synctest bubble can't handle them.

# Scheduling

At the hardware level, CPU cores are responsible for running parallel tasks.

At the operating system level, a thread is the basic unit of execution. There are usually many more threads than CPU cores, so the operating system's scheduler decides which threads to run and which ones to pause.

At the Go runtime level, a goroutine is the basic unit of execution. The runtime scheduler runs a fixed number of OS threads, often one per CPU core. There can be many more goroutines than threads, so the scheduler decides which goroutines to run on the available threads and which ones to pause. The scheduler keeps switching between goroutines to make sure each one gets a turn to run on a thread, instead of waiting in line forever.

  CPU                  OS                   Go runtime
┌──────────┐  run on ┌──────────┐  run on ┌────────────┐
│ Cores    │ <────── │ Threads  │ <────── │ Goroutines │
└──────────┘         └──────────┘         └────────────┘

This is how Go handles concurrency.

Goroutine scheduler

The goroutine scheduler's job is to run M goroutines on N operating system threads, where M can be much larger than N. Here's a very simplified version of it's algorithm:

• If there's a free thread, assign it a goroutine from the queue.
• If a running goroutine gets blocked (for example, while reading from a channel), put it back in the queue and assign a different goroutine to the thread.
• If a running goroutine gets stuck in a syscall, start a new thread to run other goroutines until the blocked goroutine finishes the syscall.
• Check the running goroutines every 10 ms. Preempt long-running goroutines and return them to the queue to prevent starvation.

┌─────┐┌─────┐┌─────┐┌─────┐
│ G17 ││ G18 ││ G19 ││ G20 │                        queue
└─────┘└─────┘└─────┘└─────┘

┌─────┐      ┌─────┐      ┌─────┐      ┌─────┐
│ G15 │      │ G16 │      │ G13 │      │ G14 │      running
└─────┘      └─────┘      └─────┘      └─────┘
  │            │            │            │
┌──────────┐ ┌──────────┐ ┌──────────┐ ┌──────────┐
│ Thread E │ │ Thread F │ │ Thread C │ │ Thread D │
└──────────┘ └──────────┘ └──────────┘ └──────────┘

┌─────┐      ┌─────┐
│ G11 │      │ G12 │                                syscalls
└─────┘      └─────┘
  │            │
┌──────────┐ ┌──────────┐
│ Thread A │ │ Thread B │
└──────────┘ └──────────┘

The number of threads running Go code is controlled by the GOMAXPROCS environment variable or the runtime.GOMAXPROCS function.

A goroutine is a structure that starts out using about 2 KB of memory, mostly for its stack. The stack can grow if needed. Since goroutines are so lightweight, you can run tens of thousands or even hundreds of thousands of them on a small machine.

# Diagnostics

To troubleshoot concurrent programs in production, we use metrics, profiling, and tracing.

Metrics show how the Go runtime is performing, like how much heap memory it uses or how long garbage collection pauses take. Each metric has a unique name and a value, which can be a number or a histogram.

You can use the runtime/metrics package to get a complete list of metrics or check the values of specific ones:

samples := []metrics.Sample{
    {Name: "/sched/gomaxprocs:threads"},
    {Name: "/sched/goroutines:goroutines"},
}
metrics.Read(samples)

for _, s := range samples {
    fmt.Printf("%s: %v\n", s.Name, s.Value.Uint64())
}

/sched/gomaxprocs:threads: 8
/sched/goroutines:goroutines: 1

In practice, people rarely do this manually. Instead, all metrics are automatically exported using Prometheus or OpenTelemetry libraries.

Profiling helps you understand exactly what the program is doing, what resources it uses, and where in the code this happens. Go uses a sampling profiler that's suitable for production.

The most commonly used profiles are CPU, which shows how much processor time each function uses, and heap, which shows how much heap memory each function uses. Goroutine, block, and mutex profiles help identify problems related to concurrency.

The easiest way to add a profiler to your app is by using the net/http/pprof package. To collect a profile with the given name, call the /debug/pprof/{name} endpoint. To view the collected profile, use the go tool pprof utility:

go tool pprof -proto \
  "http://localhost:6060/debug/pprof/profile?seconds=N" > cpu.pprof
go tool pprof -http=localhost:8080 cpu.pprof

You can also profile manually:

// CPU profile.
file, _ := os.Create("cpu.prof")
defer file.Close()
pprof.StartCPUProfile(file)
defer pprof.StopCPUProfile()
// ...

// Any other profile.
file, _ := os.Create(name + ".prof")
defer file.Close()
pprof.Lookup(name).WriteTo(file, 0)

Tracing records certain types of events while the program is running, mainly those related to concurrency and memory. When the profiling server from the net/http/pprof package is running, call the /debug/pprof/trace endpoint to collect a trace. To view the results, use the go tool trace utility.

You can also collect a trace manually:

file, _ := os.Create("trace.out")
defer file.Close()
trace.Start(file)
defer trace.Stop()
// ...

You can set up automatic tracing with a sliding window that's limited by size or duration. This is called "flight recording". It lets you always keep a recent trace available in case something goes wrong:

cfg := trace.FlightRecorderConfig{
    MinAge:   5 * time.Second,
    MaxBytes: 3 << 20, // 3MB
}
rec := trace.NewFlightRecorder(cfg)
rec.Start()
defer rec.Stop()

# Final thoughts

We've covered a number of Go tools for writing concurrent programs:

• Goroutines for running concurrent tasks.
• Channels and select as flexible communication tools.
• Timers and tickers for working with time.
• Context for canceling operations.
• Wait groups for synchronizing goroutines.
• Mutexes to prevent race conditions.
• Condition variables for signaling events.
• Once for safe one-time initialization.
• Pools to reduce garbage collector load.
• Atomic operations.

If you like the book, please recommend it to your friends or colleagues. If you're interested, check out my other books and projects.

I'm glad you finished the book. Thank you, and I'll see you next time!

• Go developers who want low-level control without having to learn another language.
• C developers who like Go's style.

The new Solod release provides an easy way to call third-party C libraries, makes a large part of the standard library freestanding, and impoves the tooling.

Automatic bindings • Freestanding packages • Type assertions • C interop • Multi-package testing • Checks and targets • Windows • Wrapping up

Automatic binding generator

Sobind generates bindings — stubs for calling third-party C libraries from Solod. It parses .h files and emits a Solod source file with necessary structs, unions, constants, variables, function pointer typedefs, and function declarations.

You can then use the generated types and functions in regular Solod code:

package main

import (
    "solod.dev/raylib/libraylib"
    "solod.dev/so/c"
)

func main() {
    // Using Raylib bindings.
    libraylib.InitWindow(screenWidth, screenHeight, "☀️ Solod / Raylib")
    defer libraylib.CloseWindow()
    // ...
}

Usually, the generated bindings are good enough to use as they are, without any manual changes. I have also prepared bindings for popular C libraries like libuv, raylib, sodium, and sqlite.

Unlike Go, calling C from Solod has zero overhead — Solod code is just regular C in the end.

More freestanding packages

At some point I decided to make as many packages as possible freestanding — independent of any libc implementation or specific OS runtime. That went pretty well. Solod now has 37 standard library packages, and 31 of them work in freestanding mode.

These packages work in freestanding mode with no restrictions:

bufio  bytealg  bytes  c  cmp  encoding  encoding/binary
encoding/hex  encoding/json  errors  io  maps  math/bits
math/rand  mem  path  runtime  slices  strconv  strings
unicode  unicode/utf8  unsafe

These packages work in freestanding mode with certain limitations:

• crypto/crand depends on a user-provided hook to read random bytes.
• fmt depends on a user-provided hook to print formatted text.
• math offers a working subset of features.
• net/netip works fully, except it can't resolve an IPv6 zone name.
• sync/atomic works on targets that support lock-free instructions.
• testing depends on a user-provided hook to print test results.
• time reads the clock using user-provided hooks.
• uuid depends on hooks from both crypto/crand and time.

There's a separate post with more details if you're interested.

Type assertions

A comma-ok type assertion is now fully supported for non-empty interfaces:

var s1 Shape = &rect
r, ok := s1.(*Rect)   // r is &rect, ok is true

var s2 Shape = &circle
c, ok := s2.(*Rect)   // c is nil, ok is false

Which translates to the following C code:

main_Shape s1 = (main_Shape){.self = &rect, .Area = main_Rect_Area};
bool ok = (s1.Area == main_Rect_Area);
main_Rect* r = ok ? (main_Rect*)s1.self : NULL;
// ok == true, r == &rect

main_Shape s2 = (main_Shape){.self = &circle, .Area = main_Circle_Area};
ok = (s2.Area == main_Rect_Area);
main_Rect* c = ok ? (main_Rect*)s2.self : NULL;
// ok == false, c == NULL

Previously, the only two supported forms were a direct assertion like r := s.(*Rect) and a check-only form like _, ok := s.(*Rect).

C interop helpers

The c package now supports more common C types:

size_t      - c.Size
ssize_t     - c.SSize
ptrdiff_t   - c.Ptrdiff
intptr_t    - c.Intptr
long double - c.LongDouble

There's also a c.ConstVoid type, which maps to a C const void. You can use it where C expects a const void* pointer:

// in c
so_ssize_t find_first(const void* items, size_t count, size_t size,
                      bool (*match)(const void*));

// in solod
//so:extern
func find_first(items *c.ConstVoid, count c.Size, size c.Size,
    match func(item *c.ConstVoid) bool) c.SSize

Finally, there are some useful cast functions.

c.Bitcast reads the bits of a value as another type of the same size:

bits := c.Bitcast[uint64](1.0)   // 0x3ff0000000000000
f := c.Bitcast[float64](bits)    // 1.0

You can use c.Bitcast instead of a pointer conversion such as *(*float64)(unsafe.Pointer(&b)).

c.StringData and c.SliceData return a typed pointer to the string or slice data:

b := []byte{1, 2, 3}
p := c.SliceData[c.UChar](b)     // unsigned char*
q := c.StringData[c.UChar]("ab") // unsigned char*

They replace (*T)(unsafe.SliceData(b)) and (*T)(unsafe.StringData(s)).

Multi-package testing

so test can now run tests from multiple packages at once. If you use a pattern that ends with ..., it will select every package that has a test subdirectory under its base directory:

so test ./so/...      # the whole stdlib
so test ./so/net/...  # only the networking packages

The entire run only needs one translation, one compilation, and one execution, which is much faster than running it separately for each package.

The -pkg-file flag restricts the run to only the packages listed in a file:

# freestanding.txt
so/bytes
so/mem
so/time

so test -pkg-file=freestanding.txt ./so/...

Checks and targets

so build, so test, so bench and so run take two new flags: -target and -check.

-target specifies the target platform for cross-compilation. Use the same value that clang and zig cc accept after --target=:

export CC="zig cc"
so build -target=x86_64-windows-gnu -o app.exe .
so build -target=wasm32-freestanding -o main.wasm .

-check enables code analysis:

so test -check=warn .      # -Wall -Wextra -Werror -Wno-shadow -Wno-unused-label
so test -check=sanitize .  # warn + AddressSanitizer + UndefinedBehaviorSanitizer
so test -check=analyze .   # warn + GCC static analyzer

The default optimization level is -O2. You can use CFLAGS to change it.

Limited Windows support

The standard library now builds for windows/amd64 and windows/arm64. All packages in the freestanding set work. Packages that require POSIX (conc, flag, log/slog, net, os, sync) are not supported.

You can use zig cc to cross-compile for Windows:

export CC="zig cc"
export CFLAGS="--target=x86_64-windows-gnu"
export LDFLAGS="-lbcrypt -liphlpapi"
so build -o app.exe .

Not the first-class Windows support that Go offers, but it's better than nothing.

Wrapping up

With v0.4, Solod can work with almost any C library thanks to automatic bindings. The freestanding-aware standard library makes the language a viable option for bare metal programming. Extra interop helpers make C-calling code easy to read, and better tooling keeps tests fast.

There's still a lot to do, of course. In the next release, I plan to focus on the standard library and bring over some hashing and crypto packages from Go. More C library integrations are on the way too!

If you're interested, take a look at Solod's readme — it has everything you need to get started. Or try Solod online without installing anything.

At some point I decided to make as many packages as possible freestanding — independent of any libc implementation or specific OS runtime. That went pretty well. Solod now has 37 standard library packages, and 31 of them work in freestanding mode.

This post describes the techniques I used to get there. There's nothing genuinely novel, and if you're experienced with C, you probably already know all of them. Still, I think it's useful to document the approach — both for me and for anyone interested.

Freestanding mode • Headers • Builtins • Memory • Atomics • Pure C • Allocation • Values • Hooks • Hosted-only • Testing • Final thoughts

Freestanding mode

C has two types of environments. In a hosted environment, you get the full standard library — either the one required by the C standard or, even better, POSIX. In a freestanding environment, you get almost nothing.

The compiler tells you which one you're in:

#if __STDC_HOSTED__
// libc is available
#else
// you're on your own
#endif

Pass -ffreestanding and link with -nostdlib, and that's it: you no longer have printf, or malloc, or even memcpy. There is no entropy source, no file system operations, and no clock. If libc itself is "hard mode", this is "impossible".

Despite its limitations, freestanding mode can be really useful for microcontrollers, WebAssembly sandboxes, kernels, and anything else without an operating system to rely on.

Freestanding headers

Freestanding does not mean "just the C language". The C standard guarantees some headers even without libc, because they define types and macros instead of functions:

float.h   stdalign.h  stdbool.h  stdint.h
limits.h  stdarg.h    stddef.h   ...

Everything that requires actual function implementations is gone:

assert.h  math.h   stdlib.h  time.h
errno.h   stdio.h  string.h  ...

To reflect the hosted/freestanding split, let's introduce builtin.h, a common header included in every standard library package:

#if __STDC_HOSTED__

#include <assert.h>
#include <inttypes.h>
#include <stdalign.h>
#include <stdbool.h>
#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>

#define so_build_hosted

#else

#include <stdbool.h>
#include <stdint.h>
#include <stdalign.h>
#include <stddef.h>

#endif  // __STDC_HOSTED__