From 9d6f71fd2ed0a290781e547c7573e86010ff660f Mon Sep 17 00:00:00 2001
From: Feoramund <161657516+Feoramund@users.noreply.github.com>
Date: Sun, 8 Sep 2024 15:27:28 -0400
Subject: [PATCH 01/23] Fix `sync.Benaphore`

The calls to `atomic_add*` return the value before adding, not after, so
the previous code was causing the occasional data race.
---
 core/sync/extended.odin | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/core/sync/extended.odin b/core/sync/extended.odin
index b446fefa015..ffba40ef898 100644
--- a/core/sync/extended.odin
+++ b/core/sync/extended.odin
@@ -355,7 +355,7 @@ from entering any critical sections associated with the same benaphore, until
 until the lock is released.
 */
 benaphore_lock :: proc "contextless" (b: ^Benaphore) {
-	if atomic_add_explicit(&b.counter, 1, .Acquire) > 1 {
+	if atomic_add_explicit(&b.counter, 1, .Acquire) > 0 {
 		sema_wait(&b.sema)
 	}
 }
@@ -384,7 +384,7 @@ are waiting on the lock, exactly one thread is allowed into a critical section
 associated with the same banaphore.
 */
 benaphore_unlock :: proc "contextless" (b: ^Benaphore) {
-	if atomic_sub_explicit(&b.counter, 1, .Release) > 0 {
+	if atomic_sub_explicit(&b.counter, 1, .Release) > 1 {
 		sema_post(&b.sema)
 	}
 }
@@ -740,4 +740,4 @@ Make event available.
 one_shot_event_signal :: proc "contextless" (e: ^One_Shot_Event) {
 	atomic_store_explicit(&e.state, 1, .Release)
 	futex_broadcast(&e.state)
-}
\ No newline at end of file
+}

From 74b28f1ff91d4776475f4009fa2bcda71c655cd5 Mon Sep 17 00:00:00 2001
From: Feoramund <161657516+Feoramund@users.noreply.github.com>
Date: Sun, 8 Sep 2024 17:25:48 -0400
Subject: [PATCH 02/23] Fix rare double-join possibility in POSIX
 `thread._join`

This was occuring about 1/100 times with the test runner's thread pool.
---
 core/thread/thread_unix.odin | 16 ++++------------
 1 file changed, 4 insertions(+), 12 deletions(-)

diff --git a/core/thread/thread_unix.odin b/core/thread/thread_unix.odin
index ddc47244c16..d165560aca4 100644
--- a/core/thread/thread_unix.odin
+++ b/core/thread/thread_unix.odin
@@ -9,8 +9,6 @@ import "core:time"
 
 _IS_SUPPORTED :: true
 
-CAS :: sync.atomic_compare_exchange_strong
-
 // NOTE(tetra): Aligned here because of core/unix/pthread_linux.odin/pthread_t.
 // Also see core/sys/darwin/mach_darwin.odin/semaphore_t.
 Thread_Os_Specific :: struct #align(16) {
@@ -140,24 +138,18 @@ _is_done :: proc(t: ^Thread) -> bool {
 }
 
 _join :: proc(t: ^Thread) {
-	// sync.guard(&t.mutex)
-
 	if unix.pthread_equal(unix.pthread_self(), t.unix_thread) {
 		return
 	}
 
-	// Preserve other flags besides `.Joined`, like `.Started`.
-	unjoined := sync.atomic_load(&t.flags) - {.Joined}
-	joined   := unjoined + {.Joined}
-
-	// Try to set `t.flags` from unjoined to joined. If it returns joined,
-	// it means the previous value had that flag set and we can return.
-	if res, ok := CAS(&t.flags, unjoined, joined); res == joined && !ok {
+	// If the previous value was already `Joined`, then we can return.
+	if .Joined in sync.atomic_or(&t.flags, {.Joined}) {
 		return
 	}
+
 	// Prevent non-started threads from blocking main thread with initial wait
 	// condition.
-	if .Started not_in unjoined {
+	if .Started not_in sync.atomic_load(&t.flags) {
 		_start(t)
 	}
 	unix.pthread_join(t.unix_thread, nil)

From cbd4d5e765646ef07c4133ab65e06652a87a1916 Mon Sep 17 00:00:00 2001
From: Feoramund <161657516+Feoramund@users.noreply.github.com>
Date: Sun, 8 Sep 2024 17:54:45 -0400
Subject: [PATCH 03/23] Fix data race in `atomic_sema_wait_with_timeout`

---
 core/sync/primitives_atomic.odin | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/core/sync/primitives_atomic.odin b/core/sync/primitives_atomic.odin
index 1d8e423db72..076a74b2024 100644
--- a/core/sync/primitives_atomic.odin
+++ b/core/sync/primitives_atomic.odin
@@ -361,7 +361,7 @@ atomic_sema_wait_with_timeout :: proc "contextless" (s: ^Atomic_Sema, duration:
 			if !futex_wait_with_timeout(&s.count, u32(original_count), remaining) {
 				return false
 			}
-			original_count = s.count
+			original_count = atomic_load_explicit(&s.count, .Relaxed)
 		}
 		if original_count == atomic_compare_exchange_strong_explicit(&s.count, original_count, original_count-1, .Acquire, .Acquire) {
 			return true

From 4d14b4257e7570216826c5cbcee94aa51116e3b3 Mon Sep 17 00:00:00 2001
From: Feoramund <161657516+Feoramund@users.noreply.github.com>
Date: Sun, 8 Sep 2024 18:05:34 -0400
Subject: [PATCH 04/23] Convert POSIX `Thread` to use semaphore instead

One less value to store, and it should be less of a hack too.

Semaphores will not wait around if they have the go-ahead; they depend
on an internal value being non-zero, instead of whatever was loaded when
they started waiting, which is the case with a `Cond`.
---
 core/thread/thread_unix.odin | 16 ++++------------
 1 file changed, 4 insertions(+), 12 deletions(-)

diff --git a/core/thread/thread_unix.odin b/core/thread/thread_unix.odin
index d165560aca4..3d3b419b096 100644
--- a/core/thread/thread_unix.odin
+++ b/core/thread/thread_unix.odin
@@ -5,7 +5,6 @@ package thread
 import "base:runtime"
 import "core:sync"
 import "core:sys/unix"
-import "core:time"
 
 _IS_SUPPORTED :: true
 
@@ -13,8 +12,7 @@ _IS_SUPPORTED :: true
 // Also see core/sys/darwin/mach_darwin.odin/semaphore_t.
 Thread_Os_Specific :: struct #align(16) {
 	unix_thread: unix.pthread_t, // NOTE: very large on Darwin, small on Linux.
-	cond:        sync.Cond,
-	mutex:       sync.Mutex,
+	start_ok:    sync.Sema,
 }
 //
 // Creates a thread which will run the given procedure.
@@ -27,14 +25,10 @@ _create :: proc(procedure: Thread_Proc, priority: Thread_Priority) -> ^Thread {
 		// We need to give the thread a moment to start up before we enable cancellation.
 		can_set_thread_cancel_state := unix.pthread_setcancelstate(unix.PTHREAD_CANCEL_ENABLE, nil) == 0
 
-		sync.lock(&t.mutex)
-
 		t.id = sync.current_thread_id()
 
-		for (.Started not_in sync.atomic_load(&t.flags)) {
-			// HACK: use a timeout so in the event that the condition is signalled at THIS comment's exact point
-			// (after checking flags, before starting the wait) it gets itself out of that deadlock after a ms.
-			sync.wait_with_timeout(&t.cond, &t.mutex, time.Millisecond)
+		if .Started not_in sync.atomic_load(&t.flags) {
+			sync.wait(&t.start_ok)
 		}
 
 		if .Joined in sync.atomic_load(&t.flags) {
@@ -64,8 +58,6 @@ _create :: proc(procedure: Thread_Proc, priority: Thread_Priority) -> ^Thread {
 
 		sync.atomic_or(&t.flags, { .Done })
 
-		sync.unlock(&t.mutex)
-
 		if .Self_Cleanup in sync.atomic_load(&t.flags) {
 			res := unix.pthread_detach(t.unix_thread)
 			assert_contextless(res == 0)
@@ -130,7 +122,7 @@ _create :: proc(procedure: Thread_Proc, priority: Thread_Priority) -> ^Thread {
 
 _start :: proc(t: ^Thread) {
 	sync.atomic_or(&t.flags, { .Started })
-	sync.signal(&t.cond)
+	sync.post(&t.start_ok)
 }
 
 _is_done :: proc(t: ^Thread) -> bool {

From 45da0093774276223e3724a89e5b0a9f8ef7c9f7 Mon Sep 17 00:00:00 2001
From: Feoramund <161657516+Feoramund@users.noreply.github.com>
Date: Sun, 8 Sep 2024 18:21:55 -0400
Subject: [PATCH 05/23] Use more atomic handling of thread flags

This can prevent a data race on Linux with `Self_Cleanup`.
---
 core/thread/thread.odin         | 12 ++++++------
 core/thread/thread_windows.odin |  6 +++---
 2 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/core/thread/thread.odin b/core/thread/thread.odin
index 17ba1a0a2c2..c1cbceb42c3 100644
--- a/core/thread/thread.odin
+++ b/core/thread/thread.odin
@@ -272,7 +272,7 @@ create_and_start :: proc(fn: proc(), init_context: Maybe(runtime.Context) = nil,
 	t := create(thread_proc, priority)
 	t.data = rawptr(fn)
 	if self_cleanup {
-		t.flags += {.Self_Cleanup}
+		intrinsics.atomic_or(&t.flags, {.Self_Cleanup})
 	}
 	t.init_context = init_context
 	start(t)
@@ -307,7 +307,7 @@ create_and_start_with_data :: proc(data: rawptr, fn: proc(data: rawptr), init_co
 	t.user_index = 1
 	t.user_args[0] = data
 	if self_cleanup {
-		t.flags += {.Self_Cleanup}
+		intrinsics.atomic_or(&t.flags, {.Self_Cleanup})
 	}
 	t.init_context = init_context
 	start(t)
@@ -347,7 +347,7 @@ create_and_start_with_poly_data :: proc(data: $T, fn: proc(data: T), init_contex
 	mem.copy(&t.user_args[0], &data, size_of(T))
 
 	if self_cleanup {
-		t.flags += {.Self_Cleanup}
+		intrinsics.atomic_or(&t.flags, {.Self_Cleanup})
 	}
 
 	t.init_context = init_context
@@ -394,7 +394,7 @@ create_and_start_with_poly_data2 :: proc(arg1: $T1, arg2: $T2, fn: proc(T1, T2),
 	_  = copy(user_args[n:], mem.ptr_to_bytes(&arg2))
 
 	if self_cleanup {
-		t.flags += {.Self_Cleanup}
+		intrinsics.atomic_or(&t.flags, {.Self_Cleanup})
 	}
 
 	t.init_context = init_context
@@ -443,7 +443,7 @@ create_and_start_with_poly_data3 :: proc(arg1: $T1, arg2: $T2, arg3: $T3, fn: pr
 	_  = copy(user_args[n:], mem.ptr_to_bytes(&arg3))
 
 	if self_cleanup {
-		t.flags += {.Self_Cleanup}
+		intrinsics.atomic_or(&t.flags, {.Self_Cleanup})
 	}
 
 	t.init_context = init_context
@@ -494,7 +494,7 @@ create_and_start_with_poly_data4 :: proc(arg1: $T1, arg2: $T2, arg3: $T3, arg4:
 	_  = copy(user_args[n:], mem.ptr_to_bytes(&arg4))
 
 	if self_cleanup {
-		t.flags += {.Self_Cleanup}
+		intrinsics.atomic_or(&t.flags, {.Self_Cleanup})
 	}
 
 	t.init_context = init_context
diff --git a/core/thread/thread_windows.odin b/core/thread/thread_windows.odin
index 50a4e5fbc10..22c3eae6589 100644
--- a/core/thread/thread_windows.odin
+++ b/core/thread/thread_windows.odin
@@ -27,7 +27,7 @@ _create :: proc(procedure: Thread_Proc, priority: Thread_Priority) -> ^Thread {
 	__windows_thread_entry_proc :: proc "system" (t_: rawptr) -> win32.DWORD {
 		t := (^Thread)(t_)
 
-		if .Joined in t.flags {
+		if .Joined in sync.atomic_load(&t.flags) {
 			return 0
 		}
 
@@ -48,9 +48,9 @@ _create :: proc(procedure: Thread_Proc, priority: Thread_Priority) -> ^Thread {
 			t.procedure(t)
 		}
 
-		intrinsics.atomic_store(&t.flags, t.flags + {.Done})
+		intrinsics.atomic_or(&t.flags, {.Done})
 
-		if .Self_Cleanup in t.flags {
+		if .Self_Cleanup in sync.atomic_load(&t.flags) {
 			win32.CloseHandle(t.win32_thread)
 			t.win32_thread = win32.INVALID_HANDLE
 			// NOTE(ftphikari): It doesn't matter which context 'free' received, right?

From dbb783fbf20df1bba899b7a2bcbd65f71eb32fef Mon Sep 17 00:00:00 2001
From: Feoramund <161657516+Feoramund@users.noreply.github.com>
Date: Sun, 8 Sep 2024 18:59:55 -0400
Subject: [PATCH 06/23] Fix atomic memory order for `sync.ticket_mutex_unlock`

---
 core/sync/extended.odin | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/core/sync/extended.odin b/core/sync/extended.odin
index ffba40ef898..fd2bda08ac0 100644
--- a/core/sync/extended.odin
+++ b/core/sync/extended.odin
@@ -297,7 +297,7 @@ waiting to acquire the lock, exactly one of those threads is unblocked and
 allowed into the critical section.
 */
 ticket_mutex_unlock :: #force_inline proc "contextless" (m: ^Ticket_Mutex) {
-	atomic_add_explicit(&m.serving, 1, .Relaxed)
+	atomic_add_explicit(&m.serving, 1, .Release)
 }
 
 /*

From c3f363cfbcee453c7d90b37429c92115e91216af Mon Sep 17 00:00:00 2001
From: Feoramund <161657516+Feoramund@users.noreply.github.com>
Date: Sun, 8 Sep 2024 21:59:55 -0400
Subject: [PATCH 07/23] Fix data race when `pool_stop_task` is called

---
 core/thread/thread_pool.odin | 1 +
 1 file changed, 1 insertion(+)

diff --git a/core/thread/thread_pool.odin b/core/thread/thread_pool.odin
index 9bcc42968f0..d9166b45090 100644
--- a/core/thread/thread_pool.odin
+++ b/core/thread/thread_pool.odin
@@ -60,6 +60,7 @@ pool_thread_runner :: proc(t: ^Thread) {
 		if task, ok := pool_pop_waiting(pool); ok {
 			data.task = task
 			pool_do_work(pool, task)
+			sync.guard(&pool.mutex)
 			data.task = {}
 		}
 	}

From 0a594147afbfdacece1d221d2dee744e612362c6 Mon Sep 17 00:00:00 2001
From: Feoramund <161657516+Feoramund@users.noreply.github.com>
Date: Mon, 9 Sep 2024 00:23:30 -0400
Subject: [PATCH 08/23] Use `contextless` procs in `core:sync` instead

---
 core/sync/extended.odin          | 10 +++++-----
 core/sync/futex_darwin.odin      | 12 ++++++------
 core/sync/futex_freebsd.odin     |  8 ++++----
 core/sync/futex_linux.odin       |  8 ++++----
 core/sync/futex_netbsd.odin      |  8 ++++----
 core/sync/futex_openbsd.odin     |  8 ++++----
 core/sync/futex_wasm.odin        |  8 ++++----
 core/sync/primitives.odin        | 18 +-----------------
 core/sync/primitives_atomic.odin |  2 +-
 9 files changed, 33 insertions(+), 49 deletions(-)

diff --git a/core/sync/extended.odin b/core/sync/extended.odin
index fd2bda08ac0..83cc648b477 100644
--- a/core/sync/extended.odin
+++ b/core/sync/extended.odin
@@ -48,12 +48,12 @@ wait_group_add :: proc "contextless" (wg: ^Wait_Group, delta: int) {
 
 	atomic_add(&wg.counter, delta)
 	if wg.counter < 0 {
-		_panic("sync.Wait_Group negative counter")
+		panic_contextless("sync.Wait_Group negative counter")
 	}
 	if wg.counter == 0 {
 		cond_broadcast(&wg.cond)
 		if wg.counter != 0 {
-			_panic("sync.Wait_Group misuse: sync.wait_group_add called concurrently with sync.wait_group_wait")
+			panic_contextless("sync.Wait_Group misuse: sync.wait_group_add called concurrently with sync.wait_group_wait")
 		}
 	}
 }
@@ -81,7 +81,7 @@ wait_group_wait :: proc "contextless" (wg: ^Wait_Group) {
 	if wg.counter != 0 {
 		cond_wait(&wg.cond, &wg.mutex)
 		if wg.counter != 0 {
-			_panic("sync.Wait_Group misuse: sync.wait_group_add called concurrently with sync.wait_group_wait")
+			panic_contextless("sync.Wait_Group misuse: sync.wait_group_add called concurrently with sync.wait_group_wait")
 		}
 	}
 }
@@ -105,7 +105,7 @@ wait_group_wait_with_timeout :: proc "contextless" (wg: ^Wait_Group, duration: t
 			return false
 		}
 		if wg.counter != 0 {
-			_panic("sync.Wait_Group misuse: sync.wait_group_add called concurrently with sync.wait_group_wait")
+			panic_contextless("sync.Wait_Group misuse: sync.wait_group_add called concurrently with sync.wait_group_wait")
 		}
 	}
 	return true
@@ -494,7 +494,7 @@ for other threads for entering.
 */
 recursive_benaphore_unlock :: proc "contextless" (b: ^Recursive_Benaphore) {
 	tid := current_thread_id()
-	_assert(tid == b.owner, "tid != b.owner")
+	assert_contextless(tid == b.owner, "tid != b.owner")
 	b.recursion -= 1
 	recursion := b.recursion
 	if recursion == 0 {
diff --git a/core/sync/futex_darwin.odin b/core/sync/futex_darwin.odin
index fca9aadfe1f..daefd669992 100644
--- a/core/sync/futex_darwin.odin
+++ b/core/sync/futex_darwin.odin
@@ -48,7 +48,7 @@ _futex_wait_with_timeout :: proc "contextless" (f: ^Futex, expected: u32, durati
 		case -ETIMEDOUT:
 			return false
 		case:
-			_panic("darwin.os_sync_wait_on_address_with_timeout failure")
+			panic_contextless("darwin.os_sync_wait_on_address_with_timeout failure")
 		}
 	} else {
 
@@ -63,7 +63,7 @@ _futex_wait_with_timeout :: proc "contextless" (f: ^Futex, expected: u32, durati
 	case ETIMEDOUT:
 		return false
 	case:
-		_panic("futex_wait failure")
+		panic_contextless("futex_wait failure")
 	}
 	return true
 
@@ -83,7 +83,7 @@ _futex_signal :: proc "contextless" (f: ^Futex) {
 			case -ENOENT:
 				return
 			case:
-				_panic("darwin.os_sync_wake_by_address_any failure")
+				panic_contextless("darwin.os_sync_wake_by_address_any failure")
 			}
 		}
 	} else {
@@ -99,7 +99,7 @@ _futex_signal :: proc "contextless" (f: ^Futex) {
 		case ENOENT:
 			return
 		case:
-			_panic("futex_wake_single failure")
+			panic_contextless("futex_wake_single failure")
 		}
 	}
 
@@ -119,7 +119,7 @@ _futex_broadcast :: proc "contextless" (f: ^Futex) {
 			case -ENOENT:
 				return
 			case:
-				_panic("darwin.os_sync_wake_by_address_all failure")
+				panic_contextless("darwin.os_sync_wake_by_address_all failure")
 			}
 		}
 	} else {
@@ -135,7 +135,7 @@ _futex_broadcast :: proc "contextless" (f: ^Futex) {
 		case ENOENT:
 			return
 		case:
-			_panic("futex_wake_all failure")
+			panic_contextless("futex_wake_all failure")
 		}
 	}
 
diff --git a/core/sync/futex_freebsd.odin b/core/sync/futex_freebsd.odin
index ac6e2400a4f..82021a71a09 100644
--- a/core/sync/futex_freebsd.odin
+++ b/core/sync/futex_freebsd.odin
@@ -21,7 +21,7 @@ _futex_wait :: proc "contextless" (f: ^Futex, expected: u32) -> bool {
 			continue
 		}
 
-		_panic("_futex_wait failure")
+		panic_contextless("_futex_wait failure")
 	}
 
 	unreachable()
@@ -44,14 +44,14 @@ _futex_wait_with_timeout :: proc "contextless" (f: ^Futex, expected: u32, durati
 		return false
 	}
 
-	_panic("_futex_wait_with_timeout failure")
+	panic_contextless("_futex_wait_with_timeout failure")
 }
 
 _futex_signal :: proc "contextless" (f: ^Futex) {
 	errno := freebsd._umtx_op(f, .WAKE, 1, nil, nil)
 
 	if errno != nil {
-		_panic("_futex_signal failure")
+		panic_contextless("_futex_signal failure")
 	}
 }
 
@@ -59,6 +59,6 @@ _futex_broadcast :: proc "contextless" (f: ^Futex)  {
 	errno := freebsd._umtx_op(f, .WAKE, cast(c.ulong)max(i32), nil, nil)
 
 	if errno != nil {
-		_panic("_futex_broadcast failure")
+		panic_contextless("_futex_broadcast failure")
 	}
 }
diff --git a/core/sync/futex_linux.odin b/core/sync/futex_linux.odin
index fe57c12eda3..4d9101b9ff7 100644
--- a/core/sync/futex_linux.odin
+++ b/core/sync/futex_linux.odin
@@ -15,7 +15,7 @@ _futex_wait :: proc "contextless" (futex: ^Futex, expected: u32) -> bool {
 		return true
 	case:
 		// TODO(flysand): More descriptive panic messages based on the vlaue of `errno`
-		_panic("futex_wait failure")
+		panic_contextless("futex_wait failure")
 	}
 }
 
@@ -34,7 +34,7 @@ _futex_wait_with_timeout :: proc "contextless" (futex: ^Futex, expected: u32, du
 	case .NONE, .EINTR, .EAGAIN:
 		return true
 	case:
-		_panic("futex_wait_with_timeout failure")
+		panic_contextless("futex_wait_with_timeout failure")
 	}
 }
 
@@ -44,7 +44,7 @@ _futex_signal :: proc "contextless" (futex: ^Futex) {
 	case .NONE:
 		return
 	case:
-		_panic("futex_wake_single failure")
+		panic_contextless("futex_wake_single failure")
 	}
 }
 
@@ -57,6 +57,6 @@ _futex_broadcast :: proc "contextless" (futex: ^Futex)  {
 	case .NONE:
 		return
 	case:
-		_panic("_futex_wake_all failure")
+		panic_contextless("_futex_wake_all failure")
 	}
 }
diff --git a/core/sync/futex_netbsd.odin b/core/sync/futex_netbsd.odin
index d12409f3278..f81a126757b 100644
--- a/core/sync/futex_netbsd.odin
+++ b/core/sync/futex_netbsd.odin
@@ -35,7 +35,7 @@ _futex_wait :: proc "contextless" (futex: ^Futex, expected: u32) -> bool {
 		case EINTR, EAGAIN:
 			return true
 		case:
-			_panic("futex_wait failure")
+			panic_contextless("futex_wait failure")
 		}	
 	}
 	return true
@@ -55,7 +55,7 @@ _futex_wait_with_timeout :: proc "contextless" (futex: ^Futex, expected: u32, du
 		case ETIMEDOUT:
 			return false
 		case:
-			_panic("futex_wait_with_timeout failure")
+			panic_contextless("futex_wait_with_timeout failure")
 		}
 	}
 	return true
@@ -63,12 +63,12 @@ _futex_wait_with_timeout :: proc "contextless" (futex: ^Futex, expected: u32, du
 
 _futex_signal :: proc "contextless" (futex: ^Futex) {
 	if _, ok := intrinsics.syscall_bsd(unix.SYS___futex, uintptr(futex), FUTEX_WAKE_PRIVATE, 1, 0, 0, 0); !ok {
-		_panic("futex_wake_single failure")
+		panic_contextless("futex_wake_single failure")
 	}
 }
 
 _futex_broadcast :: proc "contextless" (futex: ^Futex)  {
 	if _, ok := intrinsics.syscall_bsd(unix.SYS___futex, uintptr(futex), FUTEX_WAKE_PRIVATE, uintptr(max(i32)), 0, 0, 0); !ok {
-		_panic("_futex_wake_all failure")
+		panic_contextless("_futex_wake_all failure")
 	}
 }
diff --git a/core/sync/futex_openbsd.odin b/core/sync/futex_openbsd.odin
index 4883a084145..1ffe4a9a562 100644
--- a/core/sync/futex_openbsd.odin
+++ b/core/sync/futex_openbsd.odin
@@ -36,7 +36,7 @@ _futex_wait :: proc "contextless" (f: ^Futex, expected: u32) -> bool {
 		return false
 	}
 
-	_panic("futex_wait failure")
+	panic_contextless("futex_wait failure")
 }
 
 _futex_wait_with_timeout :: proc "contextless" (f: ^Futex, expected: u32, duration: time.Duration) -> bool {
@@ -62,14 +62,14 @@ _futex_wait_with_timeout :: proc "contextless" (f: ^Futex, expected: u32, durati
 		return false
 	}
 
-	_panic("futex_wait_with_timeout failure")
+	panic_contextless("futex_wait_with_timeout failure")
 }
 
 _futex_signal :: proc "contextless" (f: ^Futex) {
 	res := _unix_futex(f, FUTEX_WAKE_PRIVATE, 1, nil)
 
 	if res == -1 {
-		_panic("futex_wake_single failure")
+		panic_contextless("futex_wake_single failure")
 	}
 }
 
@@ -77,6 +77,6 @@ _futex_broadcast :: proc "contextless" (f: ^Futex)  {
 	res := _unix_futex(f, FUTEX_WAKE_PRIVATE, u32(max(i32)), nil)
 
 	if res == -1 {
-		_panic("_futex_wake_all failure")
+		panic_contextless("_futex_wake_all failure")
 	}
 }
diff --git a/core/sync/futex_wasm.odin b/core/sync/futex_wasm.odin
index de88e8198a1..27532587c5b 100644
--- a/core/sync/futex_wasm.odin
+++ b/core/sync/futex_wasm.odin
@@ -10,7 +10,7 @@ import "core:time"
 
 _futex_wait :: proc "contextless" (f: ^Futex, expected: u32) -> bool {
 	when !intrinsics.has_target_feature("atomics") {
-		_panic("usage of `core:sync` requires the `-target-feature:\"atomics\"` or a `-microarch` that supports it")
+		panic_contextless("usage of `core:sync` requires the `-target-feature:\"atomics\"` or a `-microarch` that supports it")
 	} else {
 		s := intrinsics.wasm_memory_atomic_wait32((^u32)(f), expected, -1)
 		return s != 0
@@ -19,7 +19,7 @@ _futex_wait :: proc "contextless" (f: ^Futex, expected: u32) -> bool {
 
 _futex_wait_with_timeout :: proc "contextless" (f: ^Futex, expected: u32, duration: time.Duration) -> bool {
 	when !intrinsics.has_target_feature("atomics") {
-		_panic("usage of `core:sync` requires the `-target-feature:\"atomics\"` or a `-microarch` that supports it")
+		panic_contextless("usage of `core:sync` requires the `-target-feature:\"atomics\"` or a `-microarch` that supports it")
 	} else {
 		s := intrinsics.wasm_memory_atomic_wait32((^u32)(f), expected, i64(duration))
 		return s != 0
@@ -28,7 +28,7 @@ _futex_wait_with_timeout :: proc "contextless" (f: ^Futex, expected: u32, durati
 
 _futex_signal :: proc "contextless" (f: ^Futex) {
 	when !intrinsics.has_target_feature("atomics") {
-		_panic("usage of `core:sync` requires the `-target-feature:\"atomics\"` or a `-microarch` that supports it")
+		panic_contextless("usage of `core:sync` requires the `-target-feature:\"atomics\"` or a `-microarch` that supports it")
 	} else {
 		loop: for {
 			s := intrinsics.wasm_memory_atomic_notify32((^u32)(f), 1)
@@ -41,7 +41,7 @@ _futex_signal :: proc "contextless" (f: ^Futex) {
 
 _futex_broadcast :: proc "contextless" (f: ^Futex) {
 	when !intrinsics.has_target_feature("atomics") {
-		_panic("usage of `core:sync` requires the `-target-feature:\"atomics\"` or a `-microarch` that supports it")
+		panic_contextless("usage of `core:sync` requires the `-target-feature:\"atomics\"` or a `-microarch` that supports it")
 	} else {
 		loop: for {
 			s := intrinsics.wasm_memory_atomic_notify32((^u32)(f), ~u32(0))
diff --git a/core/sync/primitives.odin b/core/sync/primitives.odin
index a2282448115..8187c904bbd 100644
--- a/core/sync/primitives.odin
+++ b/core/sync/primitives.odin
@@ -1,6 +1,5 @@
 package sync
 
-import "base:runtime"
 import "core:time"
 
 /*
@@ -560,7 +559,7 @@ futex_wait :: proc "contextless" (f: ^Futex, expected: u32) {
 		return
 	}
 	ok := _futex_wait(f, expected)
-	_assert(ok, "futex_wait failure")
+	assert_contextless(ok, "futex_wait failure")
 }
 
 /*
@@ -597,18 +596,3 @@ Wake up multiple threads waiting on a futex.
 futex_broadcast :: proc "contextless" (f: ^Futex) {
 	_futex_broadcast(f)
 }
-
-
-@(private)
-_assert :: proc "contextless" (cond: bool, msg: string) {
-	if !cond {
-		_panic(msg)
-	}
-}
-
-@(private)
-_panic :: proc "contextless" (msg: string) -> ! {
-	runtime.print_string(msg)
-	runtime.print_byte('\n')
-	runtime.trap()
-}
diff --git a/core/sync/primitives_atomic.odin b/core/sync/primitives_atomic.odin
index 076a74b2024..3c4324eb7f8 100644
--- a/core/sync/primitives_atomic.odin
+++ b/core/sync/primitives_atomic.odin
@@ -240,7 +240,7 @@ atomic_recursive_mutex_lock :: proc "contextless" (m: ^Atomic_Recursive_Mutex) {
 
 atomic_recursive_mutex_unlock :: proc "contextless" (m: ^Atomic_Recursive_Mutex) {
 	tid := current_thread_id()
-	_assert(tid == m.owner, "tid != m.owner")
+	assert_contextless(tid == m.owner, "tid != m.owner")
 	m.recursion -= 1
 	recursion := m.recursion
 	if recursion == 0 {

From 73f5ab473c4129ae209838d7967286684ac3f462 Mon Sep 17 00:00:00 2001
From: Feoramund <161657516+Feoramund@users.noreply.github.com>
Date: Mon, 9 Sep 2024 14:18:01 -0400
Subject: [PATCH 09/23] Keep `chan.can_recv` from deadlocking

---
 core/sync/chan/chan.odin | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/core/sync/chan/chan.odin b/core/sync/chan/chan.odin
index 53a3bff4b69..aca08d82ead 100644
--- a/core/sync/chan/chan.odin
+++ b/core/sync/chan/chan.odin
@@ -423,7 +423,7 @@ raw_queue_pop :: proc "contextless" (q: ^Raw_Queue) -> (data: rawptr) {
 can_recv :: proc "contextless" (c: ^Raw_Chan) -> bool {
 	sync.guard(&c.mutex)
 	if is_buffered(c) {
-		return len(c) > 0
+		return c.queue.len > 0
 	}
 	return sync.atomic_load(&c.w_waiting) > 0
 }

From 026aef69e3a42021c5d9666737c7401dc75dc89a Mon Sep 17 00:00:00 2001
From: Feoramund <161657516+Feoramund@users.noreply.github.com>
Date: Mon, 9 Sep 2024 14:42:50 -0400
Subject: [PATCH 10/23] Fix deadlock on sending to full, buffered, closed
 `Chan`

This will also keep messages from being sent to closed, buffered
channels in general.
---
 core/sync/chan/chan.odin | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/core/sync/chan/chan.odin b/core/sync/chan/chan.odin
index aca08d82ead..cb299f23ffd 100644
--- a/core/sync/chan/chan.odin
+++ b/core/sync/chan/chan.odin
@@ -164,12 +164,17 @@ send_raw :: proc "contextless" (c: ^Raw_Chan, msg_in: rawptr) -> (ok: bool) {
 	}
 	if c.queue != nil { // buffered
 		sync.guard(&c.mutex)
-		for c.queue.len == c.queue.cap {
+		for !sync.atomic_load(&c.closed) &&
+		    c.queue.len == c.queue.cap {
 			sync.atomic_add(&c.w_waiting, 1)
 			sync.wait(&c.w_cond, &c.mutex)
 			sync.atomic_sub(&c.w_waiting, 1)
 		}
 
+		if sync.atomic_load(&c.closed) {
+			return false
+		}
+
 		ok = raw_queue_push(c.queue, msg_in)
 		if sync.atomic_load(&c.r_waiting) > 0 {
 			sync.signal(&c.r_cond)

From e9a6a344809daf5c0a3b725dd52e1527382d8c41 Mon Sep 17 00:00:00 2001
From: Feoramund <161657516+Feoramund@users.noreply.github.com>
Date: Mon, 9 Sep 2024 16:04:18 -0400
Subject: [PATCH 11/23] Forbid `chan.try_send` on closed buffered channels

---
 core/sync/chan/chan.odin | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/core/sync/chan/chan.odin b/core/sync/chan/chan.odin
index cb299f23ffd..f0b04f3b445 100644
--- a/core/sync/chan/chan.odin
+++ b/core/sync/chan/chan.odin
@@ -260,6 +260,10 @@ try_send_raw :: proc "contextless" (c: ^Raw_Chan, msg_in: rawptr) -> (ok: bool)
 			return false
 		}
 
+		if sync.atomic_load(&c.closed) {
+			return false
+		}
+
 		ok = raw_queue_push(c.queue, msg_in)
 		if sync.atomic_load(&c.r_waiting) > 0 {
 			sync.signal(&c.r_cond)

From 8a14a656fbd8843628b61e5a20877b40b772482c Mon Sep 17 00:00:00 2001
From: Feoramund <161657516+Feoramund@users.noreply.github.com>
Date: Mon, 9 Sep 2024 16:05:29 -0400
Subject: [PATCH 12/23] Fix `chan.can_send` for unbuffered channels

`w_waiting` is the signal that says a caller is waiting to be able to
send something. It is incremented upon send and - in the case of an
unbuffered channel - it can only hold one message.

Therefore, check that `w_waiting` is zero instead.
---
 core/sync/chan/chan.odin | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/core/sync/chan/chan.odin b/core/sync/chan/chan.odin
index f0b04f3b445..5b9a764b47a 100644
--- a/core/sync/chan/chan.odin
+++ b/core/sync/chan/chan.odin
@@ -444,7 +444,7 @@ can_send :: proc "contextless" (c: ^Raw_Chan) -> bool {
 	if is_buffered(c) {
 		return c.queue.len < c.queue.cap
 	}
-	return sync.atomic_load(&c.r_waiting) > 0
+	return sync.atomic_load(&c.w_waiting) == 0
 }
 
 

From 074314b8874886f8956e9a7e9a773ac059051030 Mon Sep 17 00:00:00 2001
From: Feoramund <161657516+Feoramund@users.noreply.github.com>
Date: Mon, 9 Sep 2024 18:38:29 -0400
Subject: [PATCH 13/23] Fix data race in `test_core_flags`

---
 tests/core/flags/test_core_flags.odin | 33 ++++++++++++++++-----------
 1 file changed, 20 insertions(+), 13 deletions(-)

diff --git a/tests/core/flags/test_core_flags.odin b/tests/core/flags/test_core_flags.odin
index f8305f2ead9..8fcd6a8a72c 100644
--- a/tests/core/flags/test_core_flags.odin
+++ b/tests/core/flags/test_core_flags.odin
@@ -12,6 +12,26 @@ import "core:strings"
 import "core:testing"
 import "core:time/datetime"
 
+Custom_Data :: struct {
+	a: int,
+}
+
+@(init)
+init_custom_type_setter :: proc() {
+	// NOTE: This is done here so it can be out of the flow of the
+	// multi-threaded test runner, to prevent any data races that could be
+	// reported by using `-sanitize:thread`.
+	//
+	// Do mind that this means every test here acknowledges the `Custom_Data` type.
+	flags.register_type_setter(proc (data: rawptr, data_type: typeid, _, _: string) -> (string, bool, runtime.Allocator_Error) {
+		if data_type == Custom_Data {
+			(cast(^Custom_Data)data).a = 32
+			return "", true, nil
+		}
+		return "", false, nil
+	})
+}
+
 @(test)
 test_no_args :: proc(t: ^testing.T) {
 	S :: struct {
@@ -1230,9 +1250,6 @@ test_net :: proc(t: ^testing.T) {
 @(test)
 test_custom_type_setter :: proc(t: ^testing.T) {
 	Custom_Bool :: distinct bool
-	Custom_Data :: struct {
-		a: int,
-	}
 
 	S :: struct {
 		a: Custom_Data,
@@ -1240,16 +1257,6 @@ test_custom_type_setter :: proc(t: ^testing.T) {
 	}
 	s: S
 
-	// NOTE: Mind that this setter is global state, and the test runner is multi-threaded.
-	// It should be fine so long as all type setter tests are in this one test proc.
-	flags.register_type_setter(proc (data: rawptr, data_type: typeid, _, _: string) -> (string, bool, runtime.Allocator_Error) {
-		if data_type == Custom_Data {
-			(cast(^Custom_Data)data).a = 32
-			return "", true, nil
-		}
-		return "", false, nil
-	})
-	defer flags.register_type_setter(nil)
 	args := [?]string { "-a:hellope", "-b:true" }
 	result := flags.parse(&s, args[:])
 	testing.expect_value(t, result, nil)

From 3a6010918033e1548e84d57a07074cdbf802ff9b Mon Sep 17 00:00:00 2001
From: Feoramund <161657516+Feoramund@users.noreply.github.com>
Date: Mon, 9 Sep 2024 19:09:16 -0400
Subject: [PATCH 14/23] Fix signalling test child threads crashing test 0

A thread made inside a test does not share the test index of its parent,
so any time one of those threads failed an assert, it would tell the
runner to shutdown test index zero.
---
 core/testing/signal_handler_libc.odin | 11 +++++++++++
 1 file changed, 11 insertions(+)

diff --git a/core/testing/signal_handler_libc.odin b/core/testing/signal_handler_libc.odin
index 27d1a073503..e2d17059534 100644
--- a/core/testing/signal_handler_libc.odin
+++ b/core/testing/signal_handler_libc.odin
@@ -26,6 +26,8 @@ import "core:os"
 
 @(private="file", thread_local)
 local_test_index: libc.sig_atomic_t
+@(private="file", thread_local)
+local_test_index_set: bool
 
 // Windows does not appear to have a SIGTRAP, so this is defined here, instead
 // of in the libc package, just so there's no confusion about it being
@@ -45,6 +47,13 @@ stop_runner_callback :: proc "c" (sig: libc.int) {
 
 @(private="file")
 stop_test_callback :: proc "c" (sig: libc.int) {
+	if !local_test_index_set {
+		// We're a thread created by a test thread.
+		//
+		// There's nothing we can do to inform the test runner about who
+		// signalled, so hopefully the test will handle their own sub-threads.
+		return
+	}
 	if local_test_index == -1 {
 		// We're the test runner, and we ourselves have caught a signal from
 		// which there is no recovery.
@@ -114,6 +123,7 @@ This is a dire bug and should be reported to the Odin developers.
 
 _setup_signal_handler :: proc() {
 	local_test_index = -1
+	local_test_index_set = true
 
 	// Catch user interrupt / CTRL-C.
 	libc.signal(libc.SIGINT, stop_runner_callback)
@@ -135,6 +145,7 @@ _setup_signal_handler :: proc() {
 
 _setup_task_signal_handler :: proc(test_index: int) {
 	local_test_index = cast(libc.sig_atomic_t)test_index
+	local_test_index_set = true
 }
 
 _should_stop_runner :: proc() -> bool {

From b2c2235e587bb8902dbf35ef84373bb5f616a814 Mon Sep 17 00:00:00 2001
From: Feoramund <161657516+Feoramund@users.noreply.github.com>
Date: Mon, 9 Sep 2024 19:11:44 -0400
Subject: [PATCH 15/23] Fix `recursive_benaphore_try_lock`

Previously, if the owner called this, it would fail.
---
 core/sync/extended.odin | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/core/sync/extended.odin b/core/sync/extended.odin
index 83cc648b477..d5521935afe 100644
--- a/core/sync/extended.odin
+++ b/core/sync/extended.odin
@@ -474,10 +474,10 @@ recursive_benaphore_try_lock :: proc "contextless" (b: ^Recursive_Benaphore) ->
 	tid := current_thread_id()
 	if b.owner == tid {
 		atomic_add_explicit(&b.counter, 1, .Acquire)
-	}
-
-	if v, _ := atomic_compare_exchange_strong_explicit(&b.counter, 0, 1, .Acquire, .Acquire); v != 0 {
-		return false
+	} else {
+		if v, _ := atomic_compare_exchange_strong_explicit(&b.counter, 0, 1, .Acquire, .Acquire); v != 0 {
+			return false
+		}
 	}
 	// inside the lock
 	b.owner = tid

From fec1ccd7a3f0946a84b9914875845cc9f902ee74 Mon Sep 17 00:00:00 2001
From: Feoramund <161657516+Feoramund@users.noreply.github.com>
Date: Mon, 9 Sep 2024 19:15:06 -0400
Subject: [PATCH 16/23] Fix data races in `sync.Recursive_Benaphore`

---
 core/sync/extended.odin | 31 ++++++++++++++++---------------
 1 file changed, 16 insertions(+), 15 deletions(-)

diff --git a/core/sync/extended.odin b/core/sync/extended.odin
index d5521935afe..924ba92a691 100644
--- a/core/sync/extended.odin
+++ b/core/sync/extended.odin
@@ -449,13 +449,15 @@ recursive benaphore, until the lock is released.
 */
 recursive_benaphore_lock :: proc "contextless" (b: ^Recursive_Benaphore) {
 	tid := current_thread_id()
-	if atomic_add_explicit(&b.counter, 1, .Acquire) > 1 {
-		if tid != b.owner {
-			sema_wait(&b.sema)
+	check_owner: if tid != atomic_load_explicit(&b.owner, .Acquire) {
+		atomic_add_explicit(&b.counter, 1, .Relaxed)
+		if _, ok := atomic_compare_exchange_strong_explicit(&b.owner, 0, tid, .Release, .Relaxed); ok {
+			break check_owner
 		}
+		sema_wait(&b.sema)
+		atomic_store_explicit(&b.owner, tid, .Release)
 	}
 	// inside the lock
-	b.owner = tid
 	b.recursion += 1
 }
 
@@ -472,15 +474,14 @@ benaphore, until the lock is released.
 */
 recursive_benaphore_try_lock :: proc "contextless" (b: ^Recursive_Benaphore) -> bool {
 	tid := current_thread_id()
-	if b.owner == tid {
-		atomic_add_explicit(&b.counter, 1, .Acquire)
-	} else {
-		if v, _ := atomic_compare_exchange_strong_explicit(&b.counter, 0, 1, .Acquire, .Acquire); v != 0 {
-			return false
+	check_owner: if tid != atomic_load_explicit(&b.owner, .Acquire) {
+		if _, ok := atomic_compare_exchange_strong_explicit(&b.owner, 0, tid, .Release, .Relaxed); ok {
+			atomic_add_explicit(&b.counter, 1, .Relaxed)
+			break check_owner
 		}
+		return false
 	}
 	// inside the lock
-	b.owner = tid
 	b.recursion += 1
 	return true
 }
@@ -494,14 +495,14 @@ for other threads for entering.
 */
 recursive_benaphore_unlock :: proc "contextless" (b: ^Recursive_Benaphore) {
 	tid := current_thread_id()
-	assert_contextless(tid == b.owner, "tid != b.owner")
+	assert_contextless(tid == atomic_load_explicit(&b.owner, .Relaxed), "tid != b.owner")
 	b.recursion -= 1
 	recursion := b.recursion
+
 	if recursion == 0 {
-		b.owner = 0
-	}
-	if atomic_sub_explicit(&b.counter, 1, .Release) > 0 {
-		if recursion == 0 {
+		if atomic_sub_explicit(&b.counter, 1, .Relaxed) == 1 {
+			atomic_store_explicit(&b.owner, 0, .Release)
+		} else {
 			sema_post(&b.sema)
 		}
 	}

From a1435a6a904b2eb2b154a463c5d60f6c1f55abbc Mon Sep 17 00:00:00 2001
From: Feoramund <161657516+Feoramund@users.noreply.github.com>
Date: Tue, 10 Sep 2024 14:51:00 -0400
Subject: [PATCH 17/23] Fix deadlock in `Auto_Reset_Event`

---
 core/sync/extended.odin | 9 ++++-----
 1 file changed, 4 insertions(+), 5 deletions(-)

diff --git a/core/sync/extended.odin b/core/sync/extended.odin
index 924ba92a691..0971516a3d2 100644
--- a/core/sync/extended.odin
+++ b/core/sync/extended.odin
@@ -228,15 +228,14 @@ thread.
 */
 auto_reset_event_signal :: proc "contextless" (e: ^Auto_Reset_Event) {
 	old_status := atomic_load_explicit(&e.status, .Relaxed)
+	new_status := old_status + 1 if old_status < 1 else 1
 	for {
-		new_status := old_status + 1 if old_status < 1 else 1
 		if _, ok := atomic_compare_exchange_weak_explicit(&e.status, old_status, new_status, .Release, .Relaxed); ok {
 			break
 		}
-
-		if old_status < 0 {
-			sema_post(&e.sema)
-		}
+	}
+	if old_status < 0 {
+		sema_post(&e.sema)
 	}
 }
 

From b1db33b519bf52e2d0e6e42ca9daccc7470d6b8b Mon Sep 17 00:00:00 2001
From: Feoramund <161657516+Feoramund@users.noreply.github.com>
Date: Tue, 10 Sep 2024 19:04:44 -0400
Subject: [PATCH 18/23] Add `cpu_relax` to `sync.auto_reset_event_signal`

---
 core/sync/extended.odin | 1 +
 1 file changed, 1 insertion(+)

diff --git a/core/sync/extended.odin b/core/sync/extended.odin
index 0971516a3d2..0b1f79df212 100644
--- a/core/sync/extended.odin
+++ b/core/sync/extended.odin
@@ -233,6 +233,7 @@ auto_reset_event_signal :: proc "contextless" (e: ^Auto_Reset_Event) {
 		if _, ok := atomic_compare_exchange_weak_explicit(&e.status, old_status, new_status, .Release, .Relaxed); ok {
 			break
 		}
+		cpu_relax()
 	}
 	if old_status < 0 {
 		sema_post(&e.sema)

From 2938655a3d8801d1327b0076812edcf357d760df Mon Sep 17 00:00:00 2001
From: Feoramund <161657516+Feoramund@users.noreply.github.com>
Date: Wed, 11 Sep 2024 07:07:09 -0400
Subject: [PATCH 19/23] Fix CPU count detection in FreeBSD & NetBSD

---
 core/os/os_freebsd.odin |  2 +-
 core/os/os_netbsd.odin  |  2 +-
 src/gb/gb.h             | 12 ++++++++++--
 3 files changed, 12 insertions(+), 4 deletions(-)

diff --git a/core/os/os_freebsd.odin b/core/os/os_freebsd.odin
index c05a06129c8..241f42c0bb0 100644
--- a/core/os/os_freebsd.odin
+++ b/core/os/os_freebsd.odin
@@ -920,7 +920,7 @@ get_page_size :: proc() -> int {
 _processor_core_count :: proc() -> int {
 	count : int = 0
 	count_size := size_of(count)
-	if _sysctlbyname("hw.logicalcpu", &count, &count_size, nil, 0) == 0 {
+	if _sysctlbyname("hw.ncpu", &count, &count_size, nil, 0) == 0 {
 		if count > 0 {
 			return count
 		}
diff --git a/core/os/os_netbsd.odin b/core/os/os_netbsd.odin
index a56c0b78467..ba9b40fc365 100644
--- a/core/os/os_netbsd.odin
+++ b/core/os/os_netbsd.odin
@@ -978,7 +978,7 @@ get_page_size :: proc() -> int {
 _processor_core_count :: proc() -> int {
 	count : int = 0
 	count_size := size_of(count)
-	if _sysctlbyname("hw.logicalcpu", &count, &count_size, nil, 0) == 0 {
+	if _sysctlbyname("hw.ncpu", &count, &count_size, nil, 0) == 0 {
 		if count > 0 {
 			return count
 		}
diff --git a/src/gb/gb.h b/src/gb/gb.h
index 0e65696e260..1fef4b4f511 100644
--- a/src/gb/gb.h
+++ b/src/gb/gb.h
@@ -3195,11 +3195,11 @@ void gb_affinity_init(gbAffinity *a) {
 	a->core_count       = 1;
 	a->threads_per_core = 1;
 
-	if (sysctlbyname("hw.logicalcpu", &count, &count_size, NULL, 0) == 0) {
+	if (sysctlbyname("kern.smp.cpus", &count, &count_size, NULL, 0) == 0) {
 		if (count > 0) {
 			a->thread_count = count;
 			// Get # of physical cores
-			if (sysctlbyname("hw.physicalcpu", &count, &count_size, NULL, 0) == 0) {
+			if (sysctlbyname("kern.smp.cores", &count, &count_size, NULL, 0) == 0) {
 				if (count > 0) {
 					a->core_count = count;
 					a->threads_per_core = a->thread_count / count;
@@ -3210,6 +3210,14 @@ void gb_affinity_init(gbAffinity *a) {
 				}
 			}
 		}
+	} else if (sysctlbyname("hw.ncpu", &count, &count_size, NULL, 0) == 0) {
+		// SMP disabled or unavailable.
+		if (count > 0) {
+			a->is_accurate      = true;
+			a->thread_count     = count;
+			a->core_count       = count;
+			a->threads_per_core = 1;
+		}
 	}
 
 }

From 16cd16b91e9cabfba2d12e271712ca55cb16fa7d Mon Sep 17 00:00:00 2001
From: Feoramund <161657516+Feoramund@users.noreply.github.com>
Date: Sun, 8 Sep 2024 18:23:28 -0400
Subject: [PATCH 20/23] Fix comments

---
 core/sync/extended.odin   | 18 +++++++++---------
 core/sync/primitives.odin |  2 +-
 2 files changed, 10 insertions(+), 10 deletions(-)

diff --git a/core/sync/extended.odin b/core/sync/extended.odin
index 0b1f79df212..30b1b277079 100644
--- a/core/sync/extended.odin
+++ b/core/sync/extended.odin
@@ -8,7 +8,7 @@ _ :: vg
 Wait group.
 
 Wait group is a synchronization primitive used by the waiting thread to wait,
-until a all working threads finish work.
+until all working threads finish work.
 
 The waiting thread first sets the number of working threads it will expect to
 wait for using `wait_group_add` call, and start waiting using `wait_group_wait`
@@ -35,7 +35,7 @@ Wait_Group :: struct #no_copy {
 /*
 Increment an internal counter of a wait group.
 
-This procedure atomicaly increments a number to the specified wait group's
+This procedure atomically increments a number to the specified wait group's
 internal counter by a specified amount. This operation can be done on any
 thread.
 */
@@ -121,7 +121,7 @@ When `barrier_wait` procedure is called by any thread, that thread will block
 the execution, until all threads associated with the barrier reach the same
 point of execution and also call `barrier_wait`.
 
-when barrier is initialized, a `thread_count` parameter is passed, signifying
+When a barrier is initialized, a `thread_count` parameter is passed, signifying
 the amount of participant threads of the barrier. The barrier also keeps track
 of an internal atomic counter. When a thread calls `barrier_wait`, the internal
 counter is incremented. When the internal counter reaches `thread_count`, it is
@@ -208,7 +208,7 @@ Represents a thread synchronization primitive that, when signalled, releases one
 single waiting thread and then resets automatically to a state where it can be
 signalled again.
 
-When a thread calls `auto_reset_event_wait`, it's execution will be blocked,
+When a thread calls `auto_reset_event_wait`, its execution will be blocked,
 until the event is signalled by another thread. The call to
 `auto_reset_event_signal` wakes up exactly one thread waiting for the event.
 */
@@ -331,8 +331,8 @@ Benaphore.
 
 A benaphore is a combination of an atomic variable and a semaphore that can
 improve locking efficiency in a no-contention system. Acquiring a benaphore
-lock doesn't call into an internal semaphore, if no other thread in a middle of
-a critical section.
+lock doesn't call into an internal semaphore, if no other thread is in the
+middle of a critical section.
 
 Once a lock on a benaphore is acquired by a thread, no other thread is allowed
 into any critical sections, associted with the same benaphore, until the lock
@@ -381,7 +381,7 @@ Release a lock on a benaphore.
 
 This procedure releases a lock on the specified benaphore. If any of the threads
 are waiting on the lock, exactly one thread is allowed into a critical section
-associated with the same banaphore.
+associated with the same benaphore.
 */
 benaphore_unlock :: proc "contextless" (b: ^Benaphore) {
 	if atomic_sub_explicit(&b.counter, 1, .Release) > 1 {
@@ -418,8 +418,8 @@ benaphore_guard :: proc "contextless" (m: ^Benaphore) -> bool {
 /*
 Recursive benaphore.
 
-Recurisve benaphore is just like a plain benaphore, except it allows reentrancy
-into the critical section.
+A recursive benaphore is just like a plain benaphore, except it allows
+reentrancy into the critical section.
 
 When a lock is acquired on a benaphore, all other threads attempting to
 acquire a lock on the same benaphore will be blocked from any critical sections,
diff --git a/core/sync/primitives.odin b/core/sync/primitives.odin
index 8187c904bbd..f091de0452d 100644
--- a/core/sync/primitives.odin
+++ b/core/sync/primitives.odin
@@ -389,7 +389,7 @@ recursive_mutex_guard :: proc "contextless" (m: ^Recursive_Mutex) -> bool {
 A condition variable.
 
 `Cond` implements a condition variable, a rendezvous point for threads waiting
-for signalling the occurence of an event. Condition variables are used on
+for signalling the occurence of an event. Condition variables are used in
 conjuction with mutexes to provide a shared access to one or more shared
 variable.
 

From 7f7cfebc91cc319918bc0042789b0c7a931e56e2 Mon Sep 17 00:00:00 2001
From: Feoramund <161657516+Feoramund@users.noreply.github.com>
Date: Mon, 9 Sep 2024 16:17:08 -0400
Subject: [PATCH 21/23] Add tests for `core:sync` and `core:sync/chan`

---
 tests/core/normal.odin                        |   2 +
 tests/core/sync/chan/test_core_sync_chan.odin | 274 +++++++
 tests/core/sync/test_core_sync.odin           | 718 ++++++++++++++++++
 3 files changed, 994 insertions(+)
 create mode 100644 tests/core/sync/chan/test_core_sync_chan.odin
 create mode 100644 tests/core/sync/test_core_sync.odin

diff --git a/tests/core/normal.odin b/tests/core/normal.odin
index 6174f2d5cd0..0670842ac8b 100644
--- a/tests/core/normal.odin
+++ b/tests/core/normal.odin
@@ -39,6 +39,8 @@ download_assets :: proc() {
 @(require) import "slice"
 @(require) import "strconv"
 @(require) import "strings"
+@(require) import "sync"
+@(require) import "sync/chan"
 @(require) import "sys/posix"
 @(require) import "sys/windows"
 @(require) import "text/i18n"
diff --git a/tests/core/sync/chan/test_core_sync_chan.odin b/tests/core/sync/chan/test_core_sync_chan.odin
new file mode 100644
index 00000000000..9b8d9b3549d
--- /dev/null
+++ b/tests/core/sync/chan/test_core_sync_chan.odin
@@ -0,0 +1,274 @@
+package test_core_sync_chan
+
+import "base:runtime"
+import "base:intrinsics"
+import "core:log"
+import "core:math/rand"
+import "core:sync/chan"
+import "core:testing"
+import "core:thread"
+import "core:time"
+
+
+Message_Type :: enum i32 {
+	Result,
+	Add,
+	Multiply,
+	Subtract,
+	Divide,
+	End,
+}
+
+Message :: struct {
+	type: Message_Type,
+	i: i64,
+}
+
+Comm :: struct {
+	host: chan.Chan(Message),
+	client: chan.Chan(Message),
+	manual_buffering: bool,
+}
+
+BUFFER_SIZE :: 8
+MAX_RAND    :: 32
+FAIL_TIME   :: 1 * time.Second
+SLEEP_TIME  :: 1 * time.Millisecond
+
+comm_client :: proc(th: ^thread.Thread) {
+	data := cast(^Comm)th.data
+	manual_buffering := data.manual_buffering
+
+	n: i64
+
+	for manual_buffering && !chan.can_recv(data.host) {
+		thread.yield()
+	}
+
+	recv_loop: for msg in chan.recv(data.host) {
+		#partial switch msg.type {
+		case .Add:      n += msg.i
+		case .Multiply: n *= msg.i
+		case .Subtract: n -= msg.i
+		case .Divide:   n /= msg.i
+		case .End:
+			break recv_loop
+		case:
+			panic("Unknown message type for client.")
+		}
+
+		for manual_buffering && !chan.can_recv(data.host) {
+			thread.yield()
+		}
+	}
+
+	for manual_buffering && !chan.can_send(data.host) {
+		thread.yield()
+	}
+
+	chan.send(data.client, Message{.Result, n})
+	chan.close(data.client)
+}
+
+send_messages :: proc(t: ^testing.T, host: chan.Chan(Message), manual_buffering: bool = false) -> (expected: i64) {
+	expected = 1
+	for manual_buffering && !chan.can_send(host) {
+		thread.yield()
+	}
+	chan.send(host, Message{.Add, 1})
+	log.debug(Message{.Add, 1})
+
+	for _ in 0..<1+2*BUFFER_SIZE {
+		msg: Message
+		msg.i = 1 + rand.int63_max(MAX_RAND)
+		switch rand.int_max(4) {
+		case 0:
+			msg.type = .Add
+			expected += msg.i
+		case 1:
+			msg.type = .Multiply
+			expected *= msg.i
+		case 2:
+			msg.type = .Subtract
+			expected -= msg.i
+		case 3:
+			msg.type = .Divide
+			expected /= msg.i
+		}
+
+		for manual_buffering && !chan.can_send(host) {
+			thread.yield()
+		}
+		if manual_buffering {
+			testing.expect(t, chan.len(host) == 0)
+		}
+
+		chan.send(host, msg)
+		log.debug(msg)
+	}
+
+	for manual_buffering && !chan.can_send(host) {
+		thread.yield()
+	}
+	chan.send(host, Message{.End, 0})
+	log.debug(Message{.End, 0})
+	chan.close(host)
+
+	return
+}
+
+@test
+test_chan_buffered :: proc(t: ^testing.T) {
+	testing.set_fail_timeout(t, FAIL_TIME)
+
+	comm: Comm
+	alloc_err: runtime.Allocator_Error
+	comm.host,   alloc_err = chan.create_buffered(chan.Chan(Message), BUFFER_SIZE, context.allocator)
+	assert(alloc_err == nil, "allocation failed")
+	comm.client, alloc_err = chan.create_buffered(chan.Chan(Message), BUFFER_SIZE, context.allocator)
+	assert(alloc_err == nil, "allocation failed")
+	defer {
+		chan.destroy(comm.host)
+		chan.destroy(comm.client)
+	}
+
+	testing.expect(t, chan.is_buffered(comm.host))
+	testing.expect(t, chan.is_buffered(comm.client))
+	testing.expect(t, !chan.is_unbuffered(comm.host))
+	testing.expect(t, !chan.is_unbuffered(comm.client))
+	testing.expect_value(t, chan.len(comm.host), 0)
+	testing.expect_value(t, chan.len(comm.client), 0)
+	testing.expect_value(t, chan.cap(comm.host), BUFFER_SIZE)
+	testing.expect_value(t, chan.cap(comm.client), BUFFER_SIZE)
+
+	reckoner := thread.create(comm_client)
+	defer thread.destroy(reckoner)
+	reckoner.data = &comm
+	thread.start(reckoner)
+
+	expected := send_messages(t, comm.host, manual_buffering = false)
+
+	// Sleep so we can give the other thread enough time to buffer its message.
+	time.sleep(SLEEP_TIME)
+
+	testing.expect_value(t, chan.len(comm.client), 1)
+	result, ok := chan.try_recv(comm.client)
+
+	// One more sleep to ensure it has enough time to close.
+	time.sleep(SLEEP_TIME)
+
+	testing.expect_value(t, chan.is_closed(comm.client), true)
+	testing.expect_value(t, ok, true)
+	testing.expect_value(t, result.i, expected)
+	log.debug(result, expected)
+
+	// Make sure sending to closed channels fails.
+	testing.expect_value(t, chan.send(comm.host, Message{.End, 0}), false)
+	testing.expect_value(t, chan.send(comm.client, Message{.End, 0}), false)
+	testing.expect_value(t, chan.try_send(comm.host, Message{.End, 0}), false)
+	testing.expect_value(t, chan.try_send(comm.client, Message{.End, 0}), false)
+	_, ok = chan.recv(comm.host);       testing.expect_value(t, ok, false)
+	_, ok = chan.recv(comm.client);     testing.expect_value(t, ok, false)
+	_, ok = chan.try_recv(comm.host);   testing.expect_value(t, ok, false)
+	_, ok = chan.try_recv(comm.client); testing.expect_value(t, ok, false)
+}
+
+@test
+test_chan_unbuffered :: proc(t: ^testing.T) {
+	testing.set_fail_timeout(t, FAIL_TIME)
+
+	comm: Comm
+	comm.manual_buffering = true
+	alloc_err: runtime.Allocator_Error
+	comm.host,   alloc_err = chan.create_unbuffered(chan.Chan(Message), context.allocator)
+	assert(alloc_err == nil, "allocation failed")
+	comm.client, alloc_err = chan.create_unbuffered(chan.Chan(Message), context.allocator)
+	assert(alloc_err == nil, "allocation failed")
+	defer {
+		chan.destroy(comm.host)
+		chan.destroy(comm.client)
+	}
+
+	testing.expect(t, !chan.is_buffered(comm.host))
+	testing.expect(t, !chan.is_buffered(comm.client))
+	testing.expect(t, chan.is_unbuffered(comm.host))
+	testing.expect(t, chan.is_unbuffered(comm.client))
+	testing.expect_value(t, chan.len(comm.host), 0)
+	testing.expect_value(t, chan.len(comm.client), 0)
+	testing.expect_value(t, chan.cap(comm.host), 0)
+	testing.expect_value(t, chan.cap(comm.client), 0)
+
+	reckoner := thread.create(comm_client)
+	defer thread.destroy(reckoner)
+	reckoner.data = &comm
+	thread.start(reckoner)
+
+	for !chan.can_send(comm.client) {
+		thread.yield()
+	}
+
+	expected := send_messages(t, comm.host)
+	testing.expect_value(t, chan.is_closed(comm.host), true)
+
+	for !chan.can_recv(comm.client) {
+		thread.yield()
+	}
+
+	result, ok := chan.try_recv(comm.client)
+	testing.expect_value(t, ok, true)
+	testing.expect_value(t, result.i, expected)
+	log.debug(result, expected)
+
+	// Sleep so we can give the other thread enough time to close its side
+	// after we've received its message.
+	time.sleep(SLEEP_TIME)
+
+	testing.expect_value(t, chan.is_closed(comm.client), true)
+
+	// Make sure sending and receiving on closed channels fails.
+	testing.expect_value(t, chan.send(comm.host, Message{.End, 0}), false)
+	testing.expect_value(t, chan.send(comm.client, Message{.End, 0}), false)
+	testing.expect_value(t, chan.try_send(comm.host, Message{.End, 0}), false)
+	testing.expect_value(t, chan.try_send(comm.client, Message{.End, 0}), false)
+	_, ok = chan.recv(comm.host);       testing.expect_value(t, ok, false)
+	_, ok = chan.recv(comm.client);     testing.expect_value(t, ok, false)
+	_, ok = chan.try_recv(comm.host);   testing.expect_value(t, ok, false)
+	_, ok = chan.try_recv(comm.client); testing.expect_value(t, ok, false)
+}
+
+@test
+test_full_buffered_closed_chan_deadlock :: proc(t: ^testing.T) {
+	testing.set_fail_timeout(t, FAIL_TIME)
+
+	ch, alloc_err := chan.create_buffered(chan.Chan(int), 1, context.allocator)
+	assert(alloc_err == nil, "allocation failed")
+	defer chan.destroy(ch)
+
+	testing.expect(t, chan.can_send(ch))
+	testing.expect(t, chan.send(ch, 32))
+	testing.expect(t, chan.close(ch))
+	testing.expect(t, !chan.send(ch, 32))
+}
+
+// This test guarantees a buffered channel's messages can still be received
+// even after closing. This is currently how the API works. If that changes,
+// this test will need to change.
+@test
+test_accept_message_from_closed_buffered_chan :: proc(t: ^testing.T) {
+	testing.set_fail_timeout(t, FAIL_TIME)
+
+	ch, alloc_err := chan.create_buffered(chan.Chan(int), 2, context.allocator)
+	assert(alloc_err == nil, "allocation failed")
+	defer chan.destroy(ch)
+
+	testing.expect(t, chan.can_send(ch))
+	testing.expect(t, chan.send(ch, 32))
+	testing.expect(t, chan.send(ch, 64))
+	testing.expect(t, chan.close(ch))
+	result, ok := chan.recv(ch)
+	testing.expect_value(t, result, 32)
+	testing.expect(t, ok)
+	result, ok = chan.try_recv(ch)
+	testing.expect_value(t, result, 64)
+	testing.expect(t, ok)
+}
diff --git a/tests/core/sync/test_core_sync.odin b/tests/core/sync/test_core_sync.odin
new file mode 100644
index 00000000000..32c08f9351d
--- /dev/null
+++ b/tests/core/sync/test_core_sync.odin
@@ -0,0 +1,718 @@
+// NOTE(Feoramund): These tests should be run a few hundred times, with and
+// without `-sanitize:thread` enabled, to ensure maximum safety.
+//
+// Keep in mind that running with the debug logs uncommented can result in
+// failures disappearing due to the delay of sending the log message causing
+// different synchronization patterns.
+//
+// These tests are temporarily disabled on Darwin, as there is currently a
+// stall occurring which I cannot debug.
+
+//+build !darwin
+package test_core_sync
+
+import "base:intrinsics"
+// import "core:log"
+import "core:sync"
+import "core:testing"
+import "core:thread"
+import "core:time"
+
+FAIL_TIME        :: 1 * time.Second
+SLEEP_TIME       :: 1 * time.Millisecond
+SMALL_SLEEP_TIME :: 10 * time.Microsecond
+
+// This needs to be high enough to cause a data race if any of the
+// synchronization primitives fail.
+THREADS :: 8
+
+// Manually wait on all threads to finish.
+//
+// This reduces a dependency on a `Wait_Group` or similar primitives.
+//
+// It's also important that we wait for every thread to finish, as it's
+// possible for a thread to finish after the test if we don't check, despite
+// joining it to the test thread.
+wait_for :: proc(threads: []^thread.Thread) {
+	wait_loop: for {
+		count := len(threads)
+		for v in threads {
+			if thread.is_done(v) {
+				count -= 1
+			}
+		}
+		if count == 0 {
+			break wait_loop
+		}
+		thread.yield()
+	}
+	for t in threads {
+		thread.join(t)
+		thread.destroy(t)
+	}
+}
+
+//
+// core:sync/primitives.odin
+//
+
+@test
+test_mutex :: proc(t: ^testing.T) {
+	testing.set_fail_timeout(t, FAIL_TIME)
+
+	Data :: struct {
+		m: sync.Mutex,
+		number: int,
+	}
+
+	p :: proc(th: ^thread.Thread) {
+		data := cast(^Data)th.data
+
+		// log.debugf("MUTEX-%v> locking", th.id)
+		sync.mutex_lock(&data.m)
+		data.number += 1
+		// log.debugf("MUTEX-%v> unlocking", th.id)
+		sync.mutex_unlock(&data.m)
+		// log.debugf("MUTEX-%v> leaving", th.id)
+	}
+
+	data: Data
+	threads: [THREADS]^thread.Thread
+
+	for &v in threads {
+		v = thread.create(p)
+		v.data = &data
+		v.init_context = context
+		thread.start(v)
+	}
+
+	wait_for(threads[:])
+
+	testing.expect_value(t, data.number, THREADS)
+}
+
+@test
+test_rw_mutex :: proc(t: ^testing.T) {
+	testing.set_fail_timeout(t, FAIL_TIME)
+
+	Data :: struct {
+		m1: sync.RW_Mutex,
+		m2: sync.RW_Mutex,
+		number1: int,
+		number2: int,
+	}
+
+	p :: proc(th: ^thread.Thread) {
+		data := cast(^Data)th.data
+
+		sync.rw_mutex_shared_lock(&data.m1)
+		n := data.number1
+		sync.rw_mutex_shared_unlock(&data.m1)
+
+		sync.rw_mutex_lock(&data.m2)
+		data.number2 += n
+		sync.rw_mutex_unlock(&data.m2)
+	}
+
+	data: Data
+	threads: [THREADS]^thread.Thread
+
+	sync.rw_mutex_lock(&data.m1)
+
+	for &v in threads {
+		v = thread.create(p)
+		v.data = &data
+		v.init_context = context
+		thread.start(v)
+	}
+
+	data.number1 = 1
+	sync.rw_mutex_unlock(&data.m1)
+
+	wait_for(threads[:])
+
+	testing.expect_value(t, data.number2, THREADS)
+}
+
+@test
+test_recursive_mutex :: proc(t: ^testing.T) {
+	testing.set_fail_timeout(t, FAIL_TIME)
+
+	Data :: struct {
+		m: sync.Recursive_Mutex,
+		number: int,
+	}
+
+	p :: proc(th: ^thread.Thread) {
+		data := cast(^Data)th.data
+
+		// log.debugf("REC_MUTEX-%v> locking", th.id)
+		tried1 := sync.recursive_mutex_try_lock(&data.m)
+		for _ in 0..<3 {
+			sync.recursive_mutex_lock(&data.m)
+		}
+		tried2 := sync.recursive_mutex_try_lock(&data.m)
+		// log.debugf("REC_MUTEX-%v> locked", th.id)
+		data.number += 1
+		// log.debugf("REC_MUTEX-%v> unlocking", th.id)
+		for _ in 0..<3 {
+			sync.recursive_mutex_unlock(&data.m)
+		}
+		if tried1 { sync.recursive_mutex_unlock(&data.m) }
+		if tried2 { sync.recursive_mutex_unlock(&data.m) }
+		// log.debugf("REC_MUTEX-%v> leaving", th.id)
+	}
+
+	data: Data
+	threads: [THREADS]^thread.Thread
+
+	for &v in threads {
+		v = thread.create(p)
+		v.data = &data
+		v.init_context = context
+		thread.start(v)
+	}
+
+	wait_for(threads[:])
+
+	testing.expect_value(t, data.number, THREADS)
+}
+
+@test
+test_cond :: proc(t: ^testing.T) {
+	testing.set_fail_timeout(t, FAIL_TIME)
+
+	Data :: struct {
+		c: sync.Cond,
+		m: sync.Mutex,
+		i: int,
+		number: int,
+	}
+
+	p :: proc(th: ^thread.Thread) {
+		data := cast(^Data)th.data
+
+		sync.mutex_lock(&data.m)
+
+		for intrinsics.atomic_load(&data.i) != 1 {
+			sync.cond_wait(&data.c, &data.m)
+		}
+
+		data.number += intrinsics.atomic_load(&data.i)
+
+		sync.mutex_unlock(&data.m)
+	}
+
+	data: Data
+	threads: [THREADS]^thread.Thread
+	data.i = -1
+
+	sync.mutex_lock(&data.m)
+
+	for &v in threads {
+		v = thread.create(p)
+		v.data = &data
+		v.init_context = context
+		thread.start(v)
+	}
+
+	time.sleep(SLEEP_TIME)
+	data.i = 1
+	sync.mutex_unlock(&data.m)
+	sync.cond_broadcast(&data.c)
+
+	wait_for(threads[:])
+
+	testing.expect_value(t, data.number, THREADS)
+}
+
+@test
+test_cond_with_timeout :: proc(t: ^testing.T) {
+	testing.set_fail_timeout(t, FAIL_TIME)
+
+	c: sync.Cond
+	m: sync.Mutex
+	sync.mutex_lock(&m)
+	sync.cond_wait_with_timeout(&c, &m, SLEEP_TIME)
+}
+
+@test
+test_semaphore :: proc(t: ^testing.T) {
+	testing.set_fail_timeout(t, FAIL_TIME)
+
+	Data :: struct {
+		s: sync.Sema,
+		number: int,
+	}
+
+	p :: proc(th: ^thread.Thread) {
+		data := cast(^Data)th.data
+
+		// log.debugf("SEM-%v> waiting", th.id)
+		sync.sema_wait(&data.s)
+		data.number += 1
+		// log.debugf("SEM-%v> posting", th.id)
+		sync.sema_post(&data.s)
+		// log.debugf("SEM-%v> leaving", th.id)
+	}
+
+	data: Data
+	threads: [THREADS]^thread.Thread
+
+	for &v in threads {
+		v = thread.create(p)
+		v.data = &data
+		v.init_context = context
+		thread.start(v)
+	}
+	sync.sema_post(&data.s)
+
+	wait_for(threads[:])
+
+	testing.expect_value(t, data.number, THREADS)
+}
+
+@test
+test_semaphore_with_timeout :: proc(t: ^testing.T) {
+	testing.set_fail_timeout(t, FAIL_TIME)
+
+	s: sync.Sema
+	sync.sema_wait_with_timeout(&s, SLEEP_TIME)
+}
+
+@test
+test_futex :: proc(t: ^testing.T) {
+	testing.set_fail_timeout(t, FAIL_TIME)
+
+	Data :: struct {
+		f: sync.Futex,
+		i: int,
+		number: int,
+	}
+
+	p :: proc(th: ^thread.Thread) {
+		data := cast(^Data)th.data
+
+		// log.debugf("FUTEX-%v> waiting", th.id)
+		sync.futex_wait(&data.f, 3)
+		// log.debugf("FUTEX-%v> done", th.id)
+
+		n := data.i
+		intrinsics.atomic_add(&data.number, n)
+	}
+
+	data: Data
+	data.i = -1
+	data.f = 3
+	threads: [THREADS]^thread.Thread
+
+	for &v in threads {
+		v = thread.create(p)
+		v.data = &data
+		v.init_context = context
+		thread.start(v)
+	}
+
+	data.i = 1
+	// Change the futex variable to keep late-starters from stalling.
+	data.f = 0
+	sync.futex_broadcast(&data.f)
+
+	wait_for(threads[:])
+
+	testing.expect_value(t, data.number, THREADS)
+}
+
+@test
+test_futex_with_timeout :: proc(t: ^testing.T) {
+	testing.set_fail_timeout(t, FAIL_TIME)
+
+	f: sync.Futex = 1
+	sync.futex_wait_with_timeout(&f, 1, SLEEP_TIME)
+}
+
+//
+// core:sync/extended.odin
+//
+
+@test
+test_wait_group :: proc(t: ^testing.T) {
+	testing.set_fail_timeout(t, FAIL_TIME)
+
+	Data :: struct {
+		step1: sync.Wait_Group,
+		step2: sync.Wait_Group,
+		i: int,
+		number: int,
+	}
+
+	p :: proc(th: ^thread.Thread) {
+		data := cast(^Data)th.data
+
+		sync.wait_group_wait(&data.step1)
+
+		n := data.i
+		intrinsics.atomic_add(&data.number, n)
+
+		sync.wait_group_done(&data.step2)
+	}
+
+	data: Data
+	data.i = -1
+	threads: [THREADS]^thread.Thread
+
+	sync.wait_group_add(&data.step1, 1)
+	sync.wait_group_add(&data.step2, THREADS)
+
+	for &v in threads {
+		v = thread.create(p)
+		v.data = &data
+		v.init_context = context
+		thread.start(v)
+	}
+
+	time.sleep(SMALL_SLEEP_TIME)
+	data.i = 1
+	sync.wait_group_done(&data.step1)
+
+	sync.wait_group_wait(&data.step2)
+
+	wait_for(threads[:])
+
+	testing.expect_value(t, data.step1.counter, 0)
+	testing.expect_value(t, data.step2.counter, 0)
+	testing.expect_value(t, data.number, THREADS)
+}
+
+@test
+test_wait_group_with_timeout :: proc(t: ^testing.T) {
+	testing.set_fail_timeout(t, FAIL_TIME)
+
+	wg: sync.Wait_Group
+	sync.wait_group_wait_with_timeout(&wg, SLEEP_TIME)
+}
+
+@test
+test_barrier :: proc(t: ^testing.T) {
+	testing.set_fail_timeout(t, FAIL_TIME)
+
+	Data :: struct {
+		b: sync.Barrier,
+		i: int,
+		number: int,
+
+	}
+
+	p :: proc(th: ^thread.Thread) {
+		data := cast(^Data)th.data
+
+		sync.barrier_wait(&data.b)
+
+		intrinsics.atomic_add(&data.number, data.i)
+	}
+
+	data: Data
+	data.i = -1
+	threads: [THREADS]^thread.Thread
+
+	sync.barrier_init(&data.b, THREADS + 1) // +1 for this thread, of course.
+
+	for &v in threads {
+		v = thread.create(p)
+		v.data = &data
+		v.init_context = context
+		thread.start(v)
+	}
+	time.sleep(SMALL_SLEEP_TIME)
+	data.i = 1
+	sync.barrier_wait(&data.b)
+
+	wait_for(threads[:])
+
+	testing.expect_value(t, data.b.index, 0)
+	testing.expect_value(t, data.b.generation_id, 1)
+	testing.expect_value(t, data.b.thread_count, THREADS + 1)
+	testing.expect_value(t, data.number, THREADS)
+}
+
+@test
+test_auto_reset :: proc(t: ^testing.T) {
+	testing.set_fail_timeout(t, FAIL_TIME)
+
+	Data :: struct {
+		a: sync.Auto_Reset_Event,
+		number: int,
+	}
+
+	p :: proc(th: ^thread.Thread) {
+		data := cast(^Data)th.data
+
+		// log.debugf("AUR-%v> entering", th.id)
+		sync.auto_reset_event_wait(&data.a)
+		// log.debugf("AUR-%v> adding", th.id)
+		data.number += 1
+		// log.debugf("AUR-%v> signalling", th.id)
+		sync.auto_reset_event_signal(&data.a)
+		// log.debugf("AUR-%v> leaving", th.id)
+	}
+
+	data: Data
+	threads: [THREADS]^thread.Thread
+
+	for &v in threads {
+		v = thread.create(p)
+		v.data = &data
+		v.init_context = context
+		thread.start(v)
+	}
+
+	// There is a chance that this test can stall if a signal is sent before
+	// all threads are queued, because it's possible for some number of threads
+	// to get to the waiting state, the signal to fire, all of the waited
+	// threads to pass successfully, then the other threads come in with no-one
+	// to run a signal.
+	//
+	// So we'll just test a fully-waited queue of cascading threads.
+	for {
+		status := intrinsics.atomic_load(&data.a.status)
+		if status == -THREADS {
+			// log.debug("All Auto_Reset_Event threads have queued.")
+			break
+		}
+		intrinsics.cpu_relax()
+	}
+
+	sync.auto_reset_event_signal(&data.a)
+
+	wait_for(threads[:])
+
+	// The last thread should leave this primitive in a signalled state.
+	testing.expect_value(t, data.a.status, 1)
+	testing.expect_value(t, data.number, THREADS)
+}
+
+@test
+test_auto_reset_already_signalled :: proc(t: ^testing.T) {
+	testing.set_fail_timeout(t, FAIL_TIME)
+
+	a: sync.Auto_Reset_Event
+	sync.auto_reset_event_signal(&a)
+	sync.auto_reset_event_wait(&a)
+	testing.expect_value(t, a.status, 0)
+}
+
+@test
+test_ticket_mutex :: proc(t: ^testing.T) {
+	testing.set_fail_timeout(t, FAIL_TIME)
+
+	Data :: struct {
+		m: sync.Ticket_Mutex,
+		number: int,
+	}
+
+	p :: proc(th: ^thread.Thread) {
+		data := cast(^Data)th.data
+
+		// log.debugf("TIC-%i> entering", th.id)
+		// intrinsics.debug_trap()
+		sync.ticket_mutex_lock(&data.m)
+		// log.debugf("TIC-%i> locked", th.id)
+		data.number += 1
+		// log.debugf("TIC-%i> unlocking", th.id)
+		sync.ticket_mutex_unlock(&data.m)
+		// log.debugf("TIC-%i> leaving", th.id)
+	}
+
+	data: Data
+	threads: [THREADS]^thread.Thread
+
+	for &v in threads {
+		v = thread.create(p)
+		v.data = &data
+		v.init_context = context
+		thread.start(v)
+	}
+
+	wait_for(threads[:])
+
+	testing.expect_value(t, data.m.ticket, THREADS)
+	testing.expect_value(t, data.m.serving, THREADS)
+	testing.expect_value(t, data.number, THREADS)
+}
+
+@test
+test_benaphore :: proc(t: ^testing.T) {
+	testing.set_fail_timeout(t, FAIL_TIME)
+
+	Data :: struct {
+		b: sync.Benaphore,
+		number: int,
+	}
+
+	p :: proc(th: ^thread.Thread) {
+		data := cast(^Data)th.data
+		sync.benaphore_lock(&data.b)
+		data.number += 1
+		sync.benaphore_unlock(&data.b)
+	}
+
+	data: Data
+	threads: [THREADS]^thread.Thread
+
+	for &v in threads {
+		v = thread.create(p)
+		v.data = &data
+		v.init_context = context
+		thread.start(v)
+	}
+
+	wait_for(threads[:])
+
+	testing.expect_value(t, data.b.counter, 0)
+	testing.expect_value(t, data.number, THREADS)
+}
+
+@test
+test_recursive_benaphore :: proc(t: ^testing.T) {
+	testing.set_fail_timeout(t, FAIL_TIME)
+
+	Data :: struct {
+		b: sync.Recursive_Benaphore,
+		number: int,
+	}
+
+	p :: proc(th: ^thread.Thread) {
+		data := cast(^Data)th.data
+
+		// log.debugf("REC_BEP-%i> entering", th.id)
+		tried1 := sync.recursive_benaphore_try_lock(&data.b)
+		for _ in 0..<3 {
+			sync.recursive_benaphore_lock(&data.b)
+		}
+		tried2 := sync.recursive_benaphore_try_lock(&data.b)
+		// log.debugf("REC_BEP-%i> locked", th.id)
+		data.number += 1
+		for _ in 0..<3 {
+			sync.recursive_benaphore_unlock(&data.b)
+		}
+		if tried1 { sync.recursive_benaphore_unlock(&data.b) }
+		if tried2 { sync.recursive_benaphore_unlock(&data.b) }
+		// log.debugf("REC_BEP-%i> leaving", th.id)
+	}
+
+	data: Data
+	threads: [THREADS]^thread.Thread
+
+	for &v in threads {
+		v = thread.create(p)
+		v.data = &data
+		v.init_context = context
+		thread.start(v)
+	}
+
+	wait_for(threads[:])
+
+	// The benaphore should be unowned at the end.
+	testing.expect_value(t, data.b.counter, 0)
+	testing.expect_value(t, data.b.owner, 0)
+	testing.expect_value(t, data.b.recursion, 0)
+	testing.expect_value(t, data.number, THREADS)
+}
+
+@test
+test_once :: proc(t: ^testing.T) {
+	testing.set_fail_timeout(t, FAIL_TIME)
+
+	Data :: struct {
+		once: sync.Once,
+		number: int,
+	}
+
+	write :: proc "contextless" (data: rawptr) {
+		data := cast(^Data)data
+		data.number += 1
+	}
+
+	p :: proc(th: ^thread.Thread) {
+		data := cast(^Data)th.data
+		// log.debugf("ONCE-%v> entering", th.id)
+		sync.once_do_with_data_contextless(&data.once, write, data)
+		// log.debugf("ONCE-%v> leaving", th.id)
+	}
+
+	data: Data
+	threads: [THREADS]^thread.Thread
+
+	for &v in threads {
+		v = thread.create(p)
+		v.data = &data
+		v.init_context = context
+		thread.start(v)
+	}
+
+	wait_for(threads[:])
+
+	testing.expect_value(t, data.once.done, true)
+	testing.expect_value(t, data.number, 1)
+}
+
+@test
+test_park :: proc(t: ^testing.T) {
+	testing.set_fail_timeout(t, FAIL_TIME)
+
+	Data :: struct {
+		car: sync.Parker,
+		number: int,
+	}
+
+	data: Data
+
+	th := thread.create_and_start_with_data(&data, proc(data: rawptr) {
+		data := cast(^Data)data
+		time.sleep(SLEEP_TIME)
+		sync.unpark(&data.car)
+		data.number += 1
+	})
+
+	sync.park(&data.car)
+
+	wait_for([]^thread.Thread{ th })
+
+	PARKER_EMPTY :: 0
+	testing.expect_value(t, data.car.state, PARKER_EMPTY)
+	testing.expect_value(t, data.number, 1)
+}
+
+@test
+test_park_with_timeout :: proc(t: ^testing.T) {
+	testing.set_fail_timeout(t, FAIL_TIME)
+
+	car: sync.Parker
+	sync.park_with_timeout(&car, SLEEP_TIME)
+}
+
+@test
+test_one_shot_event :: proc(t: ^testing.T) {
+	testing.set_fail_timeout(t, FAIL_TIME)
+
+	Data :: struct {
+		event: sync.One_Shot_Event,
+		number: int,
+	}
+
+	data: Data
+
+	th := thread.create_and_start_with_data(&data, proc(data: rawptr) {
+		data := cast(^Data)data
+		time.sleep(SLEEP_TIME)
+		sync.one_shot_event_signal(&data.event)
+		data.number += 1
+	})
+
+	sync.one_shot_event_wait(&data.event)
+
+	wait_for([]^thread.Thread{ th })
+
+	testing.expect_value(t, data.event.state, 1)
+	testing.expect_value(t, data.number, 1)
+}

From d38f5ffb49733b5084b9419412acf9381e3073d4 Mon Sep 17 00:00:00 2001
From: Feoramund <161657516+Feoramund@users.noreply.github.com>
Date: Sun, 15 Sep 2024 22:00:53 -0400
Subject: [PATCH 22/23] Remove unneeded synchronizations in `Chan`

Everything was already guarded by `c.mutex`.
---
 core/sync/chan/chan.odin | 78 ++++++++++++++++++----------------------
 1 file changed, 35 insertions(+), 43 deletions(-)

diff --git a/core/sync/chan/chan.odin b/core/sync/chan/chan.odin
index 5b9a764b47a..c470d15f3e5 100644
--- a/core/sync/chan/chan.odin
+++ b/core/sync/chan/chan.odin
@@ -22,19 +22,17 @@ Raw_Chan :: struct {
 	allocator:       runtime.Allocator,
 	allocation_size: int,
 	msg_size:        u16,
-	closed:          b16, // atomic
+	closed:          b16, // guarded by `mutex`
 	mutex:           sync.Mutex,
 	r_cond:          sync.Cond,
 	w_cond:          sync.Cond,
-	r_waiting:       int,  // atomic
-	w_waiting:       int,  // atomic
+	r_waiting:       int,  // guarded by `mutex`
+	w_waiting:       int,  // guarded by `mutex`
 
 	// Buffered
 	queue: ^Raw_Queue,
 
 	// Unbuffered
-	r_mutex:         sync.Mutex,
-	w_mutex:         sync.Mutex,
 	unbuffered_data: rawptr,
 }
 
@@ -164,32 +162,30 @@ send_raw :: proc "contextless" (c: ^Raw_Chan, msg_in: rawptr) -> (ok: bool) {
 	}
 	if c.queue != nil { // buffered
 		sync.guard(&c.mutex)
-		for !sync.atomic_load(&c.closed) &&
-		    c.queue.len == c.queue.cap {
-			sync.atomic_add(&c.w_waiting, 1)
+		for !c.closed && c.queue.len == c.queue.cap {
+			c.w_waiting += 1
 			sync.wait(&c.w_cond, &c.mutex)
-			sync.atomic_sub(&c.w_waiting, 1)
+			c.w_waiting -= 1
 		}
 
-		if sync.atomic_load(&c.closed) {
+		if c.closed {
 			return false
 		}
 
 		ok = raw_queue_push(c.queue, msg_in)
-		if sync.atomic_load(&c.r_waiting) > 0 {
+		if c.r_waiting > 0 {
 			sync.signal(&c.r_cond)
 		}
 	} else if c.unbuffered_data != nil { // unbuffered
-		sync.guard(&c.w_mutex)
 		sync.guard(&c.mutex)
 
-		if sync.atomic_load(&c.closed) {
+		if c.closed {
 			return false
 		}
 
 		mem.copy(c.unbuffered_data, msg_in, int(c.msg_size))
-		sync.atomic_add(&c.w_waiting, 1)
-		if sync.atomic_load(&c.r_waiting) > 0 {
+		c.w_waiting += 1
+		if c.r_waiting > 0 {
 			sync.signal(&c.r_cond)
 		}
 		sync.wait(&c.w_cond, &c.mutex)
@@ -206,13 +202,13 @@ recv_raw :: proc "contextless" (c: ^Raw_Chan, msg_out: rawptr) -> (ok: bool) {
 	if c.queue != nil { // buffered
 		sync.guard(&c.mutex)
 		for c.queue.len == 0 {
-			if sync.atomic_load(&c.closed) {
+			if c.closed {
 				return
 			}
 
-			sync.atomic_add(&c.r_waiting, 1)
+			c.r_waiting += 1
 			sync.wait(&c.r_cond, &c.mutex)
-			sync.atomic_sub(&c.r_waiting, 1)
+			c.r_waiting -= 1
 		}
 
 		msg := raw_queue_pop(c.queue)
@@ -220,27 +216,26 @@ recv_raw :: proc "contextless" (c: ^Raw_Chan, msg_out: rawptr) -> (ok: bool) {
 			mem.copy(msg_out, msg, int(c.msg_size))
 		}
 
-		if sync.atomic_load(&c.w_waiting) > 0 {
+		if c.w_waiting > 0 {
 			sync.signal(&c.w_cond)
 		}
 		ok = true
 	} else if c.unbuffered_data != nil { // unbuffered
-		sync.guard(&c.r_mutex)
 		sync.guard(&c.mutex)
 
-		for !sync.atomic_load(&c.closed) &&
-		    sync.atomic_load(&c.w_waiting) == 0 {
-			sync.atomic_add(&c.r_waiting, 1)
+		for !c.closed &&
+			c.w_waiting == 0 {
+			c.r_waiting += 1
 			sync.wait(&c.r_cond, &c.mutex)
-			sync.atomic_sub(&c.r_waiting, 1)
+			c.r_waiting -= 1
 		}
 
-		if sync.atomic_load(&c.closed) {
+		if c.closed {
 			return
 		}
 
 		mem.copy(msg_out, c.unbuffered_data, int(c.msg_size))
-		sync.atomic_sub(&c.w_waiting, 1)
+		c.w_waiting -= 1
 
 		sync.signal(&c.w_cond)
 		ok = true
@@ -260,25 +255,24 @@ try_send_raw :: proc "contextless" (c: ^Raw_Chan, msg_in: rawptr) -> (ok: bool)
 			return false
 		}
 
-		if sync.atomic_load(&c.closed) {
+		if c.closed {
 			return false
 		}
 
 		ok = raw_queue_push(c.queue, msg_in)
-		if sync.atomic_load(&c.r_waiting) > 0 {
+		if c.r_waiting > 0 {
 			sync.signal(&c.r_cond)
 		}
 	} else if c.unbuffered_data != nil { // unbuffered
-		sync.guard(&c.w_mutex)
 		sync.guard(&c.mutex)
 
-		if sync.atomic_load(&c.closed) {
+		if c.closed {
 			return false
 		}
 
 		mem.copy(c.unbuffered_data, msg_in, int(c.msg_size))
-		sync.atomic_add(&c.w_waiting, 1)
-		if sync.atomic_load(&c.r_waiting) > 0 {
+		c.w_waiting += 1
+		if c.r_waiting > 0 {
 			sync.signal(&c.r_cond)
 		}
 		sync.wait(&c.w_cond, &c.mutex)
@@ -303,21 +297,19 @@ try_recv_raw :: proc "contextless" (c: ^Raw_Chan, msg_out: rawptr) -> bool {
 			mem.copy(msg_out, msg, int(c.msg_size))
 		}
 
-		if sync.atomic_load(&c.w_waiting) > 0 {
+		if c.w_waiting > 0 {
 			sync.signal(&c.w_cond)
 		}
 		return true
 	} else if c.unbuffered_data != nil { // unbuffered
-		sync.guard(&c.r_mutex)
 		sync.guard(&c.mutex)
 
-		if sync.atomic_load(&c.closed) ||
-		   sync.atomic_load(&c.w_waiting) == 0 {
+		if c.closed || c.w_waiting == 0 {
 			return false
 		}
 
 		mem.copy(msg_out, c.unbuffered_data, int(c.msg_size))
-		sync.atomic_sub(&c.w_waiting, 1)
+		c.w_waiting -= 1
 
 		sync.signal(&c.w_cond)
 		return true
@@ -360,10 +352,10 @@ close :: proc "contextless" (c: ^Raw_Chan) -> bool {
 		return false
 	}
 	sync.guard(&c.mutex)
-	if sync.atomic_load(&c.closed) {
+	if c.closed {
 		return false
 	}
-	sync.atomic_store(&c.closed, true)
+	c.closed = true
 	sync.broadcast(&c.r_cond)
 	sync.broadcast(&c.w_cond)
 	return true
@@ -375,7 +367,7 @@ is_closed :: proc "contextless" (c: ^Raw_Chan) -> bool {
 		return true
 	}
 	sync.guard(&c.mutex)
-	return bool(sync.atomic_load(&c.closed))
+	return bool(c.closed)
 }
 
 
@@ -434,7 +426,7 @@ can_recv :: proc "contextless" (c: ^Raw_Chan) -> bool {
 	if is_buffered(c) {
 		return c.queue.len > 0
 	}
-	return sync.atomic_load(&c.w_waiting) > 0
+	return c.w_waiting > 0
 }
 
 
@@ -444,7 +436,7 @@ can_send :: proc "contextless" (c: ^Raw_Chan) -> bool {
 	if is_buffered(c) {
 		return c.queue.len < c.queue.cap
 	}
-	return sync.atomic_load(&c.w_waiting) == 0
+	return c.w_waiting == 0
 }
 
 
@@ -493,4 +485,4 @@ select_raw :: proc "odin" (recvs: []^Raw_Chan, sends: []^Raw_Chan, send_msgs: []
 		ok = send_raw(sends[sel.idx], send_msgs[sel.idx])
 	}
 	return
-}
\ No newline at end of file
+}

From 16ef59700b68989beff48039a450ef6153b2e6af Mon Sep 17 00:00:00 2001
From: Feoramund <161657516+Feoramund@users.noreply.github.com>
Date: Sun, 15 Sep 2024 23:58:03 -0400
Subject: [PATCH 23/23] Check for `EINTR` in `sys/posix` test

---
 tests/core/sys/posix/structs.odin | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/tests/core/sys/posix/structs.odin b/tests/core/sys/posix/structs.odin
index bdb1c24e316..95b6258f4e9 100644
--- a/tests/core/sys/posix/structs.odin
+++ b/tests/core/sys/posix/structs.odin
@@ -63,6 +63,9 @@ execute_struct_checks :: proc(t: ^testing.T) {
 		waiting: for {
 			status: i32
 			wpid := posix.waitpid(pid, &status, {})
+			if status == posix.EINTR {
+				continue
+			}
 			if !testing.expectf(t, wpid != -1, "waitpid() failure: %v", posix.strerror()) {
 				return false
 			}