Improve error message on undefined operators

Added missing Windows functions for console manipulation

.github/workflows/stale.yml

@@ -13,7 +13,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Close Stale Issues
-        uses: actions/stale@v4.1.0
+        uses: actions/stale@v7.0.0
         with:
 #          stale-issue-message: |
 #            Hello!
@@ -36,7 +36,7 @@ jobs:
 #           The motivation for this automation is to help prioritize issues in the backlog and not ignore, reject, or belittle anyone..
 
           days-before-stale: 120
-          days-before-close: 30
+          days-before-close: -1
           exempt-draft-pr: true
           ascending: true
           operations-per-run: 1000

build.bat

@@ -62,12 +62,14 @@ if %release_mode% EQU 0 ( rem Debug
 set compiler_warnings= ^
 	-W4 -WX ^
 	-wd4100 -wd4101 -wd4127 -wd4146 ^
+	-wd4505 ^
 	-wd4456 -wd4457
 
 set compiler_includes= ^
 	/Isrc\
 set libs= ^
 	kernel32.lib ^
+	Synchronization.lib ^
 	bin\llvm\windows\LLVM-C.lib
 
 set linker_flags= -incremental:no -opt:ref -subsystem:console
@@ -94,4 +96,4 @@ if %release_mode% EQU 0 odin run examples/demo
 
 del *.obj > NUL 2> NUL
 
-:end_of_build
+:end_of_build

build_odin.sh

@@ -50,7 +50,7 @@ config_darwin() {
 		panic "Requirement: llvm-config must be base version smaller than 15"
 	fi
 
-	LDFLAGS="$LDFLAGS -liconv -ldl"
+	LDFLAGS="$LDFLAGS -liconv -ldl -framework System"
 	CXXFLAGS="$CXXFLAGS $($LLVM_CONFIG --cxxflags --ldflags)"
 	LDFLAGS="$LDFLAGS -lLLVM-C"
 }

core/builtin/builtin.odin

@@ -109,7 +109,7 @@ jmag       :: proc(value: Quaternion) -> Float ---
 kmag       :: proc(value: Quaternion) -> Float ---
 conj       :: proc(value: Complex_Or_Quaternion) -> Complex_Or_Quaternion ---
 
-expand_to_tuple :: proc(value: Struct_Or_Array) -> (A, B, C, ...) ---
+expand_values :: proc(value: Struct_Or_Array) -> (A, B, C, ...) ---
 
 min   :: proc(values: ..T) -> T ---
 max   :: proc(values: ..T) -> T ---

core/c/libc/threads.odin

@@ -108,7 +108,7 @@ when ODIN_OS == .Linux {
 		cnd_destroy   :: proc(cond: ^cnd_t) ---
 		cnd_init      :: proc(cond: ^cnd_t) -> int ---
 		cnd_signal    :: proc(cond: ^cnd_t) -> int ---
-		cnd_timedwait :: proc(cond: ^cnd_t, ts: ^timespec) -> int ---
+		cnd_timedwait :: proc(cond: ^cnd_t, mtx: ^mtx_t, ts: ^timespec) -> int ---
 		cnd_wait      :: proc(cond: ^cnd_t, mtx: ^mtx_t) -> int ---
 		
 		// 7.26.4 Mutex functions

core/compress/common.odin

@@ -212,6 +212,8 @@ read_slice_from_memory :: #force_inline proc(z: ^Context_Memory_Input, size: int
 
 @(optimization_mode="speed")
 read_slice_from_stream :: #force_inline proc(z: ^Context_Stream_Input, size: int) -> (res: []u8, err: io.Error) {
+	// TODO: REMOVE ALL USE OF context.temp_allocator here
+	// the is literally no need for it
 	b := make([]u8, size, context.temp_allocator)
 	_, e := z.input->impl_read(b[:])
 	if e == .None {

core/container/small_array/small_array.odin

@@ -1,6 +1,8 @@
 package container_small_array
 
 import "core:builtin"
+import "core:runtime"
+_ :: runtime
 
 Small_Array :: struct($N: int, $T: typeid) where N >= 0 {
 	data: [N]T,
@@ -32,6 +34,20 @@ get_ptr :: proc "contextless" (a: ^$A/Small_Array($N, $T), index: int) -> ^T {
 	return &a.data[index]
 }
 
+get_safe :: proc(a: $A/Small_Array($N, $T), index: int) -> (T, bool) #no_bounds_check {
+	if index < 0 || index >= a.len {
+		return {}, false
+	}
+	return a.data[index], true
+}
+
+get_ptr_safe :: proc(a: ^$A/Small_Array($N, $T), index: int) -> (^T, bool) #no_bounds_check {
+	if index < 0 || index >= a.len {
+		return {}, false
+	}
+	return &a.data[index], true
+}
+
 set :: proc "contextless" (a: ^$A/Small_Array($N, $T), index: int, item: T) {
 	a.data[index] = item
 }
@@ -93,7 +109,7 @@ pop_front_safe :: proc "contextless" (a: ^$A/Small_Array($N, $T)) -> (item: T, o
 		copy(s[:], s[1:])
 		a.len -= 1
 		ok = true
-	} 
+	}
 	return
 }
 
@@ -102,6 +118,23 @@ consume :: proc "odin" (a: ^$A/Small_Array($N, $T), count: int, loc := #caller_l
 	a.len -= count
 }
 
+ordered_remove :: proc "contextless" (a: ^$A/Small_Array($N, $T), index: int, loc := #caller_location) #no_bounds_check {
+    runtime.bounds_check_error_loc(loc, index, a.len)
+    if index+1 < a.len {
+		copy(a.data[index:], a.data[index+1:])
+	}
+	a.len -= 1
+}
+
+unordered_remove :: proc "contextless" (a: ^$A/Small_Array($N, $T), index: int, loc := #caller_location) #no_bounds_check {
+    runtime.bounds_check_error_loc(loc, index, a.len)
+	n := a.len-1
+	if index != n {
+		a.data[index] = a.data[n]
+	}
+    a.len -= 1
+}
+
 clear :: proc "contextless" (a: ^$A/Small_Array($N, $T)) {
 	resize(a, 0)
 }
@@ -111,6 +144,18 @@ push_back_elems :: proc "contextless" (a: ^$A/Small_Array($N, $T), items: ..T) {
 	a.len += n
 }
 
+inject_at :: proc "contextless" (a: ^$A/Small_Array($N, $T), item: T, index: int) -> bool #no_bounds_check {
+	if a.len < cap(a^) && index >= 0 && index <= len(a^) {
+		a.len += 1
+		for i := a.len - 1; i >= index + 1; i -= 1 {
+			a.data[i] = a.data[i - 1]
+		}
+		a.data[index] = item
+		return true
+	}
+	return false
+}
+
 append_elem  :: push_back
 append_elems :: push_back_elems
 push   :: proc{push_back, push_back_elems}

core/crypto/rand_linux.odin

@@ -12,7 +12,7 @@ _rand_bytes :: proc (dst: []byte) {
 
 	for l > 0 {
 		to_read := min(l, _MAX_PER_CALL_BYTES)
-		ret := unix.sys_getrandom(raw_data(dst), to_read, 0)
+		ret := unix.sys_getrandom(raw_data(dst), uint(to_read), 0)
 		if ret < 0 {
 			switch os.Errno(-ret) {
 			case os.EINTR:

core/dynlib/doc.odin

@@ -0,0 +1,7 @@
+/*
+Package core:dynlib implements loading of shared libraries/DLLs and their symbols.
+
+The behaviour of dynamically loaded libraries is specific to the target platform of the program.
+For in depth detail on the underlying behaviour please refer to your target platform's documentation.
+*/
+package dynlib

core/dynlib/lib.odin

@@ -1,15 +1,94 @@
 package dynlib
 
+/*
+A handle to a dynamically loaded library.
+*/
 Library :: distinct rawptr
 
-load_library :: proc(path: string, global_symbols := false) -> (Library, bool) {
+/*
+Loads a dynamic library from the filesystem. The paramater `global_symbols` makes the symbols in the loaded
+library available to resolve references in subsequently loaded libraries.
+
+The paramater `global_symbols` is only used for the platforms `linux`, `darwin`, `freebsd` and `openbsd`.
+On `windows` this paramater is ignored.
+
+The underlying behaviour is platform specific.  
+On `linux`, `darwin`, `freebsd` and `openbsd` refer to `dlopen`.  
+On `windows` refer to `LoadLibraryW`.
+
+**Implicit Allocators**  
+`context.temp_allocator`
+
+Example:
+	import "core:dynlib"
+	import "core:fmt"
+
+	load_my_library :: proc() {
+		LIBRARY_PATH :: "my_library.dll"
+		library, ok := dynlib.load_library(LIBRARY_PATH)
+		if ! ok {
+			return
+		}
+		fmt.println("The library %q was successfully loaded", LIBRARY_PATH)
+	}
+*/
+load_library :: proc(path: string, global_symbols := false) -> (library: Library, did_load: bool) {
 	return _load_library(path, global_symbols)
 }
 
-unload_library :: proc(library: Library) -> bool {
+/*
+Unloads a dynamic library.
+
+The underlying behaviour is platform specific.  
+On `linux`, `darwin`, `freebsd` and `openbsd` refer to `dlclose`.  
+On `windows` refer to `FreeLibrary`.
+
+Example:
+	import "core:dynlib"
+	import "core:fmt"
+
+	load_then_unload_my_library :: proc() {
+		LIBRARY_PATH :: "my_library.dll"
+		library, ok := dynlib.load_library(LIBRARY_PATH)
+		if ! ok {
+			return
+		}
+		did_unload := dynlib.unload_library(library)
+		if ! did_unload {
+			return
+		}
+		fmt.println("The library %q was successfully unloaded", LIBRARY_PATH)
+	}
+*/
+unload_library :: proc(library: Library) -> (did_unload: bool) {
 	return _unload_library(library)
 }
 
+/*
+Loads the address of a procedure/variable from a dynamic library.
+
+The underlying behaviour is platform specific.  
+On `linux`, `darwin`, `freebsd` and `openbsd` refer to `dlsym`.  
+On `windows` refer to `GetProcAddress`.
+
+**Implicit Allocators**  
+`context.temp_allocator`
+
+Example:
+	import "core:dynlib"
+	import "core:fmt"
+
+	find_a_in_my_library :: proc() {
+		LIBRARY_PATH :: "my_library.dll"
+		library, ok := dynlib.load_library(LIBRARY_PATH)
+		if ! ok {
+			return
+		}
+
+		a, found_a := dynlib.symbol_address(library, "a")
+		if found_a do fmt.printf("The symbol %q was found at the address %v", "a", a)
+	}
+*/
 symbol_address :: proc(library: Library, symbol: string) -> (ptr: rawptr, found: bool) #optional_ok {
 	return _symbol_address(library, symbol)
 }

core/dynlib/lib_windows.odin

@@ -4,10 +4,12 @@ package dynlib
 
 import win32 "core:sys/windows"
 import "core:strings"
+import "core:runtime"
 
 _load_library :: proc(path: string, global_symbols := false) -> (Library, bool) {
 	// NOTE(bill): 'global_symbols' is here only for consistency with POSIX which has RTLD_GLOBAL
 
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	wide_path := win32.utf8_to_wstring(path, context.temp_allocator)
 	handle := cast(Library)win32.LoadLibraryW(wide_path)
 	return handle, handle != nil
@@ -19,6 +21,7 @@ _unload_library :: proc(library: Library) -> bool {
 }
 
 _symbol_address :: proc(library: Library, symbol: string) -> (ptr: rawptr, found: bool) {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	c_str := strings.clone_to_cstring(symbol, context.temp_allocator)
 	ptr = win32.GetProcAddress(cast(win32.HMODULE)library, c_str)
 	found = ptr != nil

core/encoding/json/marshal.odin

@@ -198,7 +198,7 @@ marshal_to_writer :: proc(w: io.Writer, v: any, opt: ^Marshal_Options) -> (err:
 	case runtime.Type_Info_Procedure:
 		return .Unsupported_Type
 
-	case runtime.Type_Info_Tuple:
+	case runtime.Type_Info_Parameters:
 		return .Unsupported_Type
 
 	case runtime.Type_Info_Simd_Vector:
@@ -262,10 +262,14 @@ marshal_to_writer :: proc(w: io.Writer, v: any, opt: ^Marshal_Options) -> (err:
 			}
 			map_cap := uintptr(runtime.map_cap(m^))
 			ks, vs, hs, _, _ := runtime.map_kvh_data_dynamic(m^, info.map_info)
+
+			i := 0
 			for bucket_index in 0..<map_cap {
 				if !runtime.map_hash_is_valid(hs[bucket_index]) {
 					continue
 				}
+				opt_write_iteration(w, opt, i) or_return
+				i += 1
 
 				key   := rawptr(runtime.map_cell_index_dynamic(ks, info.map_info.ks, bucket_index))
 				value := rawptr(runtime.map_cell_index_dynamic(vs, info.map_info.vs, bucket_index))

core/encoding/json/unmarshal.odin

@@ -346,6 +346,8 @@ unmarshal_object :: proc(p: ^Parser, v: any, end_token: Token_Kind) -> (err: Unm
 			
 			fields := reflect.struct_fields_zipped(ti.id)
 			
+			runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD(ignore = context.temp_allocator == context.allocator)
+
 			field_used := make([]bool, len(fields), context.temp_allocator)
 			
 			use_field_idx := -1

core/encoding/xml/xml_reader.odin

@@ -33,6 +33,7 @@ import "core:intrinsics"
 import "core:mem"
 import "core:os"
 import "core:strings"
+import "core:runtime"
 
 likely :: intrinsics.expect
 
@@ -408,7 +409,7 @@ parse_bytes :: proc(data: []u8, options := DEFAULT_OPTIONS, path := "", error_ha
 				next := scan(t)
 				#partial switch next.kind {
 				case .Ident:
-					if len(next.text) == 3 && strings.to_lower(next.text, context.temp_allocator) == "xml" {
+					if len(next.text) == 3 && strings.equal_fold(next.text, "xml") {
 						parse_prologue(doc) or_return
 					} else if len(doc.prologue) > 0 {
 						/*
@@ -614,6 +615,7 @@ parse_prologue :: proc(doc: ^Document) -> (err: Error) {
 			}
 
 		case "encoding":
+			runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 			switch strings.to_lower(attr.val, context.temp_allocator) {
 			case "utf-8", "utf8":
 				doc.encoding = .UTF_8

core/fmt/fmt.odin

@@ -547,7 +547,7 @@ _parse_int :: proc(s: string, offset: int) -> (result: int, new_offset: int, ok:
 	is_digit :: #force_inline proc(r: byte) -> bool { return '0' <= r && r <= '9' }
 
 	new_offset = offset
-	for new_offset <= len(s) {
+	for new_offset < len(s) {
 		c := s[new_offset]
 		if !is_digit(c) {
 			break
@@ -678,7 +678,7 @@ _fmt_int :: proc(fi: ^Info, u: u64, base: int, is_signed: bool, bit_size: int, d
 		}
 	} else if fi.zero && fi.width_set {
 		prec = fi.width
-		if neg || fi.plus || fi.space {
+		if neg || fi.plus {
 			// There needs to be space for the "sign"
 			prec -= 1
 		}
@@ -697,7 +697,6 @@ _fmt_int :: proc(fi: ^Info, u: u64, base: int, is_signed: bool, bit_size: int, d
 	flags: strconv.Int_Flags
 	if fi.hash && !fi.zero { flags |= {.Prefix} }
 	if fi.plus             { flags |= {.Plus}   }
-	if fi.space            { flags |= {.Space}  }
 	s := strconv.append_bits(buf[start:], u, base, is_signed, bit_size, digits, flags)
 
 	if fi.hash && fi.zero && fi.indent == 0 {
@@ -744,7 +743,7 @@ _fmt_int_128 :: proc(fi: ^Info, u: u128, base: int, is_signed: bool, bit_size: i
 		}
 	} else if fi.zero && fi.width_set {
 		prec = fi.width
-		if neg || fi.plus || fi.space {
+		if neg || fi.plus {
 			// There needs to be space for the "sign"
 			prec -= 1
 		}
@@ -763,7 +762,6 @@ _fmt_int_128 :: proc(fi: ^Info, u: u128, base: int, is_signed: bool, bit_size: i
 	flags: strconv.Int_Flags
 	if fi.hash && !fi.zero { flags |= {.Prefix} }
 	if fi.plus             { flags |= {.Plus}   }
-	if fi.space            { flags |= {.Space}  }
 	s := strconv.append_bits_128(buf[start:], u, base, is_signed, bit_size, digits, flags)
 
 	if fi.hash && fi.zero && fi.indent == 0 {
@@ -867,79 +865,30 @@ _pad :: proc(fi: ^Info, s: string) {
 	}
 }
 
+_fmt_float_as :: proc(fi: ^Info, v: f64, bit_size: int, verb: rune, float_fmt: byte) {
+	prec := fi.prec if fi.prec_set else 3
+	buf: [386]byte
+
+	// Can return "NaN", "+Inf", "-Inf", "+<value>", "-<value>".
+	str := strconv.append_float(buf[:], v, float_fmt, prec, bit_size)
+
+	if !fi.plus {
+		// Strip sign from "+<value>" but not "+Inf".
+		if str[0] == '+' && str[1] != 'I' {
+			str = str[1:] 
+		}
+	}
+
+	_pad(fi, str)
+}
+
 fmt_float :: proc(fi: ^Info, v: f64, bit_size: int, verb: rune) {
 	switch verb {
 	case 'f', 'F', 'g', 'G', 'v':
-		prec: int = 3
-		if fi.prec_set {
-			prec = fi.prec
-		}
-		buf: [386]byte
-
-		str := strconv.append_float(buf[1:], v, 'f', prec, bit_size)
-		b := buf[:len(str)+1]
-		if b[1] == '+' || b[1] == '-' {
-			b = b[1:]
-		} else {
-			b[0] = '+'
-		}
-
-		if fi.space && !fi.plus && b[0] == '+' {
-			b[0] = ' '
-		}
-
-		if len(b) > 1 && (b[1] == 'N' || b[1] == 'I') {
-			io.write_string(fi.writer, string(b), &fi.n)
-			return
-		}
-
-		if fi.plus || b[0] != '+' {
-			if fi.zero && fi.width_set && fi.width > len(b) {
-				io.write_byte(fi.writer, b[0], &fi.n)
-				fmt_write_padding(fi, fi.width - len(b))
-				io.write_string(fi.writer, string(b[1:]), &fi.n)
-			} else {
-				_pad(fi, string(b))
-			}
-		} else {
-			_pad(fi, string(b[1:]))
-		}
-
+		_fmt_float_as(fi, v, bit_size, verb, 'f')
 	case 'e', 'E':
-		prec: int = 3
-		if fi.prec_set {
-			prec = fi.prec
-		}
-		buf: [386]byte
-
-		str := strconv.append_float(buf[1:], v, 'e', prec, bit_size)
-		b := buf[:len(str)+1]
-		if b[1] == '+' || b[1] == '-' {
-			b = b[1:]
-		} else {
-			b[0] = '+'
-		}
-
-		if fi.space && !fi.plus && b[0] == '+' {
-			b[0] = ' '
-		}
-
-		if len(b) > 1 && (b[1] == 'N' || b[1] == 'I') {
-			io.write_string(fi.writer, string(b), &fi.n)
-			return
-		}
-
-		if fi.plus || str[0] != '+' {
-			if fi.zero && fi.width_set && fi.width > len(b) {
-				io.write_byte(fi.writer, b[0], &fi.n)
-				fmt_write_padding(fi, fi.width - len(b))
-				io.write_string(fi.writer, string(b[1:]), &fi.n)
-			} else {
-				_pad(fi, string(b))
-			}
-		} else {
-			_pad(fi, string(b[1:]))
-		}
+		// BUG(): "%.3e" returns "3.000e+00"
+		_fmt_float_as(fi, v, bit_size, verb, 'e')
 
 	case 'h', 'H':
 		prev_fi := fi^
@@ -1834,8 +1783,8 @@ fmt_value :: proc(fi: ^Info, v: any, verb: rune) {
 
 	type_info := type_info_of(v.id)
 	switch info in type_info.variant {
-	case runtime.Type_Info_Any:   // Ignore
-	case runtime.Type_Info_Tuple: // Ignore
+	case runtime.Type_Info_Any:        // Ignore
+	case runtime.Type_Info_Parameters: // Ignore
 
 	case runtime.Type_Info_Named:
 		fmt_named(fi, v, verb, info)

core/hash/xxhash/xxhash_3.odin

@@ -118,7 +118,7 @@ XXH_mul_64_to_128_fold_64 :: #force_inline proc(lhs, rhs: xxh_u64) -> (res: xxh_
 }
 
 @(optimization_mode="speed")
-XXH_xorshift_64 :: #force_inline proc(v: xxh_u64, auto_cast shift: uint) -> (res: xxh_u64) {
+XXH_xorshift_64 :: #force_inline proc(v: xxh_u64, #any_int shift: uint) -> (res: xxh_u64) {
 	return v ~ (v >> shift)
 }
 

core/image/netpbm/netpbm.odin

@@ -8,6 +8,7 @@ import "core:os"
 import "core:strconv"
 import "core:strings"
 import "core:unicode"
+import "core:runtime"
 
 Image        :: image.Image
 Format       :: image.Netpbm_Format
@@ -407,6 +408,8 @@ _parse_header_pam :: proc(data: []byte, allocator := context.allocator) -> (head
 	}
 	length = header_end_index + len(HEADER_END)
 
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD(ignore = context.temp_allocator == allocator)
+
 	// string buffer for the tupltype
 	tupltype: strings.Builder
 	strings.builder_init(&tupltype, context.temp_allocator); defer strings.builder_destroy(&tupltype)

core/image/png/helpers.odin

@@ -16,6 +16,7 @@ import coretime "core:time"
 import "core:strings"
 import "core:bytes"
 import "core:mem"
+import "core:runtime"
 
 /*
 	Cleanup of image-specific data.
@@ -91,6 +92,8 @@ core_time :: proc(c: image.PNG_Chunk) -> (t: coretime.Time, ok: bool) {
 }
 
 text :: proc(c: image.PNG_Chunk) -> (res: Text, ok: bool) {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD(ignore = context.temp_allocator == context.allocator)
+
 	assert(len(c.data) == int(c.header.length))
 	#partial switch c.header.type {
 	case .tEXt:
@@ -194,18 +197,18 @@ text_destroy :: proc(text: Text) {
 }
 
 iccp :: proc(c: image.PNG_Chunk) -> (res: iCCP, ok: bool) {
-	ok = true
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD(ignore = context.temp_allocator == context.allocator)
 
 	fields := bytes.split_n(s=c.data, sep=[]u8{0}, n=3, allocator=context.temp_allocator)
 
 	if len(fields[0]) < 1 || len(fields[0]) > 79 {
 		// Invalid profile name
-		ok = false; return
+		return
 	}
 
 	if len(fields[1]) != 0 {
 		// Compression method should be a zero, which the split turned into an empty slice.
-		ok = false; return
+		return
 	}
 
 	// Set up ZLIB context and decompress iCCP payload
@@ -213,12 +216,12 @@ iccp :: proc(c: image.PNG_Chunk) -> (res: iCCP, ok: bool) {
 	zlib_error := zlib.inflate_from_byte_array(fields[2], &buf)
 	if zlib_error != nil {
 		bytes.buffer_destroy(&buf)
-		ok = false; return
+		return
 	}
 
 	res.name = strings.clone(string(fields[0]))
 	res.profile = bytes.buffer_to_bytes(&buf)
-
+	ok = true
 	return
 }
 
@@ -256,18 +259,18 @@ plte :: proc(c: image.PNG_Chunk) -> (res: PLTE, ok: bool) {
 
 splt :: proc(c: image.PNG_Chunk) -> (res: sPLT, ok: bool) {
 	if c.header.type != .sPLT {
-		return {}, false
+		return
 	}
-	ok = true
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD(ignore = context.temp_allocator == context.allocator)
 
 	fields := bytes.split_n(s=c.data, sep=[]u8{0}, n=2, allocator=context.temp_allocator)
 	if len(fields) != 2 {
-		return {}, false
+		return
 	}
 
 	res.depth = fields[1][0]
 	if res.depth != 8 && res.depth != 16 {
-		return {}, false
+		return
 	}
 
 	data := fields[1][1:]
@@ -275,21 +278,21 @@ splt :: proc(c: image.PNG_Chunk) -> (res: sPLT, ok: bool) {
 
 	if res.depth == 8 {
 		if len(data) % 6 != 0 {
-			return {}, false
+			return
 		}
 		count = len(data) / 6
 		if count > 256 {
-			return {}, false
+			return
 		}
 
 		res.entries = mem.slice_data_cast([][4]u8, data)
 	} else { // res.depth == 16
 		if len(data) % 10 != 0 {
-			return {}, false
+			return
 		}
 		count = len(data) / 10
 		if count > 256 {
-			return {}, false
+			return
 		}
 
 		res.entries = mem.slice_data_cast([][4]u16, data)
@@ -297,7 +300,7 @@ splt :: proc(c: image.PNG_Chunk) -> (res: sPLT, ok: bool) {
 
 	res.name = strings.clone(string(fields[0]))
 	res.used = u16(count)
-
+	ok = true
 	return
 }
 

core/image/png/png.odin

@@ -23,6 +23,7 @@ import "core:bytes"
 import "core:io"
 import "core:mem"
 import "core:intrinsics"
+import "core:runtime"
 
 // Limit chunk sizes.
 // By default: IDAT = 8k x 8k x 16-bits + 8k filter bytes.
@@ -1247,6 +1248,8 @@ defilter_8 :: proc(params: ^Filter_Params) -> (ok: bool) {
 
 	// TODO: See about doing a Duff's #unroll where practicable
 
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
+
 	// Apron so we don't need to special case first rows.
 	up := make([]u8, row_stride, context.temp_allocator)
 	ok = true
@@ -1299,10 +1302,9 @@ defilter_8 :: proc(params: ^Filter_Params) -> (ok: bool) {
 }
 
 // @(optimization_mode="speed")
-defilter_less_than_8 :: proc(params: ^Filter_Params) -> (ok: bool) #no_bounds_check {
+defilter_less_than_8 :: proc(params: ^Filter_Params) -> bool #no_bounds_check {
 
 	using params
-	ok = true
 
 	row_stride_in  := ((channels * width * depth) + 7) >> 3
 	row_stride_out := channels * width
@@ -1314,6 +1316,8 @@ defilter_less_than_8 :: proc(params: ^Filter_Params) -> (ok: bool) #no_bounds_ch
 
 	// TODO: See about doing a Duff's #unroll where practicable
 
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
+
 	// Apron so we don't need to special case first rows.
 	up := make([]u8, row_stride_out, context.temp_allocator)
 
@@ -1457,18 +1461,18 @@ defilter_less_than_8 :: proc(params: ^Filter_Params) -> (ok: bool) #no_bounds_ch
 		}
 	}
 
-	return
+	return true
 }
 
 // @(optimization_mode="speed")
-defilter_16 :: proc(params: ^Filter_Params) -> (ok: bool) {
-
+defilter_16 :: proc(params: ^Filter_Params) -> bool {
 	using params
-	ok = true
 
 	stride := channels * 2
 	row_stride := width * stride
 
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
+
 	// TODO: See about doing a Duff's #unroll where practicable
 	// Apron so we don't need to special case first rows.
 	up := make([]u8, row_stride, context.temp_allocator)
@@ -1518,7 +1522,7 @@ defilter_16 :: proc(params: ^Filter_Params) -> (ok: bool) {
 		dest    = dest[row_stride:]
 	}
 
-	return
+	return true
 }
 
 defilter :: proc(img: ^Image, filter_bytes: ^bytes.Buffer, header: ^image.PNG_IHDR, options: Options) -> (err: Error) {

core/intrinsics/intrinsics.odin

@@ -283,7 +283,7 @@ wasm_memory_atomic_wait32   :: proc(ptr: ^u32, expected: u32, timeout_ns: i64) -
 wasm_memory_atomic_notify32 :: proc(ptr: ^u32, waiters: u32) -> (waiters_woken_up: u32) ---
 
 // x86 Targets (i386, amd64)
-x86_cpuid  :: proc(ax, cx: u32) -> (eax, ebc, ecx, edx: u32) ---
+x86_cpuid  :: proc(ax, cx: u32) -> (eax, ebx, ecx, edx: u32) ---
 x86_xgetbv :: proc(cx: u32) -> (eax, edx: u32) ---
 
 
@@ -305,4 +305,4 @@ valgrind_client_request :: proc(default: uintptr, request: uintptr, a0, a1, a2,
 
 // Internal compiler use only
 
-__entry_point :: proc() ---
+__entry_point :: proc() ---

core/math/linalg/extended.odin

@@ -531,7 +531,7 @@ not_equal          :: proc{not_equal_single, not_equal_array}
 
 any :: proc(x: $A/[$N]bool) -> (out: bool) {
 	for e in x {
-		if x {
+		if e {
 			return true
 		}
 	}

core/mem/allocators.odin

@@ -153,7 +153,7 @@ scratch_allocator_proc :: proc(allocator_data: rawptr, mode: Allocator_Mode,
 	s := (^Scratch_Allocator)(allocator_data)
 
 	if s.data == nil {
-		DEFAULT_BACKING_SIZE :: 1<<22
+		DEFAULT_BACKING_SIZE :: 4 * Megabyte
 		if !(context.allocator.procedure != scratch_allocator_proc &&
 		     context.allocator.data != allocator_data) {
 			panic("cyclic initialization of the scratch allocator with itself")

core/mem/mem.odin

@@ -3,11 +3,11 @@ package mem
 import "core:runtime"
 import "core:intrinsics"
 
-Byte     :: 1
-Kilobyte :: 1024 * Byte
-Megabyte :: 1024 * Kilobyte
-Gigabyte :: 1024 * Megabyte
-Terabyte :: 1024 * Gigabyte
+Byte     :: runtime.Byte
+Kilobyte :: runtime.Kilobyte
+Megabyte :: runtime.Megabyte
+Gigabyte :: runtime.Gigabyte
+Terabyte :: runtime.Terabyte
 
 set :: proc "contextless" (data: rawptr, value: byte, len: int) -> rawptr {
 	return runtime.memset(data, i32(value), len)

core/mem/raw.odin

@@ -23,16 +23,3 @@ make_any :: proc "contextless" (data: rawptr, id: typeid) -> any {
 }
 
 raw_data :: builtin.raw_data
-
-
-Poly_Raw_Map_Entry :: struct($Key, $Value: typeid) {
-	hash:  uintptr,
-	next:  int,
-	key:   Key,
-	value: Value,	
-}
-
-Poly_Raw_Map :: struct($Key, $Value: typeid) {
-	hashes:  []int,
-	entries: [dynamic]Poly_Raw_Map_Entry(Key, Value),
-}

core/mem/virtual/arena.odin

@@ -1,6 +1,7 @@
 package mem_virtual
 
 import "core:mem"
+import "core:sync"
 
 Arena_Kind :: enum uint {
 	Growing = 0, // Chained memory blocks (singly linked list).
@@ -8,6 +9,13 @@ Arena_Kind :: enum uint {
 	Buffer  = 2, // Uses a fixed sized buffer.
 }
 
+/*
+	Arena is a generalized arena allocator that supports 3 different variants.
+
+	Growing: A linked list of `Memory_Block`s allocated with virtual memory.
+	Static: A single `Memory_Block` allocated with virtual memory.
+	Buffer: A single `Memory_Block` created from a user provided []byte.
+*/
 Arena :: struct {
 	kind:               Arena_Kind,
 	curr_block:         ^Memory_Block,
@@ -15,18 +23,21 @@ Arena :: struct {
 	total_reserved:     uint,
 	minimum_block_size: uint,
 	temp_count:         uint,
+	mutex:              sync.Mutex,
 }
 
 
 // 1 MiB should be enough to start with
-DEFAULT_ARENA_STATIC_COMMIT_SIZE         :: 1<<20
+DEFAULT_ARENA_STATIC_COMMIT_SIZE         :: mem.Megabyte
 DEFAULT_ARENA_GROWING_MINIMUM_BLOCK_SIZE :: DEFAULT_ARENA_STATIC_COMMIT_SIZE
 
 // 1 GiB on 64-bit systems, 128 MiB on 32-bit systems by default
-DEFAULT_ARENA_STATIC_RESERVE_SIZE :: 1<<30 when size_of(uintptr) == 8 else 1<<27
+DEFAULT_ARENA_STATIC_RESERVE_SIZE :: mem.Gigabyte when size_of(uintptr) == 8 else 128 * mem.Megabyte
 
 
 
+// Initialization of an `Arena` to be a `.Growing` variant.
+// A growing arena is a linked list of `Memory_Block`s allocated with virtual memory.
 @(require_results)
 arena_init_growing :: proc(arena: ^Arena, reserved: uint = DEFAULT_ARENA_GROWING_MINIMUM_BLOCK_SIZE) -> (err: Allocator_Error) {
 	arena.kind           = .Growing
@@ -37,6 +48,8 @@ arena_init_growing :: proc(arena: ^Arena, reserved: uint = DEFAULT_ARENA_GROWING
 }
 
 
+// Initialization of an `Arena` to be a `.Static` variant.
+// A static arena contains a single `Memory_Block` allocated with virtual memory.
 @(require_results)
 arena_init_static :: proc(arena: ^Arena, reserved: uint, commit_size: uint = DEFAULT_ARENA_STATIC_COMMIT_SIZE) -> (err: Allocator_Error) {
 	arena.kind           = .Static
@@ -46,6 +59,8 @@ arena_init_static :: proc(arena: ^Arena, reserved: uint, commit_size: uint = DEF
 	return
 }
 
+// Initialization of an `Arena` to be a `.Buffer` variant.
+// A buffer arena contains single `Memory_Block` created from a user provided []byte.
 @(require_results)
 arena_init_buffer :: proc(arena: ^Arena, buffer: []byte) -> (err: Allocator_Error) {
 	if len(buffer) < size_of(Memory_Block) {
@@ -69,6 +84,7 @@ arena_init_buffer :: proc(arena: ^Arena, buffer: []byte) -> (err: Allocator_Erro
 	return
 }
 
+// Allocates memory from the provided arena.
 @(require_results)
 arena_alloc :: proc(arena: ^Arena, size: uint, alignment: uint, loc := #caller_location) -> (data: []byte, err: Allocator_Error) {
 	assert(alignment & (alignment-1) == 0, "non-power of two alignment", loc)
@@ -78,6 +94,8 @@ arena_alloc :: proc(arena: ^Arena, size: uint, alignment: uint, loc := #caller_l
 		return nil, nil
 	}
 
+	sync.mutex_guard(&arena.mutex)
+
 	switch arena.kind {
 	case .Growing:
 		if arena.curr_block == nil || (safe_add(arena.curr_block.used, size) or_else 0) > arena.curr_block.reserved {
@@ -115,7 +133,10 @@ arena_alloc :: proc(arena: ^Arena, size: uint, alignment: uint, loc := #caller_l
 	return
 }
 
+// Resets the memory of a Static or Buffer arena to a specific `pos`ition (offset) and zeroes the previously used memory.
 arena_static_reset_to :: proc(arena: ^Arena, pos: uint, loc := #caller_location) -> bool {
+	sync.mutex_guard(&arena.mutex)
+
 	if arena.curr_block != nil {
 		assert(arena.kind != .Growing, "expected a non .Growing arena", loc)
 
@@ -134,48 +155,72 @@ arena_static_reset_to :: proc(arena: ^Arena, pos: uint, loc := #caller_location)
 	return false
 }
 
+// Frees the last memory block of a Growing Arena
 arena_growing_free_last_memory_block :: proc(arena: ^Arena, loc := #caller_location) {
 	if free_block := arena.curr_block; free_block != nil {
 		assert(arena.kind == .Growing, "expected a .Growing arena", loc)
+		arena.total_used -= free_block.used
+		arena.total_reserved -= free_block.reserved
+
 		arena.curr_block = free_block.prev
 		memory_block_dealloc(free_block)
 	}
 }
 
-arena_free_all :: proc(arena: ^Arena) {
+// Deallocates all but the first memory block of the arena and resets the allocator's usage to 0.
+arena_free_all :: proc(arena: ^Arena, loc := #caller_location) {
 	switch arena.kind {
 	case .Growing:
-		for arena.curr_block != nil {
-			arena_growing_free_last_memory_block(arena)
+		sync.mutex_guard(&arena.mutex)
+		// NOTE(bill): Free all but the first memory block (if it exists)
+		for arena.curr_block != nil && arena.curr_block.prev != nil {
+			arena_growing_free_last_memory_block(arena, loc)
 		}
-		arena.total_reserved = 0
+		// Zero the first block's memory
+		if arena.curr_block != nil {
+			mem.zero(arena.curr_block.base, int(arena.curr_block.used))
+			arena.curr_block.used = 0
+		}
+		arena.total_used = 0
 	case .Static, .Buffer:
 		arena_static_reset_to(arena, 0)
 	}
 	arena.total_used = 0
 }
 
-arena_destroy :: proc(arena: ^Arena) {
-	arena_free_all(arena)
-	if arena.kind != .Buffer {
+// Frees all of the memory allocated by the arena and zeros all of the values of an arena.
+// A buffer based arena does not `delete` the provided `[]byte` bufffer.
+arena_destroy :: proc(arena: ^Arena, loc := #caller_location) {
+	sync.mutex_guard(&arena.mutex)
+	switch arena.kind {
+	case .Growing:
+		for arena.curr_block != nil {
+			arena_growing_free_last_memory_block(arena, loc)
+		}
+	case .Static:
 		memory_block_dealloc(arena.curr_block)
+	case .Buffer:
+		// nothing
 	}
-	arena.curr_block = nil
+	arena.curr_block     = nil
 	arena.total_used     = 0
 	arena.total_reserved = 0
 	arena.temp_count     = 0
 }
 
+// Ability to bootstrap allocate a struct with an arena within the struct itself using the growing variant strategy.
 arena_growing_bootstrap_new :: proc{
 	arena_growing_bootstrap_new_by_offset,
 	arena_growing_bootstrap_new_by_name,
 }
 
+// Ability to bootstrap allocate a struct with an arena within the struct itself using the static variant strategy.
 arena_static_bootstrap_new :: proc{
 	arena_static_bootstrap_new_by_offset,
 	arena_static_bootstrap_new_by_name,
 }
 
+// Ability to bootstrap allocate a struct with an arena within the struct itself using the growing variant strategy.
 @(require_results)
 arena_growing_bootstrap_new_by_offset :: proc($T: typeid, offset_to_arena: uintptr, minimum_block_size: uint = DEFAULT_ARENA_GROWING_MINIMUM_BLOCK_SIZE) -> (ptr: ^T, err: Allocator_Error) {
 	bootstrap: Arena
@@ -191,11 +236,13 @@ arena_growing_bootstrap_new_by_offset :: proc($T: typeid, offset_to_arena: uintp
 	return
 }
 
+// Ability to bootstrap allocate a struct with an arena within the struct itself using the growing variant strategy.
 @(require_results)
 arena_growing_bootstrap_new_by_name :: proc($T: typeid, $field_name: string, minimum_block_size: uint = DEFAULT_ARENA_GROWING_MINIMUM_BLOCK_SIZE) -> (ptr: ^T, err: Allocator_Error) {
 	return arena_growing_bootstrap_new_by_offset(T, offset_of_by_string(T, field_name), minimum_block_size)
 }
 
+// Ability to bootstrap allocate a struct with an arena within the struct itself using the growing variant strategy.
 @(require_results)
 arena_static_bootstrap_new_by_offset :: proc($T: typeid, offset_to_arena: uintptr, reserved: uint) -> (ptr: ^T, err: Allocator_Error) {
 	bootstrap: Arena
@@ -211,17 +258,20 @@ arena_static_bootstrap_new_by_offset :: proc($T: typeid, offset_to_arena: uintpt
 	return
 }
 
+// Ability to bootstrap allocate a struct with an arena within the struct itself using the growing variant strategy.
 @(require_results)
 arena_static_bootstrap_new_by_name :: proc($T: typeid, $field_name: string, reserved: uint) -> (ptr: ^T, err: Allocator_Error) {
 	return arena_static_bootstrap_new_by_offset(T, offset_of_by_string(T, field_name), reserved)
 }
 
 
+// Create an `Allocator` from the provided `Arena`
 @(require_results)
 arena_allocator :: proc(arena: ^Arena) -> mem.Allocator {
 	return mem.Allocator{arena_allocator_proc, arena}
 }
 
+// The allocator procedured by an `Allocator` produced by `arena_allocator`
 arena_allocator_proc :: proc(allocator_data: rawptr, mode: mem.Allocator_Mode,
                              size, alignment: int,
                              old_memory: rawptr, old_size: int,
@@ -233,17 +283,17 @@ arena_allocator_proc :: proc(allocator_data: rawptr, mode: mem.Allocator_Mode,
 
 	switch mode {
 	case .Alloc, .Alloc_Non_Zeroed:
-		return arena_alloc(arena, size, alignment)
+		return arena_alloc(arena, size, alignment, location)
 	case .Free:
 		err = .Mode_Not_Implemented
 	case .Free_All:
-		arena_free_all(arena)
+		arena_free_all(arena, location)
 	case .Resize:
 		old_data := ([^]byte)(old_memory)
 
 		switch {
 		case old_data == nil:
-			return arena_alloc(arena, size, alignment)
+			return arena_alloc(arena, size, alignment, location)
 		case size == old_size:
 			// return old memory
 			data = old_data[:size]
@@ -257,7 +307,7 @@ arena_allocator_proc :: proc(allocator_data: rawptr, mode: mem.Allocator_Mode,
 			return
 		}
 
-		new_memory := arena_alloc(arena, size, alignment) or_return
+		new_memory := arena_alloc(arena, size, alignment, location) or_return
 		if new_memory == nil {
 			return
 		}
@@ -278,15 +328,20 @@ arena_allocator_proc :: proc(allocator_data: rawptr, mode: mem.Allocator_Mode,
 
 
 
+// An `Arena_Temp` is a way to produce temporary watermarks to reset a arena to a previous state.
+// All uses of an `Arena_Temp` must be handled by ending them with `arena_temp_end` or ignoring them with `arena_temp_ignore`.
 Arena_Temp :: struct {
 	arena: ^Arena,
 	block: ^Memory_Block,
 	used:  uint,
 }
 
+// Begins the section of temporary arena memory.
 @(require_results)
 arena_temp_begin :: proc(arena: ^Arena, loc := #caller_location) -> (temp: Arena_Temp) {
 	assert(arena != nil, "nil arena", loc)
+	sync.mutex_guard(&arena.mutex)
+
 	temp.arena = arena
 	temp.block = arena.curr_block
 	if arena.curr_block != nil {
@@ -296,36 +351,51 @@ arena_temp_begin :: proc(arena: ^Arena, loc := #caller_location) -> (temp: Arena
 	return
 }
 
+// Ends the section of temporary arena memory by resetting the memory to the stored position.
 arena_temp_end :: proc(temp: Arena_Temp, loc := #caller_location) {
 	assert(temp.arena != nil, "nil arena", loc)
 	arena := temp.arena
+	sync.mutex_guard(&arena.mutex)
 
-	memory_block_found := false
-	for block := arena.curr_block; block != nil; block = block.prev {
-		if block == temp.block {
-			memory_block_found = true
-			break
+	if temp.block != nil {
+		memory_block_found := false
+		for block := arena.curr_block; block != nil; block = block.prev {
+			if block == temp.block {
+				memory_block_found = true
+				break
+			}
+		}
+		if !memory_block_found {
+			assert(arena.curr_block == temp.block, "memory block stored within Arena_Temp not owned by Arena", loc)
 		}
-	}
-	if !memory_block_found {
-		assert(arena.curr_block == temp.block, "memory block stored within Arena_Temp not owned by Arena", loc)
-	}
 
-	for arena.curr_block != temp.block {
-		arena_growing_free_last_memory_block(arena)
-	}
+		for arena.curr_block != temp.block {
+			arena_growing_free_last_memory_block(arena)
+		}
 
-	if block := arena.curr_block; block != nil {
-		assert(block.used >= temp.used, "out of order use of arena_temp_end", loc)
-		amount_to_zero := min(block.used-temp.used, block.reserved-block.used)
-		mem.zero_slice(block.base[temp.used:][:amount_to_zero])
-		block.used = temp.used
+		if block := arena.curr_block; block != nil {
+			assert(block.used >= temp.used, "out of order use of arena_temp_end", loc)
+			amount_to_zero := min(block.used-temp.used, block.reserved-block.used)
+			mem.zero_slice(block.base[temp.used:][:amount_to_zero])
+			block.used = temp.used
+		}
 	}
 
 	assert(arena.temp_count > 0, "double-use of arena_temp_end", loc)
 	arena.temp_count -= 1
 }
 
+// Ignore the use of a `arena_temp_begin` entirely by __not__ resetting to the stored position.
+arena_temp_ignore :: proc(temp: Arena_Temp, loc := #caller_location) {
+	assert(temp.arena != nil, "nil arena", loc)
+	arena := temp.arena
+	sync.mutex_guard(&arena.mutex)
+
+	assert(arena.temp_count > 0, "double-use of arena_temp_end", loc)
+	arena.temp_count -= 1
+}
+
+// Asserts that all uses of `Arena_Temp` has been used by an `Arena`
 arena_check_temp :: proc(arena: ^Arena, loc := #caller_location) {
 	assert(arena.temp_count == 0, "Arena_Temp not been ended", loc)
 }

core/mem/virtual/virtual.odin

@@ -7,6 +7,7 @@ DEFAULT_PAGE_SIZE := uint(4096)
 
 Allocator_Error :: mem.Allocator_Error
 
+@(require_results)
 reserve :: proc "contextless" (size: uint) -> (data: []byte, err: Allocator_Error) {
 	return _reserve(size)
 }
@@ -15,6 +16,7 @@ commit :: proc "contextless" (data: rawptr, size: uint) -> Allocator_Error {
 	return _commit(data, size)
 }
 
+@(require_results)
 reserve_and_commit :: proc "contextless" (size: uint) -> (data: []byte, err: Allocator_Error) {
 	data = reserve(size) or_return
 	commit(raw_data(data), size) or_return
@@ -57,6 +59,7 @@ Memory_Block_Flag :: enum u32 {
 Memory_Block_Flags :: distinct bit_set[Memory_Block_Flag; u32]
 
 
+@(require_results)
 memory_block_alloc :: proc(committed, reserved: uint, flags: Memory_Block_Flags) -> (block: ^Memory_Block, err: Allocator_Error) {
 	align_formula :: proc "contextless" (size, align: uint) -> uint {
 		result := size + align-1
@@ -100,6 +103,7 @@ memory_block_alloc :: proc(committed, reserved: uint, flags: Memory_Block_Flags)
 	return &pmblock.block, nil
 }
 
+@(require_results)
 alloc_from_memory_block :: proc(block: ^Memory_Block, min_size, alignment: uint) -> (data: []byte, err: Allocator_Error) {
 	calc_alignment_offset :: proc "contextless" (block: ^Memory_Block, alignment: uintptr) -> uint {
 		alignment_offset := uint(0)
@@ -160,7 +164,7 @@ memory_block_dealloc :: proc(block_to_free: ^Memory_Block) {
 
 
 
-@(private)
+@(private, require_results)
 safe_add :: #force_inline proc "contextless" (x, y: uint) -> (uint, bool) {
 	z, did_overflow := intrinsics.overflow_add(x, y)
 	return z, !did_overflow

core/os/dir_darwin.odin

@@ -14,11 +14,12 @@ read_dir :: proc(fd: Handle, n: int, allocator := context.allocator) -> (fi: []F
 
 	dirpath: string
 	dirpath, err = absolute_path_from_handle(fd)
-
 	if err != ERROR_NONE {
 		return
 	}
 
+	defer delete(dirpath)
+
 	n := n
 	size := n
 	if n <= 0 {

core/os/dir_freebsd.odin

@@ -50,7 +50,7 @@ read_dir :: proc(fd: Handle, n: int, allocator := context.allocator) -> (fi: []F
 			continue
 		}
 
-		fullpath := make([]byte, len(dirpath)+1+len(filename))
+		fullpath := make([]byte, len(dirpath)+1+len(filename), context.temp_allocator)
 		copy(fullpath, dirpath)
 		copy(fullpath[len(dirpath):], "/")
 		copy(fullpath[len(dirpath)+1:], filename)

core/os/dir_linux.odin

@@ -2,6 +2,7 @@ package os
 
 import "core:strings"
 import "core:mem"
+import "core:runtime"
 
 read_dir :: proc(fd: Handle, n: int, allocator := context.allocator) -> (fi: []File_Info, err: Errno) {
 	dirp: Dir
@@ -51,6 +52,7 @@ read_dir :: proc(fd: Handle, n: int, allocator := context.allocator) -> (fi: []F
 			continue
 		}
 
+		runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD(ignore = context.temp_allocator == allocator)
 		fullpath := strings.join( []string{ dirpath, filename }, "/", context.temp_allocator)
 		defer delete(fullpath, context.temp_allocator)
 

core/os/dir_windows.odin

@@ -2,6 +2,7 @@ package os
 
 import win32 "core:sys/windows"
 import "core:strings"
+import "core:runtime"
 
 read_dir :: proc(fd: Handle, n: int, allocator := context.allocator) -> (fi: []File_Info, err: Errno) {
 	find_data_to_file_info :: proc(base_path: string, d: ^win32.WIN32_FIND_DATAW) -> (fi: File_Info) {
@@ -65,13 +66,16 @@ read_dir :: proc(fd: Handle, n: int, allocator := context.allocator) -> (fi: []F
 		n = -1
 		size = 100
 	}
-	dfi := make([dynamic]File_Info, 0, size)
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD(ignore = context.temp_allocator == allocator)
 
 	wpath: []u16
-	wpath, err = cleanpath_from_handle_u16(fd)
+	wpath, err = cleanpath_from_handle_u16(fd, context.temp_allocator)
 	if len(wpath) == 0 || err != ERROR_NONE {
 		return
 	}
+
+	dfi := make([dynamic]File_Info, 0, size)
+
 	wpath_search := make([]u16, len(wpath)+3, context.temp_allocator)
 	copy(wpath_search, wpath)
 	wpath_search[len(wpath)+0] = '\\'

core/os/env_windows.odin

@@ -1,6 +1,7 @@
 package os
 
 import win32 "core:sys/windows"
+import "core:runtime"
 
 // lookup_env gets the value of the environment variable named by the key
 // If the variable is found in the environment the value (which can be empty) is returned and the boolean is true
@@ -18,6 +19,8 @@ lookup_env :: proc(key: string, allocator := context.allocator) -> (value: strin
 			return "", false
 		}
 	}
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD(ignore = context.temp_allocator == allocator)
+
 	b := make([dynamic]u16, n, context.temp_allocator)
 	n = win32.GetEnvironmentVariableW(wkey, raw_data(b), u32(len(b)))
 	if n == 0 {
@@ -87,6 +90,7 @@ environ :: proc(allocator := context.allocator) -> []string {
 
 // clear_env deletes all environment variables
 clear_env :: proc() {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	envs := environ(context.temp_allocator)
 	for env in envs {
 		for j in 1..<len(env) {

core/os/file_windows.odin

@@ -2,6 +2,7 @@ package os
 
 import win32 "core:sys/windows"
 import "core:intrinsics"
+import "core:runtime"
 import "core:unicode/utf16"
 
 is_path_separator :: proc(c: byte) -> bool {
@@ -327,6 +328,7 @@ get_std_handle :: proc "contextless" (h: uint) -> Handle {
 
 
 exists :: proc(path: string) -> bool {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	wpath := win32.utf8_to_wstring(path, context.temp_allocator)
 	attribs := win32.GetFileAttributesW(wpath)
 
@@ -334,6 +336,7 @@ exists :: proc(path: string) -> bool {
 }
 
 is_file :: proc(path: string) -> bool {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	wpath := win32.utf8_to_wstring(path, context.temp_allocator)
 	attribs := win32.GetFileAttributesW(wpath)
 
@@ -344,6 +347,7 @@ is_file :: proc(path: string) -> bool {
 }
 
 is_dir :: proc(path: string) -> bool {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	wpath := win32.utf8_to_wstring(path, context.temp_allocator)
 	attribs := win32.GetFileAttributesW(wpath)
 
@@ -359,6 +363,8 @@ is_dir :: proc(path: string) -> bool {
 get_current_directory :: proc(allocator := context.allocator) -> string {
 	win32.AcquireSRWLockExclusive(&cwd_lock)
 
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD(ignore = context.temp_allocator == allocator)
+
 	sz_utf16 := win32.GetCurrentDirectoryW(0, nil)
 	dir_buf_wstr := make([]u16, sz_utf16, context.temp_allocator) // the first time, it _includes_ the NUL.
 
@@ -387,6 +393,7 @@ set_current_directory :: proc(path: string) -> (err: Errno) {
 
 
 change_directory :: proc(path: string) -> (err: Errno) {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	wpath := win32.utf8_to_wstring(path, context.temp_allocator)
 
 	if !win32.SetCurrentDirectoryW(wpath) {
@@ -396,6 +403,7 @@ change_directory :: proc(path: string) -> (err: Errno) {
 }
 
 make_directory :: proc(path: string, mode: u32 = 0) -> (err: Errno) {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	// Mode is unused on Windows, but is needed on *nix
 	wpath := win32.utf8_to_wstring(path, context.temp_allocator)
 
@@ -407,6 +415,7 @@ make_directory :: proc(path: string, mode: u32 = 0) -> (err: Errno) {
 
 
 remove_directory :: proc(path: string) -> (err: Errno) {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	wpath := win32.utf8_to_wstring(path, context.temp_allocator)
 
 	if !win32.RemoveDirectoryW(wpath) {
@@ -479,12 +488,14 @@ fix_long_path :: proc(path: string) -> string {
 
 
 link :: proc(old_name, new_name: string) -> (err: Errno) {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	n := win32.utf8_to_wstring(fix_long_path(new_name))
 	o := win32.utf8_to_wstring(fix_long_path(old_name))
 	return Errno(win32.CreateHardLinkW(n, o, nil))
 }
 
 unlink :: proc(path: string) -> (err: Errno) {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	wpath := win32.utf8_to_wstring(path, context.temp_allocator)
 
 	if !win32.DeleteFileW(wpath) {
@@ -496,6 +507,7 @@ unlink :: proc(path: string) -> (err: Errno) {
 
 
 rename :: proc(old_path, new_path: string) -> (err: Errno) {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	from := win32.utf8_to_wstring(old_path, context.temp_allocator)
 	to := win32.utf8_to_wstring(new_path, context.temp_allocator)
 

core/os/os.odin

@@ -93,7 +93,7 @@ file_size_from_path :: proc(path: string) -> i64 {
 	return length
 }
 
-read_entire_file_from_filename :: proc(name: string, allocator := context.allocator) -> (data: []byte, success: bool) {
+read_entire_file_from_filename :: proc(name: string, allocator := context.allocator, loc := #caller_location) -> (data: []byte, success: bool) {
 	context.allocator = allocator
 
 	fd, err := open(name, O_RDONLY, 0)
@@ -102,10 +102,10 @@ read_entire_file_from_filename :: proc(name: string, allocator := context.alloca
 	}
 	defer close(fd)
 
-	return read_entire_file_from_handle(fd, allocator)
+	return read_entire_file_from_handle(fd, allocator, loc)
 }
 
-read_entire_file_from_handle :: proc(fd: Handle, allocator := context.allocator) -> (data: []byte, success: bool) {
+read_entire_file_from_handle :: proc(fd: Handle, allocator := context.allocator, loc := #caller_location) -> (data: []byte, success: bool) {
 	context.allocator = allocator
 
 	length: i64
@@ -118,7 +118,7 @@ read_entire_file_from_handle :: proc(fd: Handle, allocator := context.allocator)
 		return nil, true
 	}
 
-	data = make([]byte, int(length), allocator)
+	data = make([]byte, int(length), allocator, loc)
 	if data == nil {
 	    return nil, false
 	}
@@ -216,7 +216,7 @@ heap_allocator_proc :: proc(allocator_data: rawptr, mode: mem.Allocator_Mode,
 		}
 
 		new_memory = aligned_alloc(new_size, new_alignment, p) or_return
-		
+
 		// NOTE: heap_resize does not zero the new memory, so we do it
 		if new_size > old_size {
 			new_region := mem.raw_data(new_memory[old_size:])
@@ -261,3 +261,7 @@ heap_allocator :: proc() -> mem.Allocator {
 		data = nil,
 	}
 }
+
+processor_core_count :: proc() -> int {
+	return _processor_core_count()
+}

core/os/os2/env.odin

@@ -1,10 +1,12 @@
 package os2
 
+import "core:runtime"
+
 // get_env retrieves the value of the environment variable named by the key
 // It returns the value, which will be empty if the variable is not present
 // To distinguish between an empty value and an unset value, use lookup_env
 // NOTE: the value will be allocated with the supplied allocator
-get_env :: proc(key: string, allocator := context.allocator) -> string {
+get_env :: proc(key: string, allocator: runtime.Allocator) -> string {
 	value, _ := lookup_env(key, allocator)
 	return value
 }
@@ -13,7 +15,7 @@ get_env :: proc(key: string, allocator := context.allocator) -> string {
 // If the variable is found in the environment the value (which can be empty) is returned and the boolean is true
 // Otherwise the returned value will be empty and the boolean will be false
 // NOTE: the value will be allocated with the supplied allocator
-lookup_env :: proc(key: string, allocator := context.allocator) -> (value: string, found: bool) {
+lookup_env :: proc(key: string, allocator: runtime.Allocator) -> (value: string, found: bool) {
 	return _lookup_env(key, allocator)
 }
 
@@ -36,7 +38,7 @@ clear_env :: proc() {
 
 // environ returns a copy of strings representing the environment, in the form "key=value"
 // NOTE: the slice of strings and the strings with be allocated using the supplied allocator
-environ :: proc(allocator := context.allocator) -> []string {
+environ :: proc(allocator: runtime.Allocator) -> []string {
 	return _environ(allocator)
 }
 

core/os/os2/env_linux.odin

@@ -1,7 +1,9 @@
 //+private
 package os2
 
-_get_env :: proc(key: string, allocator := context.allocator) -> (value: string, found: bool) {
+import "core:runtime"
+
+_get_env :: proc(key: string, allocator: runtime.Allocator) -> (value: string, found: bool) {
 	//TODO
 	return
 }
@@ -20,7 +22,7 @@ _clear_env :: proc() {
 	//TODO
 }
 
-_environ :: proc(allocator := context.allocator) -> []string {
+_environ :: proc(allocator: runtime.Allocator) -> []string {
 	//TODO
 	return nil
 }

core/os/os2/env_windows.odin

@@ -65,7 +65,19 @@ _environ :: proc(allocator: runtime.Allocator) -> []string {
 	}
 	defer win32.FreeEnvironmentStringsW(envs)
 
-	r := make([dynamic]string, 0, 50, allocator)
+	n := 0
+	for from, i, p := 0, 0, envs; true; i += 1 {
+		c := ([^]u16)(p)[i]
+		if c == 0 {
+			if i <= from {
+				break
+			}
+			n += 1
+			from = i + 1
+		}
+	}
+
+	r := make([dynamic]string, 0, n, allocator)
 	for from, i, p := 0, 0, envs; true; i += 1 {
 		c := ([^]u16)(p)[i]
 		if c == 0 {

core/os/os2/file_linux.odin

@@ -39,10 +39,8 @@ _file_allocator :: proc() -> runtime.Allocator {
 }
 
 _open :: proc(name: string, flags: File_Flags, perm: File_Mode) -> (^File, Error) {
-	name_cstr, allocated := _name_to_cstring(name)
-	defer if allocated {
-		delete(name_cstr)
-	}
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
+	name_cstr := _name_to_cstring(name)
 
 	flags_i: int
 	switch flags & O_RDONLY|O_WRONLY|O_RDWR {
@@ -254,7 +252,7 @@ _symlink :: proc(old_name, new_name: string) -> Error {
 	return _ok_or_error(unix.sys_symlink(old_name_cstr, new_name_cstr))
 }
 
-_read_link_cstr :: proc(name_cstr: cstring, allocator := context.allocator) -> (string, Error) {
+_read_link_cstr :: proc(name_cstr: cstring, allocator: runtime.Allocator) -> (string, Error) {
 	bufsz : uint = 256
 	buf := make([]byte, bufsz, allocator)
 	for {
@@ -272,7 +270,7 @@ _read_link_cstr :: proc(name_cstr: cstring, allocator := context.allocator) -> (
 	}
 }
 
-_read_link :: proc(name: string, allocator := context.allocator) -> (string, Error) {
+_read_link :: proc(name: string, allocator: runtime.Allocator) -> (string, Error) {
 	name_cstr, allocated := _name_to_cstring(name)
 	defer if allocated {
 		delete(name_cstr)
@@ -411,12 +409,7 @@ _is_dir_fd :: proc(fd: int) -> bool {
 // defined as 512, however, it is well known that paths can exceed that limit.
 // So, in theory you could have a path larger than the entire temp_allocator's
 // buffer. Therefor, any large paths will use context.allocator.
-_name_to_cstring :: proc(name: string) -> (cname: cstring, allocated: bool) {
-	if len(name) > _CSTRING_NAME_HEAP_THRESHOLD {
-		cname = strings.clone_to_cstring(name)
-		allocated = true
-		return
-	}
-	cname = strings.clone_to_cstring(name, context.temp_allocator)
-	return
+@(private="file")
+_temp_name_to_cstring :: proc(name: string) -> (cname: cstring) {
+	return strings.clone_to_cstring(name, context.temp_allocator)
 }

core/os/os2/file_util.odin

@@ -1,6 +1,7 @@
 package os2
 
 import "core:mem"
+import "core:runtime"
 import "core:strconv"
 import "core:unicode/utf8"
 
@@ -74,7 +75,7 @@ read_ptr :: proc(f: ^File, data: rawptr, len: int) -> (n: int, err: Error) {
 
 
 
-read_entire_file :: proc(name: string, allocator := context.allocator) -> (data: []byte, err: Error) {
+read_entire_file :: proc(name: string, allocator: runtime.Allocator) -> (data: []byte, err: Error) {
 	f, ferr := open(name)
 	if ferr != nil {
 		return nil, ferr

core/os/os2/path_linux.odin

@@ -229,7 +229,7 @@ _setwd :: proc(dir: string) -> Error {
 	return _ok_or_error(unix.sys_chdir(dir_cstr))
 }
 
-_get_full_path :: proc(fd: int, allocator := context.allocator) -> string {
+_get_full_path :: proc(fd: int, allocator: runtime.Allocator) -> string {
 	PROC_FD_PATH :: "/proc/self/fd/"
 
 	buf: [32]u8

core/os/os2/stat_linux.odin

@@ -83,7 +83,7 @@ _Stat :: struct {
 }
 
 
-_fstat :: proc(f: ^File, allocator := context.allocator) -> (File_Info, Error) {
+_fstat :: proc(f: ^File, allocator: runtime.Allocator) -> (File_Info, Error) {
 	return _fstat_internal(f.impl.fd, allocator)
 }
 
@@ -111,7 +111,7 @@ _fstat_internal :: proc(fd: int, allocator: runtime.Allocator) -> (File_Info, Er
 }
 
 // NOTE: _stat and _lstat are using _fstat to avoid a race condition when populating fullpath
-_stat :: proc(name: string, allocator := context.allocator) -> (File_Info, Error) {
+_stat :: proc(name: string, allocator: runtime.Allocator) -> (File_Info, Error) {
 	name_cstr, allocated := _name_to_cstring(name)
 	defer if allocated {
 		delete(name_cstr)
@@ -125,7 +125,7 @@ _stat :: proc(name: string, allocator := context.allocator) -> (File_Info, Error
 	return _fstat_internal(fd, allocator)
 }
 
-_lstat :: proc(name: string, allocator := context.allocator) -> (File_Info, Error) {
+_lstat :: proc(name: string, allocator: runtime.Allocator) -> (File_Info, Error) {
 	name_cstr, allocated := _name_to_cstring(name)
 	defer if allocated {
 		delete(name_cstr)

core/os/os_darwin.odin

@@ -277,8 +277,10 @@ foreign libc {
 
 	@(link_name="open")             _unix_open          :: proc(path: cstring, flags: i32, mode: u16) -> Handle ---
 	@(link_name="close")            _unix_close         :: proc(handle: Handle) -> c.int ---
-	@(link_name="read")             _unix_read          :: proc(handle: Handle, buffer: rawptr, count: int) -> int ---
-	@(link_name="write")            _unix_write         :: proc(handle: Handle, buffer: rawptr, count: int) -> int ---
+	@(link_name="read")             _unix_read          :: proc(handle: Handle, buffer: rawptr, count: c.size_t) -> int ---
+	@(link_name="write")            _unix_write         :: proc(handle: Handle, buffer: rawptr, count: c.size_t) -> int ---
+	@(link_name="pread")            _unix_pread         :: proc(handle: Handle, buffer: rawptr, count: c.size_t, offset: i64) -> int ---
+	@(link_name="pwrite")           _unix_pwrite        :: proc(handle: Handle, buffer: rawptr, count: c.size_t, offset: i64) -> int ---
 	@(link_name="lseek")            _unix_lseek         :: proc(fs: Handle, offset: int, whence: int) -> int ---
 	@(link_name="gettid")           _unix_gettid        :: proc() -> u64 ---
 	@(link_name="getpagesize")      _unix_getpagesize   :: proc() -> i32 ---
@@ -314,6 +316,7 @@ foreign libc {
 	@(link_name="realpath") _unix_realpath :: proc(path: cstring, resolved_path: rawptr) -> rawptr ---
 
 	@(link_name="strerror") _darwin_string_error :: proc(num : c.int) -> cstring ---
+	@(link_name="sysctlbyname") _sysctlbyname    :: proc(path: cstring, oldp: rawptr, oldlenp: rawptr, newp: rawptr, newlen: int) -> c.int ---
 
 	@(link_name="exit")    _unix_exit :: proc(status: c.int) -> ! ---
 }
@@ -333,7 +336,7 @@ foreign dl {
 	@(link_name="dlerror") _unix_dlerror :: proc() -> cstring ---
 }
 
-get_last_error :: proc() -> int {
+get_last_error :: proc "contextless" () -> int {
 	return __error()^
 }
 
@@ -342,21 +345,34 @@ get_last_error_string :: proc() -> string {
 }
 
 open :: proc(path: string, flags: int = O_RDWR, mode: int = 0) -> (Handle, Errno) {
+	isDir := is_dir_path(path)
+	flags := flags
+	if isDir {
+		/*
+			@INFO(Platin): To make it impossible to use the wrong flag for dir's 
+			               as you can't write to a dir only read which makes it fail to open
+		*/
+		flags = O_RDONLY
+	}
+
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	cstr := strings.clone_to_cstring(path, context.temp_allocator)
 	handle := _unix_open(cstr, i32(flags), u16(mode))
 	if handle == -1 {
-		return INVALID_HANDLE, 1
+		return INVALID_HANDLE, cast(Errno)get_last_error()
 	}
 
-when  ODIN_OS == .Darwin && ODIN_ARCH == .arm64 {
-	if mode != 0 {
+	/*
+		@INFO(Platin): this is only done because O_CREATE for some reason fails to apply mode
+		               should not happen if the handle is a directory
+	*/
+	if mode != 0 && !isDir {
 		err := fchmod(handle, cast(u16)mode)
 		if err != 0 {
 			_unix_close(handle)
-			return INVALID_HANDLE, 1
+			return INVALID_HANDLE, cast(Errno)err
 		}
 	}
-}
 
 	return handle, 0
 }
@@ -372,45 +388,51 @@ close :: proc(fd: Handle) -> bool {
 @(private)
 MAX_RW :: 0x7fffffff // The limit on Darwin is max(i32), trying to read/write more than that fails.
 
-write :: proc(fd: Handle, data: []u8) -> (int, Errno) {
-	assert(fd != -1)
-
-	bytes_total := len(data)
-	bytes_written_total := 0
-
-	for bytes_written_total < bytes_total {
-		bytes_to_write := min(bytes_total - bytes_written_total, MAX_RW)
-		slice := data[bytes_written_total:bytes_written_total + bytes_to_write]
-		bytes_written := _unix_write(fd, raw_data(slice), bytes_to_write)
-		if bytes_written == -1 {
-			return bytes_written_total, 1
-		}
-		bytes_written_total += bytes_written
+write :: proc(fd: Handle, data: []byte) -> (int, Errno) {
+	if len(data) == 0 {
+		return 0, ERROR_NONE
 	}
 
-	return bytes_written_total, 0
+	bytes_written := _unix_write(fd, raw_data(data), c.size_t(len(data)))
+	if bytes_written < 0 {
+		return -1, Errno(get_last_error())
+	}
+	return bytes_written, ERROR_NONE
 }
 
 read :: proc(fd: Handle, data: []u8) -> (int, Errno) {
-	assert(fd != -1)
-
-	bytes_total := len(data)
-	bytes_read_total := 0
-
-	for bytes_read_total < bytes_total {
-		bytes_to_read := min(bytes_total - bytes_read_total, MAX_RW)
-		slice := data[bytes_read_total:bytes_read_total + bytes_to_read]
-		bytes_read := _unix_read(fd, raw_data(slice), bytes_to_read)
-		if bytes_read == -1 {
-			return bytes_read_total, 1
-		}
-		if bytes_read == 0 {
-			break
-		}
-		bytes_read_total += bytes_read
+	if len(data) == 0 {
+		return 0, ERROR_NONE
 	}
 
-	return bytes_read_total, 0
+	bytes_read := _unix_read(fd, raw_data(data), c.size_t(len(data)))
+	if bytes_read < 0 {
+		return -1, Errno(get_last_error())
+	}
+	return bytes_read, ERROR_NONE
+}
+read_at :: proc(fd: Handle, data: []byte, offset: i64) -> (int, Errno) {
+	if len(data) == 0 {
+		return 0, ERROR_NONE
+	}
+
+	bytes_read := _unix_pread(fd, raw_data(data), c.size_t(len(data)), offset)
+	if bytes_read < 0 {
+		return -1, Errno(get_last_error())
+	}
+	return bytes_read, ERROR_NONE
+}
+
+write_at :: proc(fd: Handle, data: []byte, offset: i64) -> (int, Errno) {
+	if len(data) == 0 {
+		return 0, ERROR_NONE
+	}
+
+	bytes_written := _unix_pwrite(fd, raw_data(data), c.size_t(len(data)), offset)
+	if bytes_written < 0 {
+		return -1, Errno(get_last_error())
+	}
+	return bytes_written, ERROR_NONE
 }
 
 seek :: proc(fd: Handle, offset: i64, whence: int) -> (i64, Errno) {
@@ -495,24 +517,28 @@ is_file :: proc {is_file_path, is_file_handle}
 is_dir :: proc {is_dir_path, is_dir_handle}
 
 exists :: proc(path: string) -> bool {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	cpath := strings.clone_to_cstring(path, context.temp_allocator)
 	res := _unix_access(cpath, O_RDONLY)
 	return res == 0
 }
 
 rename :: proc(old: string, new: string) -> bool {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	old_cstr := strings.clone_to_cstring(old, context.temp_allocator)
 	new_cstr := strings.clone_to_cstring(new, context.temp_allocator)
 	return _unix_rename(old_cstr, new_cstr) != -1
 }
 
 remove :: proc(path: string) -> bool {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	path_cstr := strings.clone_to_cstring(path, context.temp_allocator)
 	return _unix_remove(path_cstr) != -1
 }
 
 @private
 _stat :: proc(path: string) -> (OS_Stat, Errno) {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	cstr := strings.clone_to_cstring(path, context.temp_allocator)
 
 	s: OS_Stat
@@ -525,6 +551,7 @@ _stat :: proc(path: string) -> (OS_Stat, Errno) {
 
 @private
 _lstat :: proc(path: string) -> (OS_Stat, Errno) {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	cstr := strings.clone_to_cstring(path, context.temp_allocator)
 
 	s: OS_Stat
@@ -590,6 +617,7 @@ _readdir :: proc(dirp: Dir) -> (entry: Dirent, err: Errno, end_of_stream: bool)
 
 @private
 _readlink :: proc(path: string) -> (string, Errno) {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD(ignore = context.temp_allocator == context.allocator)
 	path_cstr := strings.clone_to_cstring(path, context.temp_allocator)
 
 	bufsz : uint = 256
@@ -627,6 +655,7 @@ absolute_path_from_relative :: proc(rel: string) -> (path: string, err: Errno) {
 		rel = "."
 	}
 
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD(ignore = context.temp_allocator == context.allocator)
 	rel_cstr := strings.clone_to_cstring(rel, context.temp_allocator)
 
 	path_ptr := _unix_realpath(rel_cstr, nil)
@@ -642,6 +671,7 @@ absolute_path_from_relative :: proc(rel: string) -> (path: string, err: Errno) {
 }
 
 access :: proc(path: string, mask: int) -> bool {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	cstr := strings.clone_to_cstring(path, context.temp_allocator)
 	return _unix_access(cstr, mask) == 0
 }
@@ -666,6 +696,7 @@ heap_free :: proc(ptr: rawptr) {
 }
 
 lookup_env :: proc(key: string, allocator := context.allocator) -> (value: string, found: bool) {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD(ignore = context.temp_allocator == allocator)
 	path_str := strings.clone_to_cstring(key, context.temp_allocator)
 	cstr := _unix_getenv(path_str)
 	if cstr == nil {
@@ -688,6 +719,7 @@ get_current_directory :: proc() -> string {
 			return string(cwd)
 		}
 		if Errno(get_last_error()) != ERANGE {
+			delete(buf)
 			return ""
 		}
 		resize(&buf, len(buf)+page_size)
@@ -696,6 +728,7 @@ get_current_directory :: proc() -> string {
 }
 
 set_current_directory :: proc(path: string) -> (err: Errno) {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	cstr := strings.clone_to_cstring(path, context.temp_allocator)
 	res := _unix_chdir(cstr)
 	if res == -1 {
@@ -705,6 +738,7 @@ set_current_directory :: proc(path: string) -> (err: Errno) {
 }
 
 make_directory :: proc(path: string, mode: u16 = 0o775) -> Errno {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	path_cstr := strings.clone_to_cstring(path, context.temp_allocator)
 	res := _unix_mkdir(path_cstr, mode)
 	if res == -1 {
@@ -729,12 +763,14 @@ current_thread_id :: proc "contextless" () -> int {
 }
 
 dlopen :: proc(filename: string, flags: int) -> rawptr {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	cstr := strings.clone_to_cstring(filename, context.temp_allocator)
 	handle := _unix_dlopen(cstr, flags)
 	return handle
 }
 dlsym :: proc(handle: rawptr, symbol: string) -> rawptr {
 	assert(handle != nil)
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	cstr := strings.clone_to_cstring(symbol, context.temp_allocator)
 	proc_handle := _unix_dlsym(handle, cstr)
 	return proc_handle
@@ -759,6 +795,18 @@ get_page_size :: proc() -> int {
 	return page_size
 }
 
+@(private)
+_processor_core_count :: proc() -> int {
+	count : int = 0
+	count_size := size_of(count)
+	if _sysctlbyname("hw.logicalcpu", &count, &count_size, nil, 0) == 0 {
+		if count > 0 {
+			return count
+		}
+	}
+
+	return 1
+}
 
 _alloc_command_line_arguments :: proc() -> []string {
 	res := make([]string, len(runtime.args__))

core/os/os_freebsd.odin

@@ -287,6 +287,7 @@ foreign libc {
 	
 	@(link_name="getenv")           _unix_getenv        :: proc(cstring) -> cstring ---
 	@(link_name="realpath")         _unix_realpath      :: proc(path: cstring, resolved_path: rawptr) -> rawptr ---
+	@(link_name="sysctlbyname")     _sysctlbyname       :: proc(path: cstring, oldp: rawptr, oldlenp: rawptr, newp: rawptr, newlen: int) -> c.int ---
 
 	@(link_name="exit")             _unix_exit          :: proc(status: c.int) -> ! ---
 }
@@ -303,11 +304,12 @@ is_path_separator :: proc(r: rune) -> bool {
 	return r == '/'
 }
 
-get_last_error :: proc() -> int {
+get_last_error :: proc "contextless" () -> int {
 	return __errno_location()^
 }
 
 open :: proc(path: string, flags: int = O_RDONLY, mode: int = 0) -> (Handle, Errno) {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	cstr := strings.clone_to_cstring(path, context.temp_allocator)
 	handle := _unix_open(cstr, c.int(flags), c.int(mode))
 	if handle == -1 {
@@ -360,6 +362,7 @@ file_size :: proc(fd: Handle) -> (i64, Errno) {
 }
 
 rename :: proc(old_path, new_path: string) -> Errno {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	old_path_cstr := strings.clone_to_cstring(old_path, context.temp_allocator)
 	new_path_cstr := strings.clone_to_cstring(new_path, context.temp_allocator)
 	res := _unix_rename(old_path_cstr, new_path_cstr)
@@ -370,6 +373,7 @@ rename :: proc(old_path, new_path: string) -> Errno {
 }
 
 remove :: proc(path: string) -> Errno {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	path_cstr := strings.clone_to_cstring(path, context.temp_allocator)
 	res := _unix_unlink(path_cstr)
 	if res == -1 {
@@ -379,6 +383,7 @@ remove :: proc(path: string) -> Errno {
 }
 
 make_directory :: proc(path: string, mode: mode_t = 0o775) -> Errno {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	path_cstr := strings.clone_to_cstring(path, context.temp_allocator)
 	res := _unix_mkdir(path_cstr, mode)
 	if res == -1 {
@@ -388,6 +393,7 @@ make_directory :: proc(path: string, mode: mode_t = 0o775) -> Errno {
 }
 
 remove_directory :: proc(path: string) -> Errno {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	path_cstr := strings.clone_to_cstring(path, context.temp_allocator)
 	res := _unix_rmdir(path_cstr)
 	if res == -1 {
@@ -473,6 +479,7 @@ last_write_time_by_name :: proc(name: string) -> (File_Time, Errno) {
 
 @private
 _stat :: proc(path: string) -> (OS_Stat, Errno) {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	cstr := strings.clone_to_cstring(path, context.temp_allocator)
 	s: OS_Stat = ---
 	result := _unix_lstat(cstr, &s)
@@ -484,6 +491,7 @@ _stat :: proc(path: string) -> (OS_Stat, Errno) {
 
 @private
 _lstat :: proc(path: string) -> (OS_Stat, Errno) {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	cstr := strings.clone_to_cstring(path, context.temp_allocator)
 	
 	// deliberately uninitialized
@@ -549,6 +557,8 @@ _readdir :: proc(dirp: Dir) -> (entry: Dirent, err: Errno, end_of_stream: bool)
 
 @private
 _readlink :: proc(path: string) -> (string, Errno) {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD(ignore = context.temp_allocator == context.allocator)
+
 	path_cstr := strings.clone_to_cstring(path, context.temp_allocator)
 
 	bufsz : uint = MAX_PATH
@@ -566,7 +576,8 @@ _readlink :: proc(path: string) -> (string, Errno) {
 			return strings.string_from_ptr(&buf[0], rc), ERROR_NONE
 		}	
 	}
-	unreachable()
+
+	return "", Errno{}
 }
 
 // XXX FreeBSD
@@ -579,6 +590,7 @@ absolute_path_from_relative :: proc(rel: string) -> (path: string, err: Errno) {
 	if rel == "" {
 		rel = "."
 	}
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD(ignore = context.temp_allocator == context.allocator)
 
 	rel_cstr := strings.clone_to_cstring(rel, context.temp_allocator)
 
@@ -595,6 +607,8 @@ absolute_path_from_relative :: proc(rel: string) -> (path: string, err: Errno) {
 }
 
 access :: proc(path: string, mask: int) -> (bool, Errno) {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
+
 	cstr := strings.clone_to_cstring(path, context.temp_allocator)
 	result := _unix_access(cstr, c.int(mask))
 	if result == -1 {
@@ -625,6 +639,8 @@ heap_free :: proc(ptr: rawptr) {
 }
 
 lookup_env :: proc(key: string, allocator := context.allocator) -> (value: string, found: bool) {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD(ignore = context.temp_allocator == allocator)
+
 	path_str := strings.clone_to_cstring(key, context.temp_allocator)
 	cstr := _unix_getenv(path_str)
 	if cstr == nil {
@@ -650,6 +666,7 @@ get_current_directory :: proc() -> string {
 			return string(cwd)
 		}
 		if Errno(get_last_error()) != ERANGE {
+			delete(buf)
 			return ""
 		}
 		resize(&buf, len(buf)+page_size)
@@ -658,6 +675,7 @@ get_current_directory :: proc() -> string {
 }
 
 set_current_directory :: proc(path: string) -> (err: Errno) {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	cstr := strings.clone_to_cstring(path, context.temp_allocator)
 	res := _unix_chdir(cstr)
 	if res == -1 do return Errno(get_last_error())
@@ -674,12 +692,14 @@ current_thread_id :: proc "contextless" () -> int {
 }
 
 dlopen :: proc(filename: string, flags: int) -> rawptr {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	cstr := strings.clone_to_cstring(filename, context.temp_allocator)
 	handle := _unix_dlopen(cstr, c.int(flags))
 	return handle
 }
 dlsym :: proc(handle: rawptr, symbol: string) -> rawptr {
 	assert(handle != nil)
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	cstr := strings.clone_to_cstring(symbol, context.temp_allocator)
 	proc_handle := _unix_dlsym(handle, cstr)
 	return proc_handle
@@ -702,6 +722,19 @@ get_page_size :: proc() -> int {
 	return page_size
 }
 
+@(private)
+_processor_core_count :: proc() -> int {
+	count : int = 0
+	count_size := size_of(count)
+	if _sysctlbyname("hw.logicalcpu", &count, &count_size, nil, 0) == 0 {
+		if count > 0 {
+			return count
+		}
+	}
+
+	return 1
+}
+
 
 _alloc_command_line_arguments :: proc() -> []string {
 	res := make([]string, len(runtime.args__))

core/os/os_linux.odin

@@ -236,13 +236,13 @@ S_IRUSR :: 0o0400 // R for owner
 S_IWUSR :: 0o0200 // W for owner
 S_IXUSR :: 0o0100 // X for owner
 
-	// Read, write, execute/search by group
+// Read, write, execute/search by group
 S_IRWXG :: 0o0070 // RWX mask for group
 S_IRGRP :: 0o0040 // R for group
 S_IWGRP :: 0o0020 // W for group
 S_IXGRP :: 0o0010 // X for group
 
-	// Read, write, execute/search by others
+// Read, write, execute/search by others
 S_IRWXO :: 0o0007 // RWX mask for other
 S_IROTH :: 0o0004 // R for other
 S_IWOTH :: 0o0002 // W for other
@@ -270,140 +270,11 @@ AT_FDCWD            :: ~uintptr(99)	/* -100 */
 AT_REMOVEDIR        :: uintptr(0x200)
 AT_SYMLINK_NOFOLLOW :: uintptr(0x100)
 
-_unix_personality :: proc(persona: u64) -> int {
-	return int(intrinsics.syscall(unix.SYS_personality, uintptr(persona)))
-}
-
-_unix_fork :: proc() -> Pid {
-	when ODIN_ARCH != .arm64 {
-		res := int(intrinsics.syscall(unix.SYS_fork))
-	} else {
-		res := int(intrinsics.syscall(unix.SYS_clone, unix.SIGCHLD))
-	}
-	return -1 if res < 0 else Pid(res)
-}
-
-_unix_open :: proc(path: cstring, flags: int, mode: int = 0o000) -> Handle {
-	when ODIN_ARCH != .arm64 {
-		res := int(intrinsics.syscall(unix.SYS_open, uintptr(rawptr(path)), uintptr(flags), uintptr(mode)))
-	} else { // NOTE: arm64 does not have open
-		res := int(intrinsics.syscall(unix.SYS_openat, AT_FDCWD, uintptr(rawptr(path)), uintptr(flags), uintptr(mode)))
-	}
-	return -1 if res < 0 else Handle(res)
-}
-
-_unix_close :: proc(fd: Handle) -> int {
-	return int(intrinsics.syscall(unix.SYS_close, uintptr(fd)))
-}
-
-_unix_read :: proc(fd: Handle, buf: rawptr, size: uint) -> int {
-	return int(intrinsics.syscall(unix.SYS_read, uintptr(fd), uintptr(buf), uintptr(size)))
-}
-
-_unix_write :: proc(fd: Handle, buf: rawptr, size: uint) -> int {
-	return int(intrinsics.syscall(unix.SYS_write, uintptr(fd), uintptr(buf), uintptr(size)))
-}
-
-_unix_seek :: proc(fd: Handle, offset: i64, whence: int) -> i64 {
-	when ODIN_ARCH == .amd64 || ODIN_ARCH == .arm64 {
-		return i64(intrinsics.syscall(unix.SYS_lseek, uintptr(fd), uintptr(offset), uintptr(whence)))
-	} else {
-		low := uintptr(offset & 0xFFFFFFFF)
-		high := uintptr(offset >> 32)
-		result: i64
-		res := i64(intrinsics.syscall(unix.SYS__llseek, uintptr(fd), high, low, uintptr(&result), uintptr(whence)))
-		return -1 if res < 0 else result
-	}
-}
-
-_unix_stat :: proc(path: cstring, stat: ^OS_Stat) -> int {
-	when ODIN_ARCH == .amd64 {
-		return int(intrinsics.syscall(unix.SYS_stat, uintptr(rawptr(path)), uintptr(stat)))
-	} else when ODIN_ARCH != .arm64 {
-		return int(intrinsics.syscall(unix.SYS_stat64, uintptr(rawptr(path)), uintptr(stat)))
-	} else { // NOTE: arm64 does not have stat
-		return int(intrinsics.syscall(unix.SYS_fstatat, AT_FDCWD, uintptr(rawptr(path)), uintptr(stat), 0))
-	}
-}
-
-_unix_fstat :: proc(fd: Handle, stat: ^OS_Stat) -> int {
-	when ODIN_ARCH == .amd64 || ODIN_ARCH == .arm64 {
-		return int(intrinsics.syscall(unix.SYS_fstat, uintptr(fd), uintptr(stat)))
-	} else {
-		return int(intrinsics.syscall(unix.SYS_fstat64, uintptr(fd), uintptr(stat)))
-	}
-}
-
-_unix_lstat :: proc(path: cstring, stat: ^OS_Stat) -> int {
-	when ODIN_ARCH == .amd64 {
-		return int(intrinsics.syscall(unix.SYS_lstat, uintptr(rawptr(path)), uintptr(stat)))
-	} else when ODIN_ARCH != .arm64 {
-		return int(intrinsics.syscall(unix.SYS_lstat64, uintptr(rawptr(path)), uintptr(stat)))
-	} else { // NOTE: arm64 does not have any lstat
-		return int(intrinsics.syscall(unix.SYS_fstatat, AT_FDCWD, uintptr(rawptr(path)), uintptr(stat), AT_SYMLINK_NOFOLLOW))
-	}
-}
-
-_unix_readlink :: proc(path: cstring, buf: rawptr, bufsiz: uint) -> int {
-	when ODIN_ARCH != .arm64 {
-		return int(intrinsics.syscall(unix.SYS_readlink, uintptr(rawptr(path)), uintptr(buf), uintptr(bufsiz)))
-	} else { // NOTE: arm64 does not have readlink
-		return int(intrinsics.syscall(unix.SYS_readlinkat, AT_FDCWD, uintptr(rawptr(path)), uintptr(buf), uintptr(bufsiz)))
-	}
-}
-
-_unix_access :: proc(path: cstring, mask: int) -> int {
-	when ODIN_ARCH != .arm64 {
-		return int(intrinsics.syscall(unix.SYS_access, uintptr(rawptr(path)), uintptr(mask)))
-	} else { // NOTE: arm64 does not have access
-		return int(intrinsics.syscall(unix.SYS_faccessat, AT_FDCWD, uintptr(rawptr(path)), uintptr(mask)))
-	}
-}
-
-_unix_getcwd :: proc(buf: rawptr, size: uint) -> int {
-	return int(intrinsics.syscall(unix.SYS_getcwd, uintptr(buf), uintptr(size)))
-}
-
-_unix_chdir :: proc(path: cstring) -> int {
-	return int(intrinsics.syscall(unix.SYS_chdir, uintptr(rawptr(path))))
-}
-
-_unix_rename :: proc(old, new: cstring) -> int {
-	when ODIN_ARCH != .arm64 {
-		return int(intrinsics.syscall(unix.SYS_rename, uintptr(rawptr(old)), uintptr(rawptr(new))))
-	} else { // NOTE: arm64 does not have rename
-		return int(intrinsics.syscall(unix.SYS_renameat, AT_FDCWD, uintptr(rawptr(old)), uintptr(rawptr(new))))
-	}
-}
-
-_unix_unlink :: proc(path: cstring) -> int {
-	when ODIN_ARCH != .arm64 {
-		return int(intrinsics.syscall(unix.SYS_unlink, uintptr(rawptr(path))))
-	} else { // NOTE: arm64 does not have unlink
-		return int(intrinsics.syscall(unix.SYS_unlinkat, AT_FDCWD, uintptr(rawptr(path)), 0))
-	}
-}
-
-_unix_rmdir :: proc(path: cstring) -> int {
-	when ODIN_ARCH != .arm64 {
-		return int(intrinsics.syscall(unix.SYS_rmdir, uintptr(rawptr(path))))
-	} else { // NOTE: arm64 does not have rmdir
-		return int(intrinsics.syscall(unix.SYS_unlinkat, AT_FDCWD, uintptr(rawptr(path)), AT_REMOVEDIR))
-	}
-}
-
-_unix_mkdir :: proc(path: cstring, mode: u32) -> int {
-	when ODIN_ARCH != .arm64 {
-		return int(intrinsics.syscall(unix.SYS_mkdir, uintptr(rawptr(path)), uintptr(mode)))
-	} else { // NOTE: arm64 does not have mkdir
-		return int(intrinsics.syscall(unix.SYS_mkdirat, AT_FDCWD, uintptr(rawptr(path)), uintptr(mode)))
-	}
-}
-
 foreign libc {
 	@(link_name="__errno_location") __errno_location    :: proc() -> ^int ---
 
 	@(link_name="getpagesize")      _unix_getpagesize   :: proc() -> c.int ---
+	@(link_name="get_nprocs")       _unix_get_nprocs    :: proc() -> c.int ---
 	@(link_name="fdopendir")        _unix_fdopendir     :: proc(fd: Handle) -> Dir ---
 	@(link_name="closedir")         _unix_closedir      :: proc(dirp: Dir) -> c.int ---
 	@(link_name="rewinddir")        _unix_rewinddir     :: proc(dirp: Dir) ---
@@ -414,6 +285,7 @@ foreign libc {
 	@(link_name="free")             _unix_free          :: proc(ptr: rawptr) ---
 	@(link_name="realloc")          _unix_realloc       :: proc(ptr: rawptr, size: c.size_t) -> rawptr ---
 
+	@(link_name="execvp")           _unix_execvp       :: proc(path: cstring, argv: [^]cstring) -> int ---
 	@(link_name="getenv")           _unix_getenv        :: proc(cstring) -> cstring ---
 	@(link_name="putenv")           _unix_putenv        :: proc(cstring) -> c.int ---
 	@(link_name="realpath")         _unix_realpath      :: proc(path: cstring, resolved_path: rawptr) -> rawptr ---
@@ -441,12 +313,12 @@ _get_errno :: proc(res: int) -> Errno {
 }
 
 // get errno from libc
-get_last_error :: proc() -> int {
+get_last_error :: proc "contextless" () -> int {
 	return __errno_location()^
 }
 
 personality :: proc(persona: u64) -> (Errno) {
-	res := _unix_personality(persona)
+	res := unix.sys_personality(persona)
 	if res == -1 {
 		return _get_errno(res)
 	}
@@ -454,28 +326,48 @@ personality :: proc(persona: u64) -> (Errno) {
 }
 
 fork :: proc() -> (Pid, Errno) {
-	pid := _unix_fork()
+	pid := unix.sys_fork()
 	if pid == -1 {
-		return -1, _get_errno(int(pid))
+		return -1, _get_errno(pid)
 	}
-	return pid, ERROR_NONE
+	return Pid(pid), ERROR_NONE
 }
 
-open :: proc(path: string, flags: int = O_RDONLY, mode: int = 0) -> (Handle, Errno) {
-	cstr := strings.clone_to_cstring(path, context.temp_allocator)
-	handle := _unix_open(cstr, flags, mode)
-	if handle < 0 {
-		return INVALID_HANDLE, _get_errno(int(handle))
+execvp :: proc(path: string, args: []string) -> Errno {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
+	path_cstr := strings.clone_to_cstring(path, context.temp_allocator)
+
+	args_cstrs := make([]cstring, len(args) + 2, context.temp_allocator)
+	args_cstrs[0] = strings.clone_to_cstring(path, context.temp_allocator)
+	for i := 0; i < len(args); i += 1 {
+		args_cstrs[i+1] = strings.clone_to_cstring(args[i], context.temp_allocator)
 	}
-	return handle, ERROR_NONE
+
+	_unix_execvp(path_cstr, raw_data(args_cstrs))
+	return Errno(get_last_error())
+}
+
+
+open :: proc(path: string, flags: int = O_RDONLY, mode: int = 0o000) -> (Handle, Errno) {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
+	cstr := strings.clone_to_cstring(path, context.temp_allocator)
+	handle := unix.sys_open(cstr, flags, uint(mode))
+	if handle < 0 {
+		return INVALID_HANDLE, _get_errno(handle)
+	}
+	return Handle(handle), ERROR_NONE
 }
 
 close :: proc(fd: Handle) -> Errno {
-	return _get_errno(_unix_close(fd))
+	return _get_errno(unix.sys_close(int(fd)))
 }
 
 read :: proc(fd: Handle, data: []byte) -> (int, Errno) {
-	bytes_read := _unix_read(fd, &data[0], c.size_t(len(data)))
+	if len(data) == 0 {
+		return 0, ERROR_NONE
+	}
+
+	bytes_read := unix.sys_read(int(fd), raw_data(data), len(data))
 	if bytes_read < 0 {
 		return -1, _get_errno(bytes_read)
 	}
@@ -486,50 +378,78 @@ write :: proc(fd: Handle, data: []byte) -> (int, Errno) {
 	if len(data) == 0 {
 		return 0, ERROR_NONE
 	}
-	bytes_written := _unix_write(fd, &data[0], uint(len(data)))
+
+	bytes_written := unix.sys_write(int(fd), raw_data(data), len(data))
 	if bytes_written < 0 {
 		return -1, _get_errno(bytes_written)
 	}
-	return int(bytes_written), ERROR_NONE
+	return bytes_written, ERROR_NONE
+}
+read_at :: proc(fd: Handle, data: []byte, offset: i64) -> (int, Errno) {
+	if len(data) == 0 {
+		return 0, ERROR_NONE
+	}
+
+	bytes_read := unix.sys_pread(int(fd), raw_data(data), len(data), offset)
+	if bytes_read < 0 {
+		return -1, _get_errno(bytes_read)
+	}
+	return bytes_read, ERROR_NONE
+}
+
+write_at :: proc(fd: Handle, data: []byte, offset: i64) -> (int, Errno) {
+	if len(data) == 0 {
+		return 0, ERROR_NONE
+	}
+
+	bytes_written := unix.sys_pwrite(int(fd), raw_data(data), uint(len(data)), offset)
+	if bytes_written < 0 {
+		return -1, _get_errno(bytes_written)
+	}
+	return bytes_written, ERROR_NONE
 }
 
 seek :: proc(fd: Handle, offset: i64, whence: int) -> (i64, Errno) {
-	res := _unix_seek(fd, offset, whence)
+	res := unix.sys_lseek(int(fd), offset, whence)
 	if res < 0 {
 		return -1, _get_errno(int(res))
 	}
-	return res, ERROR_NONE
+	return i64(res), ERROR_NONE
 }
 
 file_size :: proc(fd: Handle) -> (i64, Errno) {
-    // deliberately uninitialized; the syscall fills this buffer for us
-    s: OS_Stat = ---
-    result := _unix_fstat(fd, &s)
-    if result < 0 {
-        return 0, _get_errno(result)
-    }
-    return max(s.size, 0), ERROR_NONE
+	// deliberately uninitialized; the syscall fills this buffer for us
+	s: OS_Stat = ---
+	result := unix.sys_fstat(int(fd), rawptr(&s))
+	if result < 0 {
+		return 0, _get_errno(result)
+	}
+	return max(s.size, 0), ERROR_NONE
 }
 
 rename :: proc(old_path, new_path: string) -> Errno {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	old_path_cstr := strings.clone_to_cstring(old_path, context.temp_allocator)
 	new_path_cstr := strings.clone_to_cstring(new_path, context.temp_allocator)
-	return _get_errno(_unix_rename(old_path_cstr, new_path_cstr))
+	return _get_errno(unix.sys_rename(old_path_cstr, new_path_cstr))
 }
 
 remove :: proc(path: string) -> Errno {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	path_cstr := strings.clone_to_cstring(path, context.temp_allocator)
-	return _get_errno(_unix_unlink(path_cstr))
+	return _get_errno(unix.sys_unlink(path_cstr))
 }
 
 make_directory :: proc(path: string, mode: u32 = 0o775) -> Errno {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	path_cstr := strings.clone_to_cstring(path, context.temp_allocator)
-	return _get_errno(_unix_mkdir(path_cstr, mode))
+	return _get_errno(unix.sys_mkdir(path_cstr, uint(mode)))
 }
 
 remove_directory :: proc(path: string) -> Errno {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	path_cstr := strings.clone_to_cstring(path, context.temp_allocator)
-	return _get_errno(_unix_rmdir(path_cstr))
+	return _get_errno(unix.sys_rmdir(path_cstr))
 }
 
 is_file_handle :: proc(fd: Handle) -> bool {
@@ -581,8 +501,9 @@ is_file :: proc {is_file_path, is_file_handle}
 is_dir :: proc {is_dir_path, is_dir_handle}
 
 exists :: proc(path: string) -> bool {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	cpath := strings.clone_to_cstring(path, context.temp_allocator)
-	res := _unix_access(cpath, O_RDONLY)
+	res := unix.sys_access(cpath, O_RDONLY)
 	return res == 0
 }
 
@@ -616,11 +537,12 @@ last_write_time_by_name :: proc(name: string) -> (File_Time, Errno) {
 
 @private
 _stat :: proc(path: string) -> (OS_Stat, Errno) {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	cstr := strings.clone_to_cstring(path, context.temp_allocator)
 
 	// deliberately uninitialized; the syscall fills this buffer for us
 	s: OS_Stat = ---
-	result := _unix_stat(cstr, &s)
+	result := unix.sys_stat(cstr, &s)
 	if result < 0 {
 		return s, _get_errno(result)
 	}
@@ -629,11 +551,12 @@ _stat :: proc(path: string) -> (OS_Stat, Errno) {
 
 @private
 _lstat :: proc(path: string) -> (OS_Stat, Errno) {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	cstr := strings.clone_to_cstring(path, context.temp_allocator)
 
 	// deliberately uninitialized; the syscall fills this buffer for us
 	s: OS_Stat = ---
-	result := _unix_lstat(cstr, &s)
+	result := unix.sys_lstat(cstr, &s)
 	if result < 0 {
 		return s, _get_errno(result)
 	}
@@ -644,7 +567,7 @@ _lstat :: proc(path: string) -> (OS_Stat, Errno) {
 _fstat :: proc(fd: Handle) -> (OS_Stat, Errno) {
 	// deliberately uninitialized; the syscall fills this buffer for us
 	s: OS_Stat = ---
-	result := _unix_fstat(fd, &s)
+	result := unix.sys_fstat(int(fd), rawptr(&s))
 	if result < 0 {
 		return s, _get_errno(result)
 	}
@@ -696,12 +619,13 @@ _readdir :: proc(dirp: Dir) -> (entry: Dirent, err: Errno, end_of_stream: bool)
 
 @private
 _readlink :: proc(path: string) -> (string, Errno) {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD(ignore = context.temp_allocator == context.allocator)
 	path_cstr := strings.clone_to_cstring(path, context.temp_allocator)
 
 	bufsz : uint = 256
 	buf := make([]byte, bufsz)
 	for {
-		rc := _unix_readlink(path_cstr, &(buf[0]), bufsz)
+		rc := unix.sys_readlink(path_cstr, &(buf[0]), bufsz)
 		if rc < 0 {
 			delete(buf)
 			return "", _get_errno(rc)
@@ -731,6 +655,7 @@ absolute_path_from_relative :: proc(rel: string) -> (path: string, err: Errno) {
 	if rel == "" {
 		rel = "."
 	}
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD(ignore = context.temp_allocator == context.allocator)
 
 	rel_cstr := strings.clone_to_cstring(rel, context.temp_allocator)
 
@@ -747,8 +672,9 @@ absolute_path_from_relative :: proc(rel: string) -> (path: string, err: Errno) {
 }
 
 access :: proc(path: string, mask: int) -> (bool, Errno) {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	cstr := strings.clone_to_cstring(path, context.temp_allocator)
-	result := _unix_access(cstr, mask)
+	result := unix.sys_access(cstr, mask)
 	if result < 0 {
 		return false, _get_errno(result)
 	}
@@ -777,6 +703,7 @@ heap_free :: proc(ptr: rawptr) {
 }
 
 lookup_env :: proc(key: string, allocator := context.allocator) -> (value: string, found: bool) {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD(ignore = context.temp_allocator == allocator)
 	path_str := strings.clone_to_cstring(key, context.temp_allocator)
 	// NOTE(tetra): Lifetime of 'cstr' is unclear, but _unix_free(cstr) segfaults.
 	cstr := _unix_getenv(path_str)
@@ -792,6 +719,7 @@ get_env :: proc(key: string, allocator := context.allocator) -> (value: string)
 }
 
 set_env :: proc(key, value: string) -> Errno {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	s := strings.concatenate({key, "=", value, "\x00"}, context.temp_allocator)
 	res := _unix_putenv(strings.unsafe_string_to_cstring(s))
 	if res < 0 {
@@ -801,6 +729,7 @@ set_env :: proc(key, value: string) -> Errno {
 }
 
 unset_env :: proc(key: string) -> Errno {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	s := strings.clone_to_cstring(key, context.temp_allocator)
 	res := _unix_putenv(s)
 	if res < 0 {
@@ -816,12 +745,13 @@ get_current_directory :: proc() -> string {
 	page_size := get_page_size()
 	buf := make([dynamic]u8, page_size)
 	for {
-		#no_bounds_check res := _unix_getcwd(&buf[0], uint(len(buf)))
+		#no_bounds_check res := unix.sys_getcwd(&buf[0], uint(len(buf)))
 
 		if res >= 0 {
 			return strings.string_from_nul_terminated_ptr(&buf[0], len(buf))
 		}
 		if _get_errno(res) != ERANGE {
+			delete(buf)
 			return ""
 		}
 		resize(&buf, len(buf)+page_size)
@@ -830,8 +760,9 @@ get_current_directory :: proc() -> string {
 }
 
 set_current_directory :: proc(path: string) -> (err: Errno) {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	cstr := strings.clone_to_cstring(path, context.temp_allocator)
-	res := _unix_chdir(cstr)
+	res := unix.sys_chdir(cstr)
 	if res < 0 {
 		return _get_errno(res)
 	}
@@ -848,12 +779,14 @@ current_thread_id :: proc "contextless" () -> int {
 }
 
 dlopen :: proc(filename: string, flags: int) -> rawptr {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	cstr := strings.clone_to_cstring(filename, context.temp_allocator)
 	handle := _unix_dlopen(cstr, c.int(flags))
 	return handle
 }
 dlsym :: proc(handle: rawptr, symbol: string) -> rawptr {
 	assert(handle != nil)
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	cstr := strings.clone_to_cstring(symbol, context.temp_allocator)
 	proc_handle := _unix_dlsym(handle, cstr)
 	return proc_handle
@@ -878,6 +811,10 @@ get_page_size :: proc() -> int {
 	return page_size
 }
 
+@(private)
+_processor_core_count :: proc() -> int {
+	return int(_unix_get_nprocs())
+}
 
 _alloc_command_line_arguments :: proc() -> []string {
 	res := make([]string, len(runtime.args__))

core/os/os_openbsd.odin

@@ -269,6 +269,7 @@ foreign libc {
 	@(link_name="mkdir")	_unix_mkdir	:: proc(path: cstring, mode: mode_t) -> c.int ---
 
 	@(link_name="getpagesize") _unix_getpagesize :: proc() -> c.int ---
+	@(link_name="sysconf") _sysconf :: proc(name: c.int) -> c.long ---
 	@(link_name="fdopendir") _unix_fdopendir :: proc(fd: Handle) -> Dir ---
 	@(link_name="closedir")	_unix_closedir	:: proc(dirp: Dir) -> c.int ---
 	@(link_name="rewinddir") _unix_rewinddir :: proc(dirp: Dir) ---
@@ -294,7 +295,7 @@ is_path_separator :: proc(r: rune) -> bool {
 	return r == '/'
 }
 
-get_last_error :: proc() -> int {
+get_last_error :: proc "contextless" () -> int {
 	return __errno()^
 }
 
@@ -307,6 +308,7 @@ fork :: proc() -> (Pid, Errno) {
 }
 
 open :: proc(path: string, flags: int = O_RDONLY, mode: int = 0) -> (Handle, Errno) {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	cstr := strings.clone_to_cstring(path, context.temp_allocator)
 	handle := _unix_open(cstr, c.int(flags), c.int(mode))
 	if handle == -1 {
@@ -359,6 +361,7 @@ file_size :: proc(fd: Handle) -> (i64, Errno) {
 }
 
 rename :: proc(old_path, new_path: string) -> Errno {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	old_path_cstr := strings.clone_to_cstring(old_path, context.temp_allocator)
 	new_path_cstr := strings.clone_to_cstring(new_path, context.temp_allocator)
 	res := _unix_rename(old_path_cstr, new_path_cstr)
@@ -369,6 +372,7 @@ rename :: proc(old_path, new_path: string) -> Errno {
 }
 
 remove :: proc(path: string) -> Errno {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	path_cstr := strings.clone_to_cstring(path, context.temp_allocator)
 	res := _unix_unlink(path_cstr)
 	if res == -1 {
@@ -378,6 +382,7 @@ remove :: proc(path: string) -> Errno {
 }
 
 make_directory :: proc(path: string, mode: mode_t = 0o775) -> Errno {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	path_cstr := strings.clone_to_cstring(path, context.temp_allocator)
 	res := _unix_mkdir(path_cstr, mode)
 	if res == -1 {
@@ -387,6 +392,7 @@ make_directory :: proc(path: string, mode: mode_t = 0o775) -> Errno {
 }
 
 remove_directory :: proc(path: string) -> Errno {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	path_cstr := strings.clone_to_cstring(path, context.temp_allocator)
 	res := _unix_rmdir(path_cstr)
 	if res == -1 {
@@ -472,6 +478,7 @@ last_write_time_by_name :: proc(name: string) -> (File_Time, Errno) {
 
 @private
 _stat :: proc(path: string) -> (OS_Stat, Errno) {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	cstr := strings.clone_to_cstring(path, context.temp_allocator)
 
 	// deliberately uninitialized
@@ -485,6 +492,7 @@ _stat :: proc(path: string) -> (OS_Stat, Errno) {
 
 @private
 _lstat :: proc(path: string) -> (OS_Stat, Errno) {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	cstr := strings.clone_to_cstring(path, context.temp_allocator)
 
 	// deliberately uninitialized
@@ -551,6 +559,7 @@ _readdir :: proc(dirp: Dir) -> (entry: Dirent, err: Errno, end_of_stream: bool)
 
 @private
 _readlink :: proc(path: string) -> (string, Errno) {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD(ignore = context.temp_allocator == context.allocator)
 	path_cstr := strings.clone_to_cstring(path, context.temp_allocator)
 
 	bufsz : uint = MAX_PATH
@@ -568,7 +577,6 @@ _readlink :: proc(path: string) -> (string, Errno) {
 			return strings.string_from_ptr(&buf[0], rc), ERROR_NONE
 		}	
 	}
-	unreachable()
 }
 
 // XXX OpenBSD
@@ -582,6 +590,7 @@ absolute_path_from_relative :: proc(rel: string) -> (path: string, err: Errno) {
 		rel = "."
 	}
 
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD(ignore = context.temp_allocator == context.allocator)
 	rel_cstr := strings.clone_to_cstring(rel, context.temp_allocator)
 
 	path_ptr := _unix_realpath(rel_cstr, nil)
@@ -597,6 +606,7 @@ absolute_path_from_relative :: proc(rel: string) -> (path: string, err: Errno) {
 }
 
 access :: proc(path: string, mask: int) -> (bool, Errno) {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	cstr := strings.clone_to_cstring(path, context.temp_allocator)
 	res := _unix_access(cstr, c.int(mask))
 	if res == -1 {
@@ -627,6 +637,7 @@ heap_free :: proc(ptr: rawptr) {
 }
 
 lookup_env :: proc(key: string, allocator := context.allocator) -> (value: string, found: bool) {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD(ignore = context.temp_allocator == allocator)
 	path_str := strings.clone_to_cstring(key, context.temp_allocator)
 	cstr := _unix_getenv(path_str)
 	if cstr == nil {
@@ -648,6 +659,7 @@ get_current_directory :: proc() -> string {
 			return string(cwd)
 		}
 		if Errno(get_last_error()) != ERANGE {
+			delete(buf)
 			return ""
 		}
 		resize(&buf, len(buf) + MAX_PATH)
@@ -656,6 +668,7 @@ get_current_directory :: proc() -> string {
 }
 
 set_current_directory :: proc(path: string) -> (err: Errno) {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	cstr := strings.clone_to_cstring(path, context.temp_allocator)
 	res := _unix_chdir(cstr)
 	if res == -1 {
@@ -674,12 +687,14 @@ current_thread_id :: proc "contextless" () -> int {
 }
 
 dlopen :: proc(filename: string, flags: int) -> rawptr {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	cstr := strings.clone_to_cstring(filename, context.temp_allocator)
 	handle := _unix_dlopen(cstr, c.int(flags))
 	return handle
 }
 dlsym :: proc(handle: rawptr, symbol: string) -> rawptr {
 	assert(handle != nil)
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	cstr := strings.clone_to_cstring(symbol, context.temp_allocator)
 	proc_handle := _unix_dlsym(handle, cstr)
 	return proc_handle
@@ -704,6 +719,12 @@ get_page_size :: proc() -> int {
 	return page_size
 }
 
+_SC_NPROCESSORS_ONLN :: 503
+
+@(private)
+_processor_core_count :: proc() -> int {
+	return int(_sysconf(_SC_NPROCESSORS_ONLN))
+}
 
 _alloc_command_line_arguments :: proc() -> []string {
 	res := make([]string, len(runtime.args__))
@@ -711,4 +732,4 @@ _alloc_command_line_arguments :: proc() -> []string {
 		res[i] = string(arg)
 	}
 	return res
-}
+}

core/os/os_wasi.odin

@@ -89,7 +89,10 @@ seek :: proc(fd: Handle, offset: i64, whence: int) -> (i64, Errno) {
 current_thread_id :: proc "contextless" () -> int {
 	return 0
 }
-
+@(private)
+_processor_core_count :: proc() -> int {
+	return 1
+}
 
 file_size :: proc(fd: Handle) -> (i64, Errno) {
 	stat, err := wasi.fd_filestat_get(wasi.fd_t(fd))

core/os/os_windows.odin

@@ -3,6 +3,7 @@ package os
 
 import win32 "core:sys/windows"
 import "core:runtime"
+import "core:intrinsics"
 
 Handle    :: distinct uintptr
 File_Time :: distinct u64
@@ -126,7 +127,29 @@ get_page_size :: proc() -> int {
 	return page_size
 }
 
+@(private)
+_processor_core_count :: proc() -> int {
+	length : win32.DWORD = 0
+	result := win32.GetLogicalProcessorInformation(nil, &length)
 
+	thread_count := 0
+	if !result && win32.GetLastError() == 122 && length > 0 {
+		runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
+		processors := make([]win32.SYSTEM_LOGICAL_PROCESSOR_INFORMATION, length, context.temp_allocator)
+
+		result = win32.GetLogicalProcessorInformation(&processors[0], &length)
+		if result {
+			for processor in processors {
+				if processor.Relationship == .RelationProcessorCore {
+					thread := intrinsics.count_ones(processor.ProcessorMask)
+					thread_count += int(thread)
+				}
+			}
+		}
+	}
+
+	return thread_count
+}
 
 exit :: proc "contextless" (code: int) -> ! {
 	runtime._cleanup_runtime_contextless()
@@ -214,4 +237,4 @@ is_windows_10 :: proc() -> bool {
 is_windows_11 :: proc() -> bool {
 	osvi := get_windows_version_w()
 	return (osvi.dwMajorVersion == 10 && osvi.dwMinorVersion == 0 && osvi.dwBuildNumber >= WINDOWS_11_BUILD_CUTOFF)
-}
+}

core/os/stat_windows.odin

@@ -1,6 +1,7 @@
 package os
 
 import "core:time"
+import "core:runtime"
 import win32 "core:sys/windows"
 
 @(private)
@@ -11,6 +12,7 @@ full_path_from_name :: proc(name: string, allocator := context.allocator) -> (pa
 	if name == "" {
 		name = "."
 	}
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD(ignore = context.temp_allocator == allocator)
 	p := win32.utf8_to_utf16(name, context.temp_allocator)
 	buf := make([dynamic]u16, 100)
 	defer delete(buf)
@@ -36,6 +38,7 @@ _stat :: proc(name: string, create_file_attributes: u32, allocator := context.al
 
 	context.allocator = allocator
 
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD(ignore = context.temp_allocator == allocator)
 
 	wname := win32.utf8_to_wstring(fix_long_path(name), context.temp_allocator)
 	fa: win32.WIN32_FILE_ATTRIBUTE_DATA
@@ -132,14 +135,15 @@ cleanpath_strip_prefix :: proc(buf: []u16) -> []u16 {
 
 @(private)
 cleanpath_from_handle :: proc(fd: Handle) -> (string, Errno) {
-	buf, err := cleanpath_from_handle_u16(fd)
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD(ignore = context.temp_allocator == context.allocator)
+	buf, err := cleanpath_from_handle_u16(fd, context.temp_allocator)
 	if err != 0 {
 		return "", err
 	}
 	return win32.utf16_to_utf8(buf, context.allocator) or_else "", err
 }
 @(private)
-cleanpath_from_handle_u16 :: proc(fd: Handle) -> ([]u16, Errno) {
+cleanpath_from_handle_u16 :: proc(fd: Handle, allocator: runtime.Allocator) -> ([]u16, Errno) {
 	if fd == 0 {
 		return nil, ERROR_INVALID_HANDLE
 	}
@@ -149,7 +153,7 @@ cleanpath_from_handle_u16 :: proc(fd: Handle) -> ([]u16, Errno) {
 	if n == 0 {
 		return nil, Errno(win32.GetLastError())
 	}
-	buf := make([]u16, max(n, win32.DWORD(260))+1, context.temp_allocator)
+	buf := make([]u16, max(n, win32.DWORD(260))+1, allocator)
 	buf_len := win32.GetFinalPathNameByHandleW(h, raw_data(buf), n, 0)
 	return buf[:buf_len], ERROR_NONE
 }

core/path/filepath/match.odin

@@ -218,7 +218,6 @@ get_escape :: proc(chunk: string) -> (r: rune, next_chunk: string, err: Match_Er
 //
 // glob ignores file system errors
 //
-
 glob :: proc(pattern: string, allocator := context.allocator) -> (matches: []string, err: Match_Error) {
 	context.allocator = allocator
 
@@ -261,6 +260,8 @@ glob :: proc(pattern: string, allocator := context.allocator) -> (matches: []str
 	}
 	return
 }
+
+// Internal implementation of `glob`, not meant to be used by the user. Prefer `glob`.
 _glob :: proc(dir, pattern: string, matches: ^[dynamic]string, allocator := context.allocator) -> (m: [dynamic]string, e: Match_Error) {
 	context.allocator = allocator
 

core/path/filepath/path.odin

@@ -20,6 +20,8 @@ is_slash :: proc(c: byte) -> bool {
 	return c == '\\' || c == '/'
 }
 
+// Splits path immediate following the last separator; separating the path into a directory and file.
+// If no separator is found, `dir` will be empty and `path` set to `path`.
 split :: proc(path: string) -> (dir, file: string) {
 	vol := volume_name(path)
 	i := len(path) - 1
@@ -29,10 +31,18 @@ split :: proc(path: string) -> (dir, file: string) {
 	return path[:i+1], path[i+1:]
 }
 
+/*
+	Returns leading volume name.
+
+	e.g.
+	  "C:\foo\bar\baz" will return "C:" on Windows.
+	  Everything else will be "".
+*/
 volume_name :: proc(path: string) -> string {
 	return path[:volume_name_len(path)]
 }
 
+// Returns the length of the volume name in bytes.
 volume_name_len :: proc(path: string) -> int {
 	if ODIN_OS == .Windows {
 		if len(path) < 2 {
@@ -74,7 +84,7 @@ volume_name_len :: proc(path: string) -> int {
 /*
 	Gets the file name and extension from a path.
 
-	i.e:
+	e.g.
 	  'path/to/name.tar.gz' -> 'name.tar.gz'
 	  'path/to/name.txt'    -> 'name.txt'
 	  'path/to/name'        -> 'name'
@@ -114,7 +124,7 @@ base :: proc(path: string) -> string {
 	Only the last dot is considered when splitting the file extension.
 	See `short_stem`.
 
-	i.e:
+	e.g.
 	  'name.tar.gz' -> 'name.tar'
 	  'name.txt'    -> 'name'
 
@@ -147,7 +157,7 @@ stem :: proc(path: string) -> string {
 
 	The first dot is used to split off the file extension, unlike `stem` which uses the last dot.
 
-	i.e:
+	e.g.
 	  'name.tar.gz' -> 'name'
 	  'name.txt'    -> 'name'
 
@@ -170,7 +180,7 @@ short_stem :: proc(path: string) -> string {
 	Only the last dot is considered when splitting the file extension.
 	See `long_ext`.
 
-	i.e:
+	e.g.
 	  'name.tar.gz' -> '.gz'
 	  'name.txt'    -> '.txt'
 
@@ -193,7 +203,7 @@ ext :: proc(path: string) -> string {
 
 	The first dot is used to split off the file extension, unlike `ext` which uses the last dot.
 
-	i.e:
+	e.g.
 	  'name.tar.gz' -> '.tar.gz'
 	  'name.txt'    -> '.txt'
 
@@ -219,6 +229,21 @@ long_ext :: proc(path: string) -> string {
 	return ""
 }
 
+/*
+	Returns the shortest path name equivalent to `path` through solely lexical processing.
+	It applies the folliwng rules until none of them can be applied:
+
+	* Replace multiple separators with a single one
+	* Remove each current directory (`.`) path name element
+	* Remove each inner parent directory (`..`) path and the preceding paths
+	* Remove `..` that begin at the root of a path
+	* All possible separators are replaced with the OS specific separator
+
+	The return path ends in a slash only if it represents the root of a directory (`C:\` on Windows and  `/` on *nix systems).
+
+	If the result of the path is an empty string, the returned path with be `"."`.
+
+*/
 clean :: proc(path: string, allocator := context.allocator) -> string {
 	context.allocator = allocator
 
@@ -299,6 +324,7 @@ clean :: proc(path: string, allocator := context.allocator) -> string {
 	return cleaned
 }
 
+// Returns the result of replacing each forward slash `/` character in the path with the separate OS specific character.
 from_slash :: proc(path: string, allocator := context.allocator) -> (new_path: string, new_allocation: bool) {
 	if SEPARATOR == '/' {
 		return path, false
@@ -306,6 +332,7 @@ from_slash :: proc(path: string, allocator := context.allocator) -> (new_path: s
 	return strings.replace_all(path, "/", SEPARATOR_STRING, allocator)
 }
 
+// Returns the result of replacing each OS specific separator with a forward slash `/` character.
 to_slash :: proc(path: string, allocator := context.allocator) -> (new_path: string, new_allocation: bool) {
 	if SEPARATOR == '/' {
 		return path, false
@@ -320,6 +347,13 @@ Relative_Error :: enum {
 	Cannot_Relate,
 }
 
+/*
+	Returns a relative path that is lexically equivalent to the `target_path` when joined with the `base_path` with an OS specific separator.
+
+	e.g. `join(base_path, rel(base_path, target_path))` is equivalent to `target_path`
+
+	On failure, the `Relative_Error` will be state it cannot compute the necessary relative path.
+*/
 rel :: proc(base_path, target_path: string, allocator := context.allocator) -> (string, Relative_Error) {
 	context.allocator = allocator
 	base_clean, target_clean := clean(base_path), clean(target_path)
@@ -398,6 +432,11 @@ rel :: proc(base_path, target_path: string, allocator := context.allocator) -> (
 	return target[t0:], .None
 }
 
+/*
+	Returns all but the last element path, usually the path's directory. Once the final element has been removed,
+	`dir` calls `clean` on the path and trailing separators are removed. If the path consists purely of separators,
+	then `"."` is returned.
+*/
 dir :: proc(path: string, allocator := context.allocator) -> string {
         context.allocator = allocator
 	vol := volume_name(path)

core/path/filepath/path_unix.odin

@@ -7,6 +7,7 @@ when ODIN_OS == .Darwin {
 	foreign import libc "system:c"
 }
 
+import "core:runtime"
 import "core:strings"
 
 SEPARATOR :: '/'
@@ -41,6 +42,7 @@ abs :: proc(path: string, allocator := context.allocator) -> (string, bool) {
 join :: proc(elems: []string, allocator := context.allocator) -> string {
 	for e, i in elems {
 		if e != "" {
+			runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD(ignore = context.temp_allocator == allocator)
 			p := strings.join(elems[i:], SEPARATOR_STRING, context.temp_allocator)
 			return clean(p, allocator)
 		}

core/path/filepath/path_windows.odin

@@ -1,6 +1,7 @@
 package filepath
 
 import "core:strings"
+import "core:runtime"
 import "core:os"
 import win32 "core:sys/windows"
 
@@ -60,25 +61,25 @@ temp_full_path :: proc(name: string) -> (path: string, err: os.Errno) {
 	}
 
 	p := win32.utf8_to_utf16(name, ta)
-	buf := make([dynamic]u16, 100, ta)
-	for {
-		n := win32.GetFullPathNameW(raw_data(p), u32(len(buf)), raw_data(buf), nil)
-		if n == 0 {
-			delete(buf)
-			return "", os.Errno(win32.GetLastError())
-		}
-		if n <= u32(len(buf)) {
-			return win32.utf16_to_utf8(buf[:n], ta) or_else "", os.ERROR_NONE
-		}
-		resize(&buf, len(buf)*2)
+	n := win32.GetFullPathNameW(raw_data(p), 0, nil, nil)
+	if n == 0 {
+		return "", os.Errno(win32.GetLastError())
 	}
 
-	return
+	buf := make([]u16, n, ta)
+	n = win32.GetFullPathNameW(raw_data(p), u32(len(buf)), raw_data(buf), nil)
+	if n == 0 {
+		delete(buf)
+		return "", os.Errno(win32.GetLastError())
+	}
+
+	return win32.utf16_to_utf8(buf[:n], ta) or_else "", os.ERROR_NONE
 }
 
 
 
 abs :: proc(path: string, allocator := context.allocator) -> (string, bool) {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD(ignore = allocator == context.temp_allocator)
 	full_path, err := temp_full_path(path)
 	if err != 0 {
 		return "", false
@@ -99,6 +100,8 @@ join :: proc(elems: []string, allocator := context.allocator) -> string {
 
 join_non_empty :: proc(elems: []string, allocator := context.allocator) -> string {
 	context.allocator = allocator
+
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD(ignore = allocator == context.temp_allocator)
 	
 	if len(elems[0]) == 2 && elems[0][1] == ':' {
 		i := 1

core/path/slashpath/path.odin

@@ -5,6 +5,7 @@
 // To manipulate operating system specific paths, use the path/filepath package
 package slashpath
 
+import "core:runtime"
 import "core:strings"
 
 // is_separator checks whether the byte is a valid separator character
@@ -150,8 +151,9 @@ join :: proc(elems: []string, allocator := context.allocator) -> string {
 	context.allocator = allocator
 	for elem, i in elems {
 		if elem != "" {
+			runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD(ignore = context.temp_allocator == allocator)
 			s := strings.join(elems[i:], "/", context.temp_allocator)
-			return clean(s)
+			return clean(s, allocator)
 		}
 	}
 	return ""

core/prof/spall/spall.odin

@@ -0,0 +1,208 @@
+package prof_spall
+
+import "core:os"
+import "core:time"
+import "core:intrinsics"
+import "core:mem"
+
+// File Format
+
+MANUAL_MAGIC :: u64le(0x0BADF00D)
+
+Manual_Header :: struct #packed {
+	magic:           u64le,
+	version:         u64le,
+	timestamp_scale: f64le,
+	reserved:        u64le,
+}
+
+Manual_Event_Type :: enum u8 {
+	Invalid             = 0,
+
+	Begin               = 3,
+	End                 = 4,
+	Instant             = 5,
+
+	Pad_Skip            = 7,
+}
+
+Begin_Event :: struct #packed {
+	type:     Manual_Event_Type,
+	category: u8,
+	pid:      u32le,
+	tid:      u32le,
+	ts:       f64le,
+	name_len: u8,
+	args_len: u8,
+}
+BEGIN_EVENT_MAX :: size_of(Begin_Event) + 255 + 255
+
+End_Event :: struct #packed {
+	type: Manual_Event_Type,
+	pid:  u32le,
+	tid:  u32le,
+	ts:   f64le,
+}
+
+Pad_Skip :: struct #packed {
+	type: Manual_Event_Type,
+	size: u32le,
+}
+
+// User Interface
+
+Context :: struct {
+	precise_time:    bool,
+	timestamp_scale: f64,
+	fd:              os.Handle,
+}
+
+Buffer :: struct {
+	data: []u8,
+	head: int,
+	tid:  u32,
+	pid:  u32,
+}
+
+BUFFER_DEFAULT_SIZE :: 0x10_0000
+
+
+context_create :: proc(filename: string) -> (ctx: Context, ok: bool) #optional_ok {
+	fd, err := os.open(filename, os.O_WRONLY | os.O_APPEND | os.O_CREATE | os.O_TRUNC, 0o600)
+	if err != os.ERROR_NONE {
+		return
+	}
+	ctx.fd = fd
+
+	freq, freq_ok := time.tsc_frequency()
+	ctx.precise_time = freq_ok
+	ctx.timestamp_scale = ((1 / f64(freq)) * 1_000_000) if freq_ok else 1
+
+	temp := [size_of(Manual_Header)]u8{}
+	_build_header(temp[:], ctx.timestamp_scale)
+	os.write(ctx.fd, temp[:])
+	ok = true
+	return
+}
+
+context_destroy :: proc(ctx: ^Context) {
+	if ctx == nil {
+		return
+	}
+
+	os.close(ctx.fd)
+	ctx^ = Context{}
+}
+
+buffer_create :: proc(data: []byte, tid: u32 = 0, pid: u32 = 0) -> (buffer: Buffer, ok: bool) #optional_ok {
+	assert(len(data) > 0)
+	buffer.data = data
+	buffer.tid  = tid
+	buffer.pid  = pid
+	buffer.head = 0
+	ok = true
+	return
+}
+
+buffer_flush :: proc(ctx: ^Context, buffer: ^Buffer) {
+	os.write(ctx.fd, buffer.data[:buffer.head])
+	buffer.head = 0
+}
+
+buffer_destroy :: proc(ctx: ^Context, buffer: ^Buffer) {
+	buffer_flush(ctx, buffer)
+
+	buffer^ = Buffer{}
+}
+
+
+
+@(deferred_in=_scoped_buffer_end)
+SCOPED_EVENT :: proc(ctx: ^Context, buffer: ^Buffer, name: string, args: string = "", location := #caller_location) -> bool {
+	_buffer_begin(ctx, buffer, name, args, location)
+	return true
+}
+
+@(private)
+_scoped_buffer_end :: proc(ctx: ^Context, buffer: ^Buffer, _, _: string, _ := #caller_location) {
+	_buffer_end(ctx, buffer)
+}
+
+
+_trace_now :: proc "contextless" (ctx: ^Context) -> f64 {
+	if !ctx.precise_time {
+		return f64(time.tick_now()._nsec) / 1_000
+	}
+
+	return f64(intrinsics.read_cycle_counter())
+}
+
+_build_header :: proc "contextless" (buffer: []u8, timestamp_scale: f64) -> (header_size: int, ok: bool) #optional_ok {
+	header_size = size_of(Manual_Header)
+	if header_size > len(buffer) {
+		return 0, false
+	}
+
+	hdr := (^Manual_Header)(raw_data(buffer))
+	hdr.magic = MANUAL_MAGIC
+	hdr.version = 1
+	hdr.timestamp_scale = f64le(timestamp_scale)
+	hdr.reserved = 0
+	ok = true
+	return
+}
+
+_build_begin :: proc "contextless" (buffer: []u8, name: string, args: string, ts: f64, tid: u32, pid: u32) -> (event_size: int, ok: bool) #optional_ok {
+	ev := (^Begin_Event)(raw_data(buffer))
+	name_len := min(len(name), 255)
+	args_len := min(len(args), 255)
+
+	event_size = size_of(Begin_Event) + name_len + args_len
+	if event_size > len(buffer) {
+		return 0, false
+	}
+
+	ev.type = .Begin
+	ev.pid  = u32le(pid)
+	ev.tid  = u32le(tid)
+	ev.ts   = f64le(ts)
+	ev.name_len = u8(name_len)
+	ev.args_len = u8(args_len)
+	mem.copy(raw_data(buffer[size_of(Begin_Event):]), raw_data(name), name_len)
+	mem.copy(raw_data(buffer[size_of(Begin_Event)+name_len:]), raw_data(args), args_len)
+	ok = true
+	return
+}
+
+_build_end :: proc(buffer: []u8, ts: f64, tid: u32, pid: u32) -> (event_size: int, ok: bool) #optional_ok {
+	ev := (^End_Event)(raw_data(buffer))
+	event_size = size_of(End_Event)
+	if event_size > len(buffer) {
+		return 0, false
+	}
+
+	ev.type = .End
+	ev.pid  = u32le(pid)
+	ev.tid  = u32le(tid)
+	ev.ts   = f64le(ts)
+	ok = true
+	return
+}
+
+_buffer_begin :: proc(ctx: ^Context, buffer: ^Buffer, name: string, args: string = "", location := #caller_location) {
+	if buffer.head + BEGIN_EVENT_MAX > len(buffer.data) {
+		buffer_flush(ctx, buffer)
+	}
+	name := location.procedure if name == "" else name
+	buffer.head += _build_begin(buffer.data[buffer.head:], name, args, _trace_now(ctx), buffer.tid, buffer.pid)
+}
+
+_buffer_end :: proc(ctx: ^Context, buffer: ^Buffer) {
+	ts := _trace_now(ctx)
+
+	if buffer.head + size_of(End_Event) > len(buffer.data) {
+		buffer_flush(ctx, buffer)
+	}
+
+	buffer.head += _build_end(buffer.data[buffer.head:], ts, buffer.tid, buffer.pid)
+}

core/reflect/iterator.odin

@@ -2,6 +2,7 @@ package reflect
 
 import "core:runtime"
 
+@(require_results)
 iterate_array :: proc(val: any, it: ^int) -> (elem: any, index: int, ok: bool) {
 	if val == nil || it == nil {
 		return
@@ -41,6 +42,7 @@ iterate_array :: proc(val: any, it: ^int) -> (elem: any, index: int, ok: bool) {
 	return
 }
 
+@(require_results)
 iterate_map :: proc(val: any, it: ^int) -> (key, value: any, ok: bool) {
 	if val == nil || it == nil {
 		return

core/reflect/reflect.odin

@@ -25,7 +25,8 @@ Type_Info_Array            :: runtime.Type_Info_Array
 Type_Info_Enumerated_Array :: runtime.Type_Info_Enumerated_Array
 Type_Info_Dynamic_Array    :: runtime.Type_Info_Dynamic_Array
 Type_Info_Slice            :: runtime.Type_Info_Slice
-Type_Info_Tuple            :: runtime.Type_Info_Tuple
+Type_Info_Parameters       :: runtime.Type_Info_Parameters
+Type_Info_Tuple            :: runtime.Type_Info_Parameters
 Type_Info_Struct           :: runtime.Type_Info_Struct
 Type_Info_Union            :: runtime.Type_Info_Union
 Type_Info_Enum             :: runtime.Type_Info_Enum
@@ -74,6 +75,7 @@ Type_Kind :: enum {
 }
 
 
+@(require_results)
 type_kind :: proc(T: typeid) -> Type_Kind {
 	ti := type_info_of(T)
 	if ti != nil {
@@ -95,7 +97,7 @@ type_kind :: proc(T: typeid) -> Type_Kind {
 		case Type_Info_Enumerated_Array: return .Enumerated_Array
 		case Type_Info_Dynamic_Array:    return .Dynamic_Array
 		case Type_Info_Slice:            return .Slice
-		case Type_Info_Tuple:            return .Tuple
+		case Type_Info_Parameters:       return .Tuple
 		case Type_Info_Struct:           return .Struct
 		case Type_Info_Union:            return .Union
 		case Type_Info_Enum:             return .Enum
@@ -113,57 +115,31 @@ type_kind :: proc(T: typeid) -> Type_Kind {
 }
 
 // TODO(bill): Better name
+@(require_results)
 underlying_type_kind :: proc(T: typeid) -> Type_Kind {
 	return type_kind(runtime.typeid_base(T))
 }
 
 // TODO(bill): Better name
+@(require_results)
 backing_type_kind :: proc(T: typeid) -> Type_Kind {
 	return type_kind(runtime.typeid_core(T))
 }
 
 
-type_info_base :: proc(info: ^Type_Info) -> ^Type_Info {
-	if info == nil { return nil }
-
-	base := info
-	loop: for {
-		#partial switch i in base.variant {
-		case Type_Info_Named: base = i.base
-		case: break loop
-		}
-	}
-	return base
-}
-
-
-type_info_core :: proc(info: ^Type_Info) -> ^Type_Info {
-	if info == nil { return nil }
-
-	base := info
-	loop: for {
-		#partial switch i in base.variant {
-		case Type_Info_Named:  base = i.base
-		case Type_Info_Enum:   base = i.base
-		case: break loop
-		}
-	}
-	return base
-}
+type_info_base :: runtime.type_info_base
+type_info_core :: runtime.type_info_core 
 type_info_base_without_enum :: type_info_core
 
 
-typeid_base :: proc(id: typeid) -> typeid {
-	ti := type_info_of(id)
-	ti = type_info_base(ti)
-	return ti.id
+when !ODIN_DISALLOW_RTTI {
+	typeid_base :: runtime.typeid_base
+	typeid_core :: runtime.typeid_core
+	typeid_base_without_enum :: typeid_core
 }
-typeid_core :: proc(id: typeid) -> typeid {
-	ti := type_info_base_without_enum(type_info_of(id))
-	return ti.id
-}
-typeid_base_without_enum :: typeid_core
 
+
+@(require_results)
 any_base :: proc(v: any) -> any {
 	v := v
 	if v != nil {
@@ -171,6 +147,7 @@ any_base :: proc(v: any) -> any {
 	}
 	return v
 }
+@(require_results)
 any_core :: proc(v: any) -> any {
 	v := v
 	if v != nil {
@@ -179,6 +156,7 @@ any_core :: proc(v: any) -> any {
 	return v
 }
 
+@(require_results)
 typeid_elem :: proc(id: typeid) -> typeid {
 	ti := type_info_of(id)
 	if ti == nil { return nil }
@@ -208,6 +186,7 @@ typeid_elem :: proc(id: typeid) -> typeid {
 }
 
 
+@(require_results)
 size_of_typeid :: proc(T: typeid) -> int {
 	if ti := type_info_of(T); ti != nil {
 		return ti.size
@@ -215,6 +194,7 @@ size_of_typeid :: proc(T: typeid) -> int {
 	return 0
 }
 
+@(require_results)
 align_of_typeid :: proc(T: typeid) -> int {
 	if ti := type_info_of(T); ti != nil {
 		return ti.align
@@ -222,6 +202,7 @@ align_of_typeid :: proc(T: typeid) -> int {
 	return 1
 }
 
+@(require_results)
 as_bytes :: proc(v: any) -> []byte {
 	if v != nil {
 		sz := size_of_typeid(v.id)
@@ -230,10 +211,12 @@ as_bytes :: proc(v: any) -> []byte {
 	return nil
 }
 
+@(require_results)
 any_data :: #force_inline proc(v: any) -> (data: rawptr, id: typeid) {
 	return v.data, v.id
 }
 
+@(require_results)
 is_nil :: proc(v: any) -> bool {
 	if v == nil {
 		return true
@@ -250,6 +233,7 @@ is_nil :: proc(v: any) -> bool {
 	return true
 }
 
+@(require_results)
 length :: proc(val: any) -> int {
 	if val == nil { return 0 }
 
@@ -285,6 +269,7 @@ length :: proc(val: any) -> int {
 	return 0
 }
 
+@(require_results)
 capacity :: proc(val: any) -> int {
 	if val == nil { return 0 }
 
@@ -311,6 +296,7 @@ capacity :: proc(val: any) -> int {
 }
 
 
+@(require_results)
 index :: proc(val: any, i: int, loc := #caller_location) -> any {
 	if val == nil { return nil }
 
@@ -370,6 +356,7 @@ index :: proc(val: any, i: int, loc := #caller_location) -> any {
 	return nil
 }
 
+@(require_results)
 deref :: proc(val: any) -> any {
 	if val != nil {
 		ti := type_info_base(type_info_of(val.id))
@@ -399,6 +386,7 @@ Struct_Field :: struct {
 	is_using: bool,
 }
 
+@(require_results)
 struct_field_at :: proc(T: typeid, i: int) -> (field: Struct_Field) {
 	ti := runtime.type_info_base(type_info_of(T))
 	if s, ok := ti.variant.(runtime.Type_Info_Struct); ok {
@@ -413,6 +401,7 @@ struct_field_at :: proc(T: typeid, i: int) -> (field: Struct_Field) {
 	return
 }
 
+@(require_results)
 struct_field_by_name :: proc(T: typeid, name: string) -> (field: Struct_Field) {
 	ti := runtime.type_info_base(type_info_of(T))
 	if s, ok := ti.variant.(runtime.Type_Info_Struct); ok {
@@ -430,6 +419,7 @@ struct_field_by_name :: proc(T: typeid, name: string) -> (field: Struct_Field) {
 	return
 }
 
+@(require_results)
 struct_field_value_by_name :: proc(a: any, field: string, allow_using := false) -> any {
 	if a == nil { return nil }
 
@@ -461,6 +451,7 @@ struct_field_value_by_name :: proc(a: any, field: string, allow_using := false)
 
 
 
+@(require_results)
 struct_field_names :: proc(T: typeid) -> []string {
 	ti := runtime.type_info_base(type_info_of(T))
 	if s, ok := ti.variant.(runtime.Type_Info_Struct); ok {
@@ -469,6 +460,7 @@ struct_field_names :: proc(T: typeid) -> []string {
 	return nil
 }
 
+@(require_results)
 struct_field_types :: proc(T: typeid) -> []^Type_Info {
 	ti := runtime.type_info_base(type_info_of(T))
 	if s, ok := ti.variant.(runtime.Type_Info_Struct); ok {
@@ -478,6 +470,7 @@ struct_field_types :: proc(T: typeid) -> []^Type_Info {
 }
 
 
+@(require_results)
 struct_field_tags :: proc(T: typeid) -> []Struct_Tag {
 	ti := runtime.type_info_base(type_info_of(T))
 	if s, ok := ti.variant.(runtime.Type_Info_Struct); ok {
@@ -486,6 +479,7 @@ struct_field_tags :: proc(T: typeid) -> []Struct_Tag {
 	return nil
 }
 
+@(require_results)
 struct_field_offsets :: proc(T: typeid) -> []uintptr {
 	ti := runtime.type_info_base(type_info_of(T))
 	if s, ok := ti.variant.(runtime.Type_Info_Struct); ok {
@@ -494,6 +488,7 @@ struct_field_offsets :: proc(T: typeid) -> []uintptr {
 	return nil
 }
 
+@(require_results)
 struct_fields_zipped :: proc(T: typeid) -> (fields: #soa[]Struct_Field) {
 	ti := runtime.type_info_base(type_info_of(T))
 	if s, ok := ti.variant.(runtime.Type_Info_Struct); ok {
@@ -510,11 +505,13 @@ struct_fields_zipped :: proc(T: typeid) -> (fields: #soa[]Struct_Field) {
 
 
 
+@(require_results)
 struct_tag_get :: proc(tag: Struct_Tag, key: string) -> (value: Struct_Tag) {
 	value, _ = struct_tag_lookup(tag, key)
 	return
 }
 
+@(require_results)
 struct_tag_lookup :: proc(tag: Struct_Tag, key: string) -> (value: Struct_Tag, ok: bool) {
 	for t := tag; t != ""; /**/ {
 		i := 0
@@ -573,6 +570,7 @@ struct_tag_lookup :: proc(tag: Struct_Tag, key: string) -> (value: Struct_Tag, o
 }
 
 
+@(require_results)
 enum_string :: proc(a: any) -> string {
 	if a == nil { return "" }
 	ti := runtime.type_info_base(type_info_of(a.id))
@@ -591,6 +589,7 @@ enum_string :: proc(a: any) -> string {
 }
 
 // Given a enum type and a value name, get the enum value.
+@(require_results)
 enum_from_name :: proc($Enum_Type: typeid, name: string) -> (value: Enum_Type, ok: bool) {
 	ti := type_info_base(type_info_of(Enum_Type))
 	if eti, eti_ok := ti.variant.(runtime.Type_Info_Enum); eti_ok {
@@ -607,6 +606,7 @@ enum_from_name :: proc($Enum_Type: typeid, name: string) -> (value: Enum_Type, o
 	return
 }
 
+@(require_results)
 enum_from_name_any :: proc(Enum_Type: typeid, name: string) -> (value: Type_Info_Enum_Value, ok: bool) {
 	ti := runtime.type_info_base(type_info_of(Enum_Type))
 	if eti, eti_ok := ti.variant.(runtime.Type_Info_Enum); eti_ok {
@@ -623,6 +623,7 @@ enum_from_name_any :: proc(Enum_Type: typeid, name: string) -> (value: Type_Info
 }
 
 
+@(require_results)
 enum_field_names :: proc(Enum_Type: typeid) -> []string {
 	ti := runtime.type_info_base(type_info_of(Enum_Type))
 	if eti, eti_ok := ti.variant.(runtime.Type_Info_Enum); eti_ok {
@@ -630,6 +631,7 @@ enum_field_names :: proc(Enum_Type: typeid) -> []string {
 	}
 	return nil
 }
+@(require_results)
 enum_field_values :: proc(Enum_Type: typeid) -> []Type_Info_Enum_Value {
 	ti := runtime.type_info_base(type_info_of(Enum_Type))
 	if eti, eti_ok := ti.variant.(runtime.Type_Info_Enum); eti_ok {
@@ -643,6 +645,7 @@ Enum_Field :: struct {
 	value: Type_Info_Enum_Value,
 }
 
+@(require_results)
 enum_fields_zipped :: proc(Enum_Type: typeid) -> (fields: #soa[]Enum_Field) {
 	ti := runtime.type_info_base(type_info_of(Enum_Type))
 	if eti, eti_ok := ti.variant.(runtime.Type_Info_Enum); eti_ok {
@@ -653,15 +656,18 @@ enum_fields_zipped :: proc(Enum_Type: typeid) -> (fields: #soa[]Enum_Field) {
 
 
 
+@(require_results)
 union_variant_type_info :: proc(a: any) -> ^Type_Info {
 	id := union_variant_typeid(a)
 	return type_info_of(id)
 }
 
+@(require_results)
 type_info_union_is_pure_maybe :: proc(info: runtime.Type_Info_Union) -> bool {
 	return len(info.variants) == 1 && is_pointer(info.variants[0])
 }
 
+@(require_results)
 union_variant_typeid :: proc(a: any) -> typeid {
 	if a == nil { return nil }
 
@@ -690,9 +696,10 @@ union_variant_typeid :: proc(a: any) -> typeid {
 		case: unimplemented()
 		}
 
-		if a.data != nil && tag != 0 {
-			i := tag if info.no_nil else tag-1
-			return info.variants[i].id
+		if info.no_nil {
+			return info.variants[tag].id
+		} else if tag != 0 {
+			return info.variants[tag-1].id
 		}
 
 		return nil
@@ -700,6 +707,7 @@ union_variant_typeid :: proc(a: any) -> typeid {
 	panic("expected a union to reflect.union_variant_typeid")
 }
 
+@(require_results)
 get_union_variant_raw_tag :: proc(a: any) -> i64 {
 	if a == nil { return -1 }
 
@@ -730,6 +738,7 @@ get_union_variant_raw_tag :: proc(a: any) -> i64 {
 	panic("expected a union to reflect.get_union_variant_raw_tag")
 }
 
+@(require_results)
 get_union_variant :: proc(a: any) -> any {
 	if a == nil {
 		return nil
@@ -741,6 +750,7 @@ get_union_variant :: proc(a: any) -> any {
 	return any{a.data, id}
 }
 
+@(require_results)
 get_union_as_ptr_variants :: proc(val: ^$T) -> (res: intrinsics.type_convert_variants_to_pointers(T)) where intrinsics.type_is_union(T) {
 	ptr := rawptr(val)
 	tag := get_union_variant_raw_tag(val^)
@@ -881,6 +891,7 @@ set_union_value :: proc(dst: any, value: any) -> bool {
 
 
 
+@(require_results)
 as_bool :: proc(a: any) -> (value: bool, valid: bool) {
 	if a == nil { return }
 	a := a
@@ -903,6 +914,7 @@ as_bool :: proc(a: any) -> (value: bool, valid: bool) {
 	return
 }
 
+@(require_results)
 as_int :: proc(a: any) -> (value: int, valid: bool) {
 	v: i64
 	v, valid = as_i64(a)
@@ -910,6 +922,7 @@ as_int :: proc(a: any) -> (value: int, valid: bool) {
 	return
 }
 
+@(require_results)
 as_uint :: proc(a: any) -> (value: uint, valid: bool) {
 	v: u64
 	v, valid = as_u64(a)
@@ -917,6 +930,7 @@ as_uint :: proc(a: any) -> (value: uint, valid: bool) {
 	return
 }
 
+@(require_results)
 as_i64 :: proc(a: any) -> (value: i64, valid: bool) {
 	if a == nil { return }
 	a := a
@@ -1024,6 +1038,7 @@ as_i64 :: proc(a: any) -> (value: i64, valid: bool) {
 	return
 }
 
+@(require_results)
 as_u64 :: proc(a: any) -> (value: u64, valid: bool) {
 	if a == nil { return }
 	a := a
@@ -1133,6 +1148,7 @@ as_u64 :: proc(a: any) -> (value: u64, valid: bool) {
 }
 
 
+@(require_results)
 as_f64 :: proc(a: any) -> (value: f64, valid: bool) {
 	if a == nil { return }
 	a := a
@@ -1239,6 +1255,7 @@ as_f64 :: proc(a: any) -> (value: f64, valid: bool) {
 }
 
 
+@(require_results)
 as_string :: proc(a: any) -> (value: string, valid: bool) {
 	if a == nil { return }
 	a := a
@@ -1258,6 +1275,7 @@ as_string :: proc(a: any) -> (value: string, valid: bool) {
 	return
 }
 
+@(require_results)
 relative_pointer_to_absolute :: proc(a: any) -> rawptr {
 	if a == nil { return nil }
 	a := a
@@ -1272,6 +1290,7 @@ relative_pointer_to_absolute :: proc(a: any) -> rawptr {
 }
 
 
+@(require_results)
 relative_pointer_to_absolute_raw :: proc(data: rawptr, base_integer_id: typeid) -> rawptr {
 	_handle :: proc(ptr: ^$T) -> rawptr where intrinsics.type_is_integer(T) {
 		if ptr^ == 0 {
@@ -1314,6 +1333,7 @@ relative_pointer_to_absolute_raw :: proc(data: rawptr, base_integer_id: typeid)
 
 
 
+@(require_results)
 as_pointer :: proc(a: any) -> (value: rawptr, valid: bool) {
 	if a == nil { return }
 	a := a
@@ -1341,6 +1361,7 @@ as_pointer :: proc(a: any) -> (value: rawptr, valid: bool) {
 }
 
 
+@(require_results)
 as_raw_data :: proc(a: any) -> (value: rawptr, valid: bool) {
 	if a == nil { return }
 	a := a
@@ -1377,9 +1398,11 @@ ne :: not_equal
 
 DEFAULT_EQUAL_MAX_RECURSION_LEVEL :: 32
 
+@(require_results)
 not_equal :: proc(a, b: any, including_indirect_array_recursion := false, recursion_level := 0) -> bool {
 	return !equal(a, b, including_indirect_array_recursion, recursion_level)
 }
+@(require_results)
 equal :: proc(a, b: any, including_indirect_array_recursion := false, recursion_level := 0) -> bool {
 	if a == nil && b == nil {
 		return true
@@ -1416,7 +1439,7 @@ equal :: proc(a, b: any, including_indirect_array_recursion := false, recursion_
 	switch v in t.variant {
 	case Type_Info_Named:
 		unreachable()
-	case Type_Info_Tuple:
+	case Type_Info_Parameters:
 		unreachable()
 	case Type_Info_Any:
 		if !including_indirect_array_recursion {

core/reflect/types.odin

@@ -3,17 +3,16 @@ package reflect
 import "core:io"
 import "core:strings"
 
+@(require_results)
 are_types_identical :: proc(a, b: ^Type_Info) -> bool {
 	if a == b {
 		return true
 	}
 
-	if (a == nil && b != nil) ||
-	   (a != nil && b == nil) {
+	if a == nil || b == nil {
 		return false
 	}
 
-
 	switch {
 	case a.size != b.size, a.align != b.align:
 		return false
@@ -102,8 +101,8 @@ are_types_identical :: proc(a, b: ^Type_Info) -> bool {
 		y := b.variant.(Type_Info_Slice) or_return
 		return are_types_identical(x.elem, y.elem)
 
-	case Type_Info_Tuple:
-		y := b.variant.(Type_Info_Tuple) or_return
+	case Type_Info_Parameters:
+		y := b.variant.(Type_Info_Parameters) or_return
 		if len(x.types) != len(y.types) { return false }
 		for _, i in x.types {
 			xt, yt := x.types[i], y.types[i]
@@ -180,6 +179,7 @@ are_types_identical :: proc(a, b: ^Type_Info) -> bool {
 	return false
 }
 
+@(require_results)
 is_signed :: proc(info: ^Type_Info) -> bool {
 	if info == nil { return false }
 	#partial switch i in type_info_base(info).variant {
@@ -188,6 +188,7 @@ is_signed :: proc(info: ^Type_Info) -> bool {
 	}
 	return false
 }
+@(require_results)
 is_unsigned :: proc(info: ^Type_Info) -> bool {
 	if info == nil { return false }
 	#partial switch i in type_info_base(info).variant {
@@ -197,6 +198,7 @@ is_unsigned :: proc(info: ^Type_Info) -> bool {
 	return false
 }
 
+@(require_results)
 is_byte :: proc(info: ^Type_Info) -> bool {
 	if info == nil { return false }
 	#partial switch i in type_info_base(info).variant {
@@ -206,66 +208,79 @@ is_byte :: proc(info: ^Type_Info) -> bool {
 }
 
 
+@(require_results)
 is_integer :: proc(info: ^Type_Info) -> bool {
 	if info == nil { return false }
 	_, ok := type_info_base(info).variant.(Type_Info_Integer)
 	return ok
 }
+@(require_results)
 is_rune :: proc(info: ^Type_Info) -> bool {
 	if info == nil { return false }
 	_, ok := type_info_base(info).variant.(Type_Info_Rune)
 	return ok
 }
+@(require_results)
 is_float :: proc(info: ^Type_Info) -> bool {
 	if info == nil { return false }
 	_, ok := type_info_base(info).variant.(Type_Info_Float)
 	return ok
 }
+@(require_results)
 is_complex :: proc(info: ^Type_Info) -> bool {
 	if info == nil { return false }
 	_, ok := type_info_base(info).variant.(Type_Info_Complex)
 	return ok
 }
+@(require_results)
 is_quaternion :: proc(info: ^Type_Info) -> bool {
 	if info == nil { return false }
 	_, ok := type_info_base(info).variant.(Type_Info_Quaternion)
 	return ok
 }
+@(require_results)
 is_any :: proc(info: ^Type_Info) -> bool {
 	if info == nil { return false }
 	_, ok := type_info_base(info).variant.(Type_Info_Any)
 	return ok
 }
+@(require_results)
 is_string :: proc(info: ^Type_Info) -> bool {
 	if info == nil { return false }
 	_, ok := type_info_base(info).variant.(Type_Info_String)
 	return ok
 }
+@(require_results)
 is_cstring :: proc(info: ^Type_Info) -> bool {
 	if info == nil { return false }
 	v, ok := type_info_base(info).variant.(Type_Info_String)
 	return ok && v.is_cstring
 }
+@(require_results)
 is_boolean :: proc(info: ^Type_Info) -> bool {
 	if info == nil { return false }
 	_, ok := type_info_base(info).variant.(Type_Info_Boolean)
 	return ok
 }
+@(require_results)
 is_pointer :: proc(info: ^Type_Info) -> bool {
 	if info == nil { return false }
 	_, ok := type_info_base(info).variant.(Type_Info_Pointer)
 	return ok
 }
+@(require_results)
 is_multi_pointer :: proc(info: ^Type_Info) -> bool {
 	if info == nil { return false }
 	_, ok := type_info_base(info).variant.(Type_Info_Multi_Pointer)
 	return ok
 }
+@(require_results)
 is_soa_pointer :: proc(info: ^Type_Info) -> bool {
 	if info == nil { return false }
 	_, ok := type_info_base(info).variant.(Type_Info_Soa_Pointer)
 	return ok
 }
+@(require_results)
 is_pointer_internally :: proc(info: ^Type_Info) -> bool {
 	if info == nil { return false }
 	#partial switch v in info.variant {
@@ -277,76 +292,97 @@ is_pointer_internally :: proc(info: ^Type_Info) -> bool {
 	}
 	return false
 }
+@(require_results)
 is_procedure :: proc(info: ^Type_Info) -> bool {
 	if info == nil { return false }
 	_, ok := type_info_base(info).variant.(Type_Info_Procedure)
 	return ok
 }
+@(require_results)
 is_array :: proc(info: ^Type_Info) -> bool {
 	if info == nil { return false }
 	_, ok := type_info_base(info).variant.(Type_Info_Array)
 	return ok
 }
+@(require_results)
 is_enumerated_array :: proc(info: ^Type_Info) -> bool {
 	if info == nil { return false }
 	_, ok := type_info_base(info).variant.(Type_Info_Enumerated_Array)
 	return ok
 }
+@(require_results)
 is_dynamic_array :: proc(info: ^Type_Info) -> bool {
 	if info == nil { return false }
 	_, ok := type_info_base(info).variant.(Type_Info_Dynamic_Array)
 	return ok
 }
+@(require_results)
 is_dynamic_map :: proc(info: ^Type_Info) -> bool {
 	if info == nil { return false }
 	_, ok := type_info_base(info).variant.(Type_Info_Map)
 	return ok
 }
+@(require_results)
 is_bit_set :: proc(info: ^Type_Info) -> bool {
 	if info == nil { return false }
 	_, ok := type_info_base(info).variant.(Type_Info_Bit_Set)
 	return ok
 }
+@(require_results)
 is_slice :: proc(info: ^Type_Info) -> bool {
 	if info == nil { return false }
 	_, ok := type_info_base(info).variant.(Type_Info_Slice)
 	return ok
 }
-is_tuple :: proc(info: ^Type_Info) -> bool {
+@(require_results)
+is_parameters :: proc(info: ^Type_Info) -> bool {
 	if info == nil { return false }
-	_, ok := type_info_base(info).variant.(Type_Info_Tuple)
+	_, ok := type_info_base(info).variant.(Type_Info_Parameters)
 	return ok
 }
+@(require_results, deprecated="prefer is_parameters")
+is_tuple :: proc(info: ^Type_Info) -> bool {
+	if info == nil { return false }
+	_, ok := type_info_base(info).variant.(Type_Info_Parameters)
+	return ok
+}
+@(require_results)
 is_struct :: proc(info: ^Type_Info) -> bool {
 	if info == nil { return false }
 	s, ok := type_info_base(info).variant.(Type_Info_Struct)
 	return ok && !s.is_raw_union
 }
+@(require_results)
 is_raw_union :: proc(info: ^Type_Info) -> bool {
 	if info == nil { return false }
 	s, ok := type_info_base(info).variant.(Type_Info_Struct)
 	return ok && s.is_raw_union
 }
+@(require_results)
 is_union :: proc(info: ^Type_Info) -> bool {
 	if info == nil { return false }
 	_, ok := type_info_base(info).variant.(Type_Info_Union)
 	return ok
 }
+@(require_results)
 is_enum :: proc(info: ^Type_Info) -> bool {
 	if info == nil { return false }
 	_, ok := type_info_base(info).variant.(Type_Info_Enum)
 	return ok
 }
+@(require_results)
 is_simd_vector :: proc(info: ^Type_Info) -> bool {
 	if info == nil { return false }
 	_, ok := type_info_base(info).variant.(Type_Info_Simd_Vector)
 	return ok
 }
+@(require_results)
 is_relative_pointer :: proc(info: ^Type_Info) -> bool {
 	if info == nil { return false }
 	_, ok := type_info_base(info).variant.(Type_Info_Relative_Pointer)
 	return ok
 }
+@(require_results)
 is_relative_slice :: proc(info: ^Type_Info) -> bool {
 	if info == nil { return false }
 	_, ok := type_info_base(info).variant.(Type_Info_Relative_Slice)
@@ -460,7 +496,7 @@ write_type_writer :: proc(w: io.Writer, ti: ^Type_Info, n_written: ^int = nil) -
 		if info.params == nil {
 			io.write_string(w, "()", &n) or_return
 		} else {
-			t := info.params.variant.(Type_Info_Tuple)
+			t := info.params.variant.(Type_Info_Parameters)
 			io.write_string(w, "(", &n) or_return
 			for t, i in t.types {
 				if i > 0 {
@@ -474,7 +510,7 @@ write_type_writer :: proc(w: io.Writer, ti: ^Type_Info, n_written: ^int = nil) -
 			io.write_string(w, " -> ", &n)  or_return
 			write_type(w, info.results, &n) or_return
 		}
-	case Type_Info_Tuple:
+	case Type_Info_Parameters:
 		count := len(info.names)
 		if count != 1 { 
 			io.write_string(w, "(", &n) or_return 

core/runtime/core.odin

@@ -83,8 +83,8 @@ Type_Info_Multi_Pointer :: struct {
 	elem: ^Type_Info,
 }
 Type_Info_Procedure :: struct {
-	params:     ^Type_Info, // Type_Info_Tuple
-	results:    ^Type_Info, // Type_Info_Tuple
+	params:     ^Type_Info, // Type_Info_Parameters
+	results:    ^Type_Info, // Type_Info_Parameters
 	variadic:   bool,
 	convention: Calling_Convention,
 }
@@ -104,10 +104,12 @@ Type_Info_Enumerated_Array :: struct {
 }
 Type_Info_Dynamic_Array :: struct {elem: ^Type_Info, elem_size: int}
 Type_Info_Slice         :: struct {elem: ^Type_Info, elem_size: int}
-Type_Info_Tuple :: struct { // Only used for procedures parameters and results
+
+Type_Info_Parameters :: struct { // Only used for procedures parameters and results
 	types:        []^Type_Info,
 	names:        []string,
 }
+Type_Info_Tuple :: Type_Info_Parameters // Will be removed eventually
 
 Type_Info_Struct :: struct {
 	types:        []^Type_Info,
@@ -208,7 +210,7 @@ Type_Info :: struct {
 		Type_Info_Enumerated_Array,
 		Type_Info_Dynamic_Array,
 		Type_Info_Slice,
-		Type_Info_Tuple,
+		Type_Info_Parameters,
 		Type_Info_Struct,
 		Type_Info_Union,
 		Type_Info_Enum,
@@ -329,6 +331,12 @@ Allocator :: struct {
 	data:      rawptr,
 }
 
+Byte     :: 1
+Kilobyte :: 1024 * Byte
+Megabyte :: 1024 * Kilobyte
+Gigabyte :: 1024 * Megabyte
+Terabyte :: 1024 * Gigabyte
+
 // Logging stuff
 
 Logger_Level :: enum uint {
@@ -499,11 +507,8 @@ Odin_Endian_Type :: type_of(ODIN_ENDIAN)
 foreign {
 	@(link_name="__$startup_runtime")
 	_startup_runtime :: proc "odin" () ---
-}
-
-@(link_name="__$cleanup_runtime")
-_cleanup_runtime :: proc() {
-	default_temp_allocator_destroy(&global_default_temp_allocator_data)
+	@(link_name="__$cleanup_runtime")
+	_cleanup_runtime :: proc "odin" () ---
 }
 
 _cleanup_runtime_contextless :: proc "contextless" () {

core/runtime/core_builtin.odin

@@ -231,13 +231,12 @@ make_dynamic_array_len_cap :: proc($T: typeid/[dynamic]$E, #any_int len: int, #a
 	return
 }
 @(builtin)
-make_map :: proc($T: typeid/map[$K]$E, #any_int capacity: int = 1<<MAP_MIN_LOG2_CAPACITY, allocator := context.allocator, loc := #caller_location) -> T {
+make_map :: proc($T: typeid/map[$K]$E, #any_int capacity: int = 1<<MAP_MIN_LOG2_CAPACITY, allocator := context.allocator, loc := #caller_location) -> (m: T, err: Allocator_Error) #optional_allocator_error {
 	make_map_expr_error_loc(loc, capacity)
 	context.allocator = allocator
 
-	m: T
-	reserve_map(&m, capacity, loc)
-	return m
+	err = reserve_map(&m, capacity, loc)
+	return
 }
 @(builtin)
 make_multi_pointer :: proc($T: typeid/[^]$E, #any_int len: int, allocator := context.allocator, loc := #caller_location) -> (mp: T, err: Allocator_Error) #optional_allocator_error {
@@ -276,10 +275,8 @@ clear_map :: proc "contextless" (m: ^$T/map[$K]$V) {
 }
 
 @builtin
-reserve_map :: proc(m: ^$T/map[$K]$V, capacity: int, loc := #caller_location) {
-	if m != nil {
-		__dynamic_map_reserve((^Raw_Map)(m), map_info(T), uint(capacity), loc)
-	}
+reserve_map :: proc(m: ^$T/map[$K]$V, capacity: int, loc := #caller_location) -> Allocator_Error {
+	return __dynamic_map_reserve((^Raw_Map)(m), map_info(T), uint(capacity), loc) if m != nil else nil
 }
 
 /*
@@ -615,7 +612,7 @@ shrink_dynamic_array :: proc(array: ^$T/[dynamic]$E, new_cap := -1, loc := #call
 	old_size := a.cap * size_of(E)
 	new_size := new_cap * size_of(E)
 
-	new_data, err := mem_resize(a.data, old_size, new_size, align_of(E), allocator, loc)
+	new_data, err := mem_resize(a.data, old_size, new_size, align_of(E), a.allocator, loc)
 	if err != nil {
 		return
 	}

core/runtime/core_builtin_soa.odin

@@ -150,14 +150,14 @@ make_soa_dynamic_array :: proc($T: typeid/#soa[dynamic]$E, allocator := context.
 }
 
 @builtin
-make_soa_dynamic_array_len :: proc($T: typeid/#soa[dynamic]$E, auto_cast length: int, allocator := context.allocator, loc := #caller_location) -> (array: T) {
+make_soa_dynamic_array_len :: proc($T: typeid/#soa[dynamic]$E, #any_int length: int, allocator := context.allocator, loc := #caller_location) -> (array: T) {
 	context.allocator = allocator
 	resize_soa(&array, length, loc)
 	return
 }
 
 @builtin
-make_soa_dynamic_array_len_cap :: proc($T: typeid/#soa[dynamic]$E, auto_cast length, capacity: int, allocator := context.allocator, loc := #caller_location) -> (array: T) {
+make_soa_dynamic_array_len_cap :: proc($T: typeid/#soa[dynamic]$E, #any_int length, capacity: int, allocator := context.allocator, loc := #caller_location) -> (array: T) {
 	context.allocator = allocator
 	if reserve_soa(&array, capacity, loc) {
 		resize_soa(&array, length, loc)

core/runtime/default_allocators_arena.odin

@@ -0,0 +1,304 @@
+package runtime
+
+import "core:intrinsics"
+
+DEFAULT_ARENA_GROWING_MINIMUM_BLOCK_SIZE :: uint(DEFAULT_TEMP_ALLOCATOR_BACKING_SIZE)
+
+Memory_Block :: struct {
+	prev:      ^Memory_Block,
+	allocator: Allocator,
+	base:      [^]byte,
+	used:      uint,
+	capacity:  uint,
+}
+
+Arena :: struct {
+	backing_allocator:  Allocator,
+	curr_block:         ^Memory_Block,
+	total_used:         uint,
+	total_capacity:     uint,
+	minimum_block_size: uint,
+	temp_count:         uint,
+}
+
+@(private, require_results)
+safe_add :: #force_inline proc "contextless" (x, y: uint) -> (uint, bool) {
+	z, did_overflow := intrinsics.overflow_add(x, y)
+	return z, !did_overflow
+}
+
+@(require_results)
+memory_block_alloc :: proc(allocator: Allocator, capacity: uint, loc := #caller_location) -> (block: ^Memory_Block, err: Allocator_Error) {
+	total_size := uint(capacity + size_of(Memory_Block))
+	base_offset    := uintptr(size_of(Memory_Block))
+
+	min_alignment: int = max(16, align_of(Memory_Block))
+	data := mem_alloc(int(total_size), min_alignment, allocator, loc) or_return
+	block = (^Memory_Block)(raw_data(data))
+	end := uintptr(raw_data(data)[len(data):])
+
+	block.allocator = allocator
+	block.base = ([^]byte)(uintptr(block) + base_offset)
+	block.capacity = uint(end - uintptr(block.base))
+
+	// Should be zeroed
+	assert(block.used == 0)
+	assert(block.prev == nil)
+	return
+}
+
+memory_block_dealloc :: proc(block_to_free: ^Memory_Block, loc := #caller_location) {
+	if block_to_free != nil {
+		allocator := block_to_free.allocator
+		mem_free(block_to_free, allocator, loc)
+	}
+}
+
+@(require_results)
+alloc_from_memory_block :: proc(block: ^Memory_Block, min_size, alignment: uint) -> (data: []byte, err: Allocator_Error) {
+	calc_alignment_offset :: proc "contextless" (block: ^Memory_Block, alignment: uintptr) -> uint {
+		alignment_offset := uint(0)
+		ptr := uintptr(block.base[block.used:])
+		mask := alignment-1
+		if ptr & mask != 0 {
+			alignment_offset = uint(alignment - (ptr & mask))
+		}
+		return alignment_offset
+
+	}
+	if block == nil {
+		return nil, .Out_Of_Memory
+	}
+	alignment_offset := calc_alignment_offset(block, uintptr(alignment))
+	size, size_ok := safe_add(min_size, alignment_offset)
+	if !size_ok {
+		err = .Out_Of_Memory
+		return
+	}
+
+	if to_be_used, ok := safe_add(block.used, size); !ok || to_be_used > block.capacity {
+		err = .Out_Of_Memory
+		return
+	}
+	data = block.base[block.used+alignment_offset:][:min_size]
+	block.used += size
+	return
+}
+
+@(require_results)
+arena_alloc :: proc(arena: ^Arena, size, alignment: uint, loc := #caller_location) -> (data: []byte, err: Allocator_Error) {
+	align_forward_uint :: proc "contextless" (ptr, align: uint) -> uint {
+		p := ptr
+		modulo := p & (align-1)
+		if modulo != 0 {
+			p += align - modulo
+		}
+		return p
+	}
+
+	assert(alignment & (alignment-1) == 0, "non-power of two alignment", loc)
+
+	size := size
+	if size == 0 {
+		return
+	}
+
+	if arena.curr_block == nil || (safe_add(arena.curr_block.used, size) or_else 0) > arena.curr_block.capacity {
+		size = align_forward_uint(size, alignment)
+		if arena.minimum_block_size == 0 {
+			arena.minimum_block_size = DEFAULT_ARENA_GROWING_MINIMUM_BLOCK_SIZE
+		}
+
+		block_size := max(size, arena.minimum_block_size)
+
+		if arena.backing_allocator.procedure == nil {
+			arena.backing_allocator = default_allocator()
+		}
+
+		new_block := memory_block_alloc(arena.backing_allocator, block_size, loc) or_return
+		new_block.prev = arena.curr_block
+		arena.curr_block = new_block
+		arena.total_capacity += new_block.capacity
+	}
+
+	prev_used := arena.curr_block.used
+	data, err = alloc_from_memory_block(arena.curr_block, size, alignment)
+	arena.total_used += arena.curr_block.used - prev_used
+	return
+}
+
+// `arena_init` will initialize the arena with a usuable block.
+// This procedure is not necessary to use the Arena as the default zero as `arena_alloc` will set things up if necessary
+@(require_results)
+arena_init :: proc(arena: ^Arena, size: uint, backing_allocator: Allocator, loc := #caller_location) -> Allocator_Error {
+	arena^ = {}
+	arena.backing_allocator = backing_allocator
+	arena.minimum_block_size = max(size, 1<<12) // minimum block size of 4 KiB
+	new_block := memory_block_alloc(arena.backing_allocator, arena.minimum_block_size, loc) or_return
+	arena.curr_block = new_block
+	arena.total_capacity += new_block.capacity
+	return nil
+}
+
+
+arena_free_last_memory_block :: proc(arena: ^Arena, loc := #caller_location) {
+	if free_block := arena.curr_block; free_block != nil {
+		arena.curr_block = free_block.prev
+
+		arena.total_capacity -= free_block.capacity
+		memory_block_dealloc(free_block, loc)
+	}
+}
+
+// `arena_free_all` will free all but the first memory block, and then reset the memory block
+arena_free_all :: proc(arena: ^Arena, loc := #caller_location) {
+	for arena.curr_block != nil && arena.curr_block.prev != nil {
+		arena_free_last_memory_block(arena, loc)
+	}
+
+	if arena.curr_block != nil {
+		intrinsics.mem_zero(arena.curr_block.base, arena.curr_block.used)
+		arena.curr_block.used = 0
+	}
+	arena.total_used = 0
+}
+
+arena_destroy :: proc(arena: ^Arena, loc := #caller_location) {
+	for arena.curr_block != nil {
+		free_block := arena.curr_block
+		arena.curr_block = free_block.prev
+
+		arena.total_capacity -= free_block.capacity
+		memory_block_dealloc(free_block, loc)
+	}
+	arena.total_used = 0
+	arena.total_capacity = 0
+}
+
+arena_allocator :: proc(arena: ^Arena) -> Allocator {
+	return Allocator{arena_allocator_proc, arena}
+}
+
+arena_allocator_proc :: proc(allocator_data: rawptr, mode: Allocator_Mode,
+                             size, alignment: int,
+                             old_memory: rawptr, old_size: int,
+                             location := #caller_location) -> (data: []byte, err: Allocator_Error) {
+	arena := (^Arena)(allocator_data)
+
+	size, alignment := uint(size), uint(alignment)
+	old_size := uint(old_size)
+
+	switch mode {
+	case .Alloc, .Alloc_Non_Zeroed:
+		return arena_alloc(arena, size, alignment, location)
+	case .Free:
+		err = .Mode_Not_Implemented
+	case .Free_All:
+		arena_free_all(arena, location)
+	case .Resize:
+		old_data := ([^]byte)(old_memory)
+
+		switch {
+		case old_data == nil:
+			return arena_alloc(arena, size, alignment, location)
+		case size == old_size:
+			// return old memory
+			data = old_data[:size]
+			return
+		case size == 0:
+			err = .Mode_Not_Implemented
+			return
+		case (uintptr(old_data) & uintptr(alignment-1) == 0) && size < old_size:
+			// shrink data in-place
+			data = old_data[:size]
+			return
+		}
+
+		new_memory := arena_alloc(arena, size, alignment, location) or_return
+		if new_memory == nil {
+			return
+		}
+		copy(new_memory, old_data[:old_size])
+		return new_memory, nil
+	case .Query_Features:
+		set := (^Allocator_Mode_Set)(old_memory)
+		if set != nil {
+			set^ = {.Alloc, .Alloc_Non_Zeroed, .Free_All, .Resize, .Query_Features}
+		}
+	case .Query_Info:
+		err = .Mode_Not_Implemented
+	}
+
+	return
+}
+
+
+
+
+Arena_Temp :: struct {
+	arena: ^Arena,
+	block: ^Memory_Block,
+	used:  uint,
+}
+
+@(require_results)
+arena_temp_begin :: proc(arena: ^Arena, loc := #caller_location) -> (temp: Arena_Temp) {
+	assert(arena != nil, "nil arena", loc)
+
+	temp.arena = arena
+	temp.block = arena.curr_block
+	if arena.curr_block != nil {
+		temp.used = arena.curr_block.used
+	}
+	arena.temp_count += 1
+	return
+}
+
+arena_temp_end :: proc(temp: Arena_Temp, loc := #caller_location) {
+	if temp.arena == nil {
+		assert(temp.block == nil)
+		assert(temp.used == 0)
+		return
+	}
+	arena := temp.arena
+
+	if temp.block != nil {
+		memory_block_found := false
+		for block := arena.curr_block; block != nil; block = block.prev {
+			if block == temp.block {
+				memory_block_found = true
+				break
+			}
+		}
+		if !memory_block_found {
+			assert(arena.curr_block == temp.block, "memory block stored within Arena_Temp not owned by Arena", loc)
+		}
+
+		for arena.curr_block != temp.block {
+			arena_free_last_memory_block(arena)
+		}
+
+		if block := arena.curr_block; block != nil {
+			assert(block.used >= temp.used, "out of order use of arena_temp_end", loc)
+			amount_to_zero := min(block.used-temp.used, block.capacity-block.used)
+			intrinsics.mem_zero(block.base[temp.used:], amount_to_zero)
+			block.used = temp.used
+		}
+	}
+
+	assert(arena.temp_count > 0, "double-use of arena_temp_end", loc)
+	arena.temp_count -= 1
+}
+
+// Ignore the use of a `arena_temp_begin` entirely
+arena_temp_ignore :: proc(temp: Arena_Temp, loc := #caller_location) {
+	assert(temp.arena != nil, "nil arena", loc)
+	arena := temp.arena
+
+	assert(arena.temp_count > 0, "double-use of arena_temp_end", loc)
+	arena.temp_count -= 1
+}
+
+arena_check_temp :: proc(arena: ^Arena, loc := #caller_location) {
+	assert(arena.temp_count == 0, "Arena_Temp not been ended", loc)
+}

core/runtime/default_temporary_allocator.odin

@@ -1,159 +1,38 @@
 package runtime
 
-DEFAULT_TEMP_ALLOCATOR_BACKING_SIZE: int : #config(DEFAULT_TEMP_ALLOCATOR_BACKING_SIZE, 1<<22)
+DEFAULT_TEMP_ALLOCATOR_BACKING_SIZE: int : #config(DEFAULT_TEMP_ALLOCATOR_BACKING_SIZE, 4 * Megabyte)
 
 
 when ODIN_OS == .Freestanding || ODIN_OS == .JS || ODIN_DEFAULT_TO_NIL_ALLOCATOR {
 	Default_Temp_Allocator :: struct {}
 	
-	default_temp_allocator_init :: proc(s: ^Default_Temp_Allocator, size: int, backup_allocator := context.allocator) {}
+	default_temp_allocator_init :: proc(s: ^Default_Temp_Allocator, size: int, backing_allocator := context.allocator) {}
 	
 	default_temp_allocator_destroy :: proc(s: ^Default_Temp_Allocator) {}
 	
 	default_temp_allocator_proc :: nil_allocator_proc
+
+	@(require_results)
+	default_temp_allocator_temp_begin :: proc(loc := #caller_location) -> (temp: Arena_Temp) {
+		return
+	}
+
+	default_temp_allocator_temp_end :: proc(temp: Arena_Temp, loc := #caller_location) {
+	}
 } else {
 	Default_Temp_Allocator :: struct {
-		data:               []byte,
-		curr_offset:        int,
-		prev_allocation:    rawptr,
-		backup_allocator:   Allocator,
-		leaked_allocations: [dynamic][]byte,
+		arena: Arena,
 	}
 	
-	default_temp_allocator_init :: proc(s: ^Default_Temp_Allocator, size: int, backup_allocator := context.allocator) {
-		s.data = make_aligned([]byte, size, 2*align_of(rawptr), backup_allocator)
-		s.curr_offset = 0
-		s.prev_allocation = nil
-		s.backup_allocator = backup_allocator
-		s.leaked_allocations.allocator = backup_allocator
+	default_temp_allocator_init :: proc(s: ^Default_Temp_Allocator, size: int, backing_allocator := context.allocator) {
+		_ = arena_init(&s.arena, uint(size), backing_allocator)
 	}
 
 	default_temp_allocator_destroy :: proc(s: ^Default_Temp_Allocator) {
-		if s == nil {
-			return
+		if s != nil {
+			arena_destroy(&s.arena)
+			s^ = {}
 		}
-		for ptr in s.leaked_allocations {
-			free(raw_data(ptr), s.backup_allocator)
-		}
-		delete(s.leaked_allocations)
-		delete(s.data, s.backup_allocator)
-		s^ = {}
-	}
-
-	@(private)
-	default_temp_allocator_alloc :: proc(s: ^Default_Temp_Allocator, size, alignment: int, loc := #caller_location) -> ([]byte, Allocator_Error) {
-		size := size
-		size = align_forward_int(size, alignment)
-
-		switch {
-		case s.curr_offset+size <= len(s.data):
-			start := uintptr(raw_data(s.data))
-			ptr := start + uintptr(s.curr_offset)
-			ptr = align_forward_uintptr(ptr, uintptr(alignment))
-			mem_zero(rawptr(ptr), size)
-
-			s.prev_allocation = rawptr(ptr)
-			offset := int(ptr - start)
-			s.curr_offset = offset + size
-			return byte_slice(rawptr(ptr), size), .None
-
-		case size <= len(s.data):
-			start := uintptr(raw_data(s.data))
-			ptr := align_forward_uintptr(start, uintptr(alignment))
-			mem_zero(rawptr(ptr), size)
-
-			s.prev_allocation = rawptr(ptr)
-			offset := int(ptr - start)
-			s.curr_offset = offset + size
-			return byte_slice(rawptr(ptr), size), .None
-		}
-		a := s.backup_allocator
-		if a.procedure == nil {
-			a = context.allocator
-			s.backup_allocator = a
-		}
-
-		data, err := mem_alloc_bytes(size, alignment, a, loc)
-		if err != nil {
-			return data, err
-		}
-		if s.leaked_allocations == nil {
-			s.leaked_allocations = make([dynamic][]byte, a)
-		}
-		append(&s.leaked_allocations, data)
-
-		// TODO(bill): Should leaks be notified about?
-		if logger := context.logger; logger.lowest_level <= .Warning {
-			if logger.procedure != nil {
-				logger.procedure(logger.data, .Warning, "default temp allocator resorted to backup_allocator" , logger.options, loc)
-			}
-		}
-
-		return data, .None
-	}
-
-	@(private)
-	default_temp_allocator_free :: proc(s: ^Default_Temp_Allocator, old_memory: rawptr, loc := #caller_location) -> Allocator_Error {
-		if old_memory == nil {
-			return .None
-		}
-
-		start := uintptr(raw_data(s.data))
-		end := start + uintptr(len(s.data))
-		old_ptr := uintptr(old_memory)
-
-		if s.prev_allocation == old_memory {
-			s.curr_offset = int(uintptr(s.prev_allocation) - start)
-			s.prev_allocation = nil
-			return .None
-		}
-
-		if start <= old_ptr && old_ptr < end {
-			// NOTE(bill): Cannot free this pointer but it is valid
-			return .None
-		}
-
-		if len(s.leaked_allocations) != 0 {
-			for data, i in s.leaked_allocations {
-				ptr := raw_data(data)
-				if ptr == old_memory {
-					free(ptr, s.backup_allocator)
-					ordered_remove(&s.leaked_allocations, i)
-					return .None
-				}
-			}
-		}
-		return .Invalid_Pointer
-		// panic("invalid pointer passed to default_temp_allocator");
-	}
-
-	@(private)
-	default_temp_allocator_free_all :: proc(s: ^Default_Temp_Allocator, loc := #caller_location) {
-		s.curr_offset = 0
-		s.prev_allocation = nil
-		for data in s.leaked_allocations {
-			free(raw_data(data), s.backup_allocator)
-		}
-		clear(&s.leaked_allocations)
-	}
-
-	@(private)
-	default_temp_allocator_resize :: proc(s: ^Default_Temp_Allocator, old_memory: rawptr, old_size, size, alignment: int, loc := #caller_location) -> ([]byte, Allocator_Error) {
-		begin := uintptr(raw_data(s.data))
-		end := begin + uintptr(len(s.data))
-		old_ptr := uintptr(old_memory)
-		if old_memory == s.prev_allocation && old_ptr & uintptr(alignment)-1 == 0 {
-			if old_ptr+uintptr(size) < end {
-				s.curr_offset = int(old_ptr-begin)+size
-				return byte_slice(old_memory, size), .None
-			}
-		}
-		data, err := default_temp_allocator_alloc(s, size, alignment, loc)
-		if err == .None {
-			copy(data, byte_slice(old_memory, old_size))
-			err = default_temp_allocator_free(s, old_memory, loc)
-		}
-		return data, err
 	}
 
 	default_temp_allocator_proc :: proc(allocator_data: rawptr, mode: Allocator_Mode,
@@ -161,40 +40,40 @@ when ODIN_OS == .Freestanding || ODIN_OS == .JS || ODIN_DEFAULT_TO_NIL_ALLOCATOR
 	                                    old_memory: rawptr, old_size: int, loc := #caller_location) -> (data: []byte, err: Allocator_Error) {
 
 		s := (^Default_Temp_Allocator)(allocator_data)
+		return arena_allocator_proc(&s.arena, mode, size, alignment, old_memory, old_size, loc)
+	}
 
-		if s.data == nil {
-			default_temp_allocator_init(s, DEFAULT_TEMP_ALLOCATOR_BACKING_SIZE, default_allocator())
+	@(require_results)
+	default_temp_allocator_temp_begin :: proc(loc := #caller_location) -> (temp: Arena_Temp) {
+		if context.temp_allocator.data == &global_default_temp_allocator_data {
+			temp = arena_temp_begin(&global_default_temp_allocator_data.arena, loc)
 		}
-
-		switch mode {
-		case .Alloc, .Alloc_Non_Zeroed:
-			data, err = default_temp_allocator_alloc(s, size, alignment, loc)
-		case .Free:
-			err = default_temp_allocator_free(s, old_memory, loc)
-
-		case .Free_All:
-			default_temp_allocator_free_all(s, loc)
-
-		case .Resize:
-			data, err = default_temp_allocator_resize(s, old_memory, old_size, size, alignment, loc)
-
-		case .Query_Features:
-			set := (^Allocator_Mode_Set)(old_memory)
-			if set != nil {
-				set^ = {.Alloc, .Alloc_Non_Zeroed, .Free, .Free_All, .Resize, .Query_Features}
-			}
-
-		case .Query_Info:
-			return nil, .Mode_Not_Implemented
-		}
-
 		return
 	}
+
+	default_temp_allocator_temp_end :: proc(temp: Arena_Temp, loc := #caller_location) {
+		arena_temp_end(temp, loc)
+	}
 }
 
+@(deferred_out=default_temp_allocator_temp_end)
+DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD :: #force_inline proc(ignore := false, loc := #caller_location) -> (Arena_Temp, Source_Code_Location) {
+	if ignore {
+		return {}, loc
+	} else {
+		return default_temp_allocator_temp_begin(loc), loc
+	}
+}
+
+
 default_temp_allocator :: proc(allocator: ^Default_Temp_Allocator) -> Allocator {
 	return Allocator{
 		procedure = default_temp_allocator_proc,
-		data = allocator,
+		data      = allocator,
 	}
-}
+}
+
+@(fini, private)
+_destroy_temp_allocator_fini :: proc() {
+	default_temp_allocator_destroy(&global_default_temp_allocator_data)
+}

core/runtime/internal.odin

@@ -184,32 +184,33 @@ mem_free_all :: #force_inline proc(allocator := context.allocator, loc := #calle
 	return
 }
 
-mem_resize :: proc(ptr: rawptr, old_size, new_size: int, alignment: int = DEFAULT_ALIGNMENT, allocator := context.allocator, loc := #caller_location) -> ([]byte, Allocator_Error) {
+mem_resize :: proc(ptr: rawptr, old_size, new_size: int, alignment: int = DEFAULT_ALIGNMENT, allocator := context.allocator, loc := #caller_location) -> (data: []byte, err: Allocator_Error) {
 	if allocator.procedure == nil {
 		return nil, nil
 	}
 	if new_size == 0 {
 		if ptr != nil {
-			_, err := allocator.procedure(allocator.data, .Free, 0, 0, ptr, old_size, loc)
-			return nil, err
+			_, err = allocator.procedure(allocator.data, .Free, 0, 0, ptr, old_size, loc)
+			return
 		}
-		return nil, nil
+		return
 	} else if ptr == nil {
 		return allocator.procedure(allocator.data, .Alloc, new_size, alignment, nil, 0, loc)
 	} else if old_size == new_size && uintptr(ptr) % uintptr(alignment) == 0 {
-		return ([^]byte)(ptr)[:old_size], nil
+		data = ([^]byte)(ptr)[:old_size]
+		return
 	}
 
-	data, err := allocator.procedure(allocator.data, .Resize, new_size, alignment, ptr, old_size, loc)
+	data, err = allocator.procedure(allocator.data, .Resize, new_size, alignment, ptr, old_size, loc)
 	if err == .Mode_Not_Implemented {
 		data, err = allocator.procedure(allocator.data, .Alloc, new_size, alignment, nil, 0, loc)
 		if err != nil {
-			return data, err
+			return
 		}
 		copy(data, ([^]byte)(ptr)[:old_size])
 		_, err = allocator.procedure(allocator.data, .Free, 0, 0, ptr, old_size, loc)
 	}
-	return data, err
+	return
 }
 
 memory_equal :: proc "contextless" (x, y: rawptr, n: int) -> bool {
@@ -223,7 +224,7 @@ memory_equal :: proc "contextless" (x, y: rawptr, n: int) -> bool {
 	
 	when size_of(uint) == 8 {
 		if word_length := length >> 3; word_length != 0 {
-			for i in 0..<word_length {
+			for _ in 0..<word_length {
 				if intrinsics.unaligned_load((^u64)(a)) != intrinsics.unaligned_load((^u64)(b)) {
 					return false
 				}
@@ -254,7 +255,7 @@ memory_equal :: proc "contextless" (x, y: rawptr, n: int) -> bool {
 		return true
 	} else {
 		if word_length := length >> 2; word_length != 0 {
-			for i in 0..<word_length {
+			for _ in 0..<word_length {
 				if intrinsics.unaligned_load((^u32)(a)) != intrinsics.unaligned_load((^u32)(b)) {
 					return false
 				}

core/runtime/print.odin

@@ -2,6 +2,76 @@ package runtime
 
 _INTEGER_DIGITS :: "0123456789abcdefghijklmnopqrstuvwxyz"
 
+@(private="file")
+_INTEGER_DIGITS_VAR := _INTEGER_DIGITS
+
+when !ODIN_DISALLOW_RTTI {
+	print_any_single :: proc(arg: any) {
+		x := arg
+		if loc, ok := x.(Source_Code_Location); ok {
+			print_caller_location(loc)
+			return
+		}
+		x.id = typeid_base(x.id)
+		switch v in x {
+		case typeid:     print_typeid(v)
+		case ^Type_Info: print_type(v)
+
+		case string:  print_string(v)
+		case cstring: print_string(string(v))
+		case []byte:  print_string(string(v))
+
+		case rune:  print_rune(v)
+
+		case u8:    print_u64(u64(v))
+		case u16:   print_u64(u64(v))
+		case u16le: print_u64(u64(v))
+		case u16be: print_u64(u64(v))
+		case u32:   print_u64(u64(v))
+		case u32le: print_u64(u64(v))
+		case u32be: print_u64(u64(v))
+		case u64:   print_u64(u64(v))
+		case u64le: print_u64(u64(v))
+		case u64be: print_u64(u64(v))
+
+		case i8:    print_i64(i64(v))
+		case i16:   print_i64(i64(v))
+		case i16le: print_i64(i64(v))
+		case i16be: print_i64(i64(v))
+		case i32:   print_i64(i64(v))
+		case i32le: print_i64(i64(v))
+		case i32be: print_i64(i64(v))
+		case i64:   print_i64(i64(v))
+		case i64le: print_i64(i64(v))
+		case i64be: print_i64(i64(v))
+
+		case int:     print_int(v)
+		case uint:    print_uint(v)
+		case uintptr: print_uintptr(v)
+
+		case:
+			ti := type_info_of(x.id)
+			#partial switch v in ti.variant {
+			case Type_Info_Pointer:
+				print_uintptr((^uintptr)(x.data)^)
+				return
+			}
+
+			print_string("<invalid-value>")
+		}
+	}
+	println_any :: proc(args: ..any) {
+		loop: for arg, i in args {
+			if i != 0 {
+				print_string(" ")
+			}
+			print_any_single(arg)
+		}
+		print_string("\n")
+	}
+}
+
+
 encode_rune :: proc "contextless" (c: rune) -> ([4]u8, int) {
 	r := c
 
@@ -38,14 +108,14 @@ encode_rune :: proc "contextless" (c: rune) -> ([4]u8, int) {
 	return buf, 4
 }
 
-print_string :: proc "contextless" (str: string) -> (int, _OS_Errno) {
-	return os_write(transmute([]byte)str)
+print_string :: proc "contextless" (str: string) -> (n: int) {
+	n, _ = os_write(transmute([]byte)str)
+	return
 }
 
-print_strings :: proc "contextless" (args: ..string) -> (n: int, err: _OS_Errno) {
+print_strings :: proc "contextless" (args: ..string) -> (n: int) {
 	for str in args {
-		m: int
-		m, err = os_write(transmute([]byte)str)
+		m, err := os_write(transmute([]byte)str)
 		n += m
 		if err != 0 {
 			break
@@ -54,8 +124,9 @@ print_strings :: proc "contextless" (args: ..string) -> (n: int, err: _OS_Errno)
 	return
 }
 
-print_byte :: proc "contextless" (b: byte) -> (int, _OS_Errno) {
-	return os_write([]byte{b})
+print_byte :: proc "contextless" (b: byte) -> (n: int) {
+	n, _ = os_write([]byte{b})
+	return
 }
 
 print_encoded_rune :: proc "contextless" (r: rune) {
@@ -74,11 +145,10 @@ print_encoded_rune :: proc "contextless" (r: rune) {
 		if r <= 0 {
 			print_string("\\x00")
 		} else if r < 32 {
-			digits := _INTEGER_DIGITS
 			n0, n1 := u8(r) >> 4, u8(r) & 0xf
 			print_string("\\x")
-			print_byte(digits[n0])
-			print_byte(digits[n1])
+			print_byte(_INTEGER_DIGITS_VAR[n0])
+			print_byte(_INTEGER_DIGITS_VAR[n1])
 		} else {
 			print_rune(r)
 		}
@@ -86,7 +156,7 @@ print_encoded_rune :: proc "contextless" (r: rune) {
 	print_byte('\'')
 }
 
-print_rune :: proc "contextless" (r: rune) -> (int, _OS_Errno) #no_bounds_check {
+print_rune :: proc "contextless" (r: rune) -> int #no_bounds_check {
 	RUNE_SELF :: 0x80
 
 	if r < RUNE_SELF {
@@ -94,29 +164,27 @@ print_rune :: proc "contextless" (r: rune) -> (int, _OS_Errno) #no_bounds_check
 	}
 
 	b, n := encode_rune(r)
-	return os_write(b[:n])
+	m, _ := os_write(b[:n])
+	return m
 }
 
 
 print_u64 :: proc "contextless" (x: u64) #no_bounds_check {
-	digits := _INTEGER_DIGITS
-
 	a: [129]byte
 	i := len(a)
 	b := u64(10)
 	u := x
 	for u >= b {
-		i -= 1; a[i] = digits[u % b]
+		i -= 1; a[i] = _INTEGER_DIGITS_VAR[u % b]
 		u /= b
 	}
-	i -= 1; a[i] = digits[u % b]
+	i -= 1; a[i] = _INTEGER_DIGITS_VAR[u % b]
 
 	os_write(a[i:])
 }
 
 
 print_i64 :: proc "contextless" (x: i64) #no_bounds_check {
-	digits := _INTEGER_DIGITS
 	b :: i64(10)
 
 	u := x
@@ -126,10 +194,10 @@ print_i64 :: proc "contextless" (x: i64) #no_bounds_check {
 	a: [129]byte
 	i := len(a)
 	for u >= b {
-		i -= 1; a[i] = digits[u % b]
+		i -= 1; a[i] = _INTEGER_DIGITS_VAR[u % b]
 		u /= b
 	}
-	i -= 1; a[i] = digits[u % b]
+	i -= 1; a[i] = _INTEGER_DIGITS_VAR[u % b]
 	if neg {
 		i -= 1; a[i] = '-'
 	}
@@ -236,7 +304,7 @@ print_type :: proc "contextless" (ti: ^Type_Info) {
 		if info.params == nil {
 			print_string("()")
 		} else {
-			t := info.params.variant.(Type_Info_Tuple)
+			t := info.params.variant.(Type_Info_Parameters)
 			print_byte('(')
 			for t, i in t.types {
 				if i > 0 { print_string(", ") }
@@ -248,7 +316,7 @@ print_type :: proc "contextless" (ti: ^Type_Info) {
 			print_string(" -> ")
 			print_type(info.results)
 		}
-	case Type_Info_Tuple:
+	case Type_Info_Parameters:
 		count := len(info.names)
 		if count != 1 { print_byte('(') }
 		for name, i in info.names {

core/slice/map.odin

@@ -37,7 +37,7 @@ Map_Entry_Info :: struct($Key, $Value: typeid) {
 }
 
 
-map_entries :: proc(m: $M/map[$K]$V, allocator := context.allocator) -> (entries: []Map_Entry(K, V), err: runtime.Allocator) {
+map_entries :: proc(m: $M/map[$K]$V, allocator := context.allocator) -> (entries: []Map_Entry(K, V), err: runtime.Allocator_Error) {
 	entries = make(type_of(entries), len(m), allocator) or_return
 	i := 0
 	for key, value in m {

core/slice/ptr.odin

@@ -73,24 +73,26 @@ ptr_rotate :: proc(left: int, mid: ^$T, right: int) {
 		left, mid, right := left, mid, right
 
 		// TODO(bill): Optimization with a buffer for smaller ranges
-		if left >= right {
-			for {
-				ptr_swap_non_overlapping(ptr_sub(mid, right), mid, right)
-				mid = ptr_sub(mid, right)
+		for left > 0 && right > 0 {
+			if left >= right {
+				for {
+					ptr_swap_non_overlapping(ptr_sub(mid, right), mid, right * size_of(T))
+					mid = ptr_sub(mid, right)
 
-				left -= right
-				if left < right {
-					break
+					left -= right
+					if left < right {
+						break
+					}
 				}
-			}
-		} else {
-			for {
-				ptr_swap_non_overlapping(ptr_sub(mid, left), mid, left)
-				mid = ptr_add(mid, left)
+			} else {
+				for {
+					ptr_swap_non_overlapping(ptr_sub(mid, left), mid, left * size_of(T))
+					mid = ptr_add(mid, left)
 
-				right -= left
-				if right < left {
-					break
+					right -= left
+					if right < left {
+						break
+					}
 				}
 			}
 		}

core/slice/slice.odin

@@ -77,8 +77,7 @@ swap_between :: proc(a, b: $T/[]$E) {
 reverse :: proc(array: $T/[]$E) {
 	n := len(array)/2
 	for i in 0..<n {
-		a, b := i, len(array)-i-1
-		array[a], array[b] = array[b], array[a]
+		swap(array, i, len(array)-i-1)
 	}
 }
 
@@ -218,8 +217,10 @@ rotate_left :: proc(array: $T/[]$E, mid: int) {
 	n := len(array)
 	m := mid %% n
 	k := n - m
-	p := raw_data(array)
-	ptr_rotate(mid, ptr_add(p, mid), k)
+	// FIXME: (ap29600) this cast is a temporary fix for the compiler not matching
+	// [^T] with $P/^$T
+	p := cast(^E)raw_data(array)
+	ptr_rotate(m, ptr_add(p, m), k)
 }
 rotate_right :: proc(array: $T/[]$E, k: int) {
 	rotate_left(array, -k)
@@ -515,4 +516,4 @@ dot_product :: proc(a, b: $S/[]$T) -> (r: T, ok: bool)
 enumerated_array :: proc(ptr: ^$T) -> []intrinsics.type_elem_type(T)
 	where intrinsics.type_is_enumerated_array(T) {
 	return ([^]intrinsics.type_elem_type(T))(ptr)[:len(T)]
-}
+}

core/sort/map.odin

@@ -1,36 +0,0 @@
-package sort
-
-import "core:intrinsics"
-import "core:runtime"
-import "core:slice"
-
-_ :: runtime
-_ :: slice
-
-map_entries_by_key :: proc(m: ^$M/map[$K]$V, loc := #caller_location) where intrinsics.type_is_ordered(K) {
-	Entry :: struct {
-		hash:  uintptr,
-		next:  int,
-		key:   K,
-		value: V,
-	}
-	
-	header := runtime.__get_map_header(m)
-	entries := (^[dynamic]Entry)(&header.m.entries)
-	slice.sort_by_key(entries[:], proc(e: Entry) -> K { return e.key })
-	runtime.__dynamic_map_reset_entries(header, loc)
-}
-
-map_entries_by_value :: proc(m: ^$M/map[$K]$V, loc := #caller_location) where intrinsics.type_is_ordered(V) {
-	Entry :: struct {
-		hash:  uintptr,
-		next:  int,
-		key:   K,
-		value: V,
-	}
-	
-	header := runtime.__get_map_header(m)
-	entries := (^[dynamic]Entry)(&header.m.entries)
-	slice.sort_by_key(entries[:], proc(e: Entry) -> V { return e.value })
-	runtime.__dynamic_map_reset_entries(header, loc)
-}

core/strconv/decimal/decimal.odin

@@ -214,25 +214,128 @@ shift_right :: proc(a: ^Decimal, k: uint) {
 	trim(a)
 }
 
-shift_left :: proc(a: ^Decimal, k: uint) {
-	// NOTE(bill): used to determine buffer size required for the decimal from the binary shift
-	// 'k' means `1<<k` == `2^k` which equates to roundup(k*log10(2)) digits required
-	log10_2 :: 0.301029995663981195213738894724493026768189881462108541310
-	capacity := int(f64(k)*log10_2 + 1)
+import "core:runtime"
+println :: proc(args: ..any) {
+	for arg, i in args {
+		if i != 0 {
+			runtime.print_string(" ")
+		}
+		switch v in arg {
+		case string:  runtime.print_string(v)
+		case rune:    runtime.print_rune(v)
+		case int:     runtime.print_int(v)
+		case uint:    runtime.print_uint(v)
+		case u8:      runtime.print_u64(u64(v))
+		case u16:     runtime.print_u64(u64(v))
+		case u32:     runtime.print_u64(u64(v))
+		case u64:     runtime.print_u64(v)
+		case i8:      runtime.print_i64(i64(v))
+		case i16:     runtime.print_i64(i64(v))
+		case i32:     runtime.print_i64(i64(v))
+		case i64:     runtime.print_i64(v)
+		case uintptr: runtime.print_uintptr(v)
+		case bool:    runtime.print_string("true" if v else "false")
+		}
+	}
+	runtime.print_string("\n")
+}
 
-	r := a.count          // read index
-	w := a.count+capacity // write index
+@(private="file")
+_shift_left_offsets := [?]struct{delta: int, cutoff: string}{
+	{ 0, ""},
+	{ 1, "5"},
+	{ 1, "25"},
+	{ 1, "125"},
+	{ 2, "625"},
+	{ 2, "3125"},
+	{ 2, "15625"},
+	{ 3, "78125"},
+	{ 3, "390625"},
+	{ 3, "1953125"},
+	{ 4, "9765625"},
+	{ 4, "48828125"},
+	{ 4, "244140625"},
+	{ 4, "1220703125"},
+	{ 5, "6103515625"},
+	{ 5, "30517578125"},
+	{ 5, "152587890625"},
+	{ 6, "762939453125"},
+	{ 6, "3814697265625"},
+	{ 6, "19073486328125"},
+	{ 7, "95367431640625"},
+	{ 7, "476837158203125"},
+	{ 7, "2384185791015625"},
+	{ 7, "11920928955078125"},
+	{ 8, "59604644775390625"},
+	{ 8, "298023223876953125"},
+	{ 8, "1490116119384765625"},
+	{ 9, "7450580596923828125"},
+	{ 9, "37252902984619140625"},
+	{ 9, "186264514923095703125"},
+	{10, "931322574615478515625"},
+	{10, "4656612873077392578125"},
+	{10, "23283064365386962890625"},
+	{10, "116415321826934814453125"},
+	{11, "582076609134674072265625"},
+	{11, "2910383045673370361328125"},
+	{11, "14551915228366851806640625"},
+	{12, "72759576141834259033203125"},
+	{12, "363797880709171295166015625"},
+	{12, "1818989403545856475830078125"},
+	{13, "9094947017729282379150390625"},
+	{13, "45474735088646411895751953125"},
+	{13, "227373675443232059478759765625"},
+	{13, "1136868377216160297393798828125"},
+	{14, "5684341886080801486968994140625"},
+	{14, "28421709430404007434844970703125"},
+	{14, "142108547152020037174224853515625"},
+	{15, "710542735760100185871124267578125"},
+	{15, "3552713678800500929355621337890625"},
+	{15, "17763568394002504646778106689453125"},
+	{16, "88817841970012523233890533447265625"},
+	{16, "444089209850062616169452667236328125"},
+	{16, "2220446049250313080847263336181640625"},
+	{16, "11102230246251565404236316680908203125"},
+	{17, "55511151231257827021181583404541015625"},
+	{17, "277555756156289135105907917022705078125"},
+	{17, "1387778780781445675529539585113525390625"},
+	{18, "6938893903907228377647697925567626953125"},
+	{18, "34694469519536141888238489627838134765625"},
+	{18, "173472347597680709441192448139190673828125"},
+	{19, "867361737988403547205962240695953369140625"},
+}
 
-	d := len(a.digits)
+shift_left :: proc(a: ^Decimal, k: uint) #no_bounds_check {
+	prefix_less :: #force_inline proc "contextless" (b: []byte, s: string) -> bool #no_bounds_check {
+		for i in 0..<len(s) {
+			if i >= len(b) {
+				return true
+			}
+			if b[i] != s[i] {
+				return b[i] < s[i]
+			}
+		}
+		return false
+	}
+
+	assert(k < 61)
+
+	delta := _shift_left_offsets[k].delta
+	if prefix_less(a.digits[:a.count], _shift_left_offsets[k].cutoff) {
+		delta -= 1
+	}
+
+	read_index  := a.count
+	write_index := a.count+delta
 
 	n: uint
-	for r -= 1; r >= 0; r -= 1 {
-		n += (uint(a.digits[r]) - '0') << k
+	for read_index -= 1; read_index >= 0; read_index -= 1 {
+		n += (uint(a.digits[read_index]) - '0') << k
 		quo := n/10
 		rem := n - 10*quo
-		w -= 1
-		if w < d {
-			a.digits[w] = byte('0' + rem)
+		write_index -= 1
+		if write_index < len(a.digits) {
+			a.digits[write_index] = byte('0' + rem)
 		} else if rem != 0 {
 			a.trunc = true
 		}
@@ -242,21 +345,18 @@ shift_left :: proc(a: ^Decimal, k: uint) {
 	for n > 0 {
 		quo := n/10
 		rem := n - 10*quo
-		w -= 1
-		if w < d {
-			a.digits[w] = byte('0' + rem)
+		write_index -= 1
+		if write_index < len(a.digits) {
+			a.digits[write_index] = byte('0' + rem)
 		} else if rem != 0 {
 			a.trunc = true
 		}
 		n = quo
 	}
 
-	// NOTE(bill): Remove unused buffer size
-	assert(w >= 0)
-	capacity -= w
+	a.decimal_point += delta
 
-	a.count = min(a.count+capacity, d)
-	a.decimal_point += capacity
+	a.count = clamp(a.count, 0, len(a.digits))
 	trim(a)
 }
 

core/strconv/generic_float.odin

@@ -287,13 +287,13 @@ round_shortest :: proc(d: ^decimal.Decimal, mant: u64, exp: int, flt: ^Float_Inf
 
 @(private)
 decimal_to_float_bits :: proc(d: ^decimal.Decimal, info: ^Float_Info) -> (b: u64, overflow: bool) {
-	end :: proc "contextless" (d: ^decimal.Decimal, mant: u64, exp: int, info: ^Float_Info) -> (b: u64) {
-		bits := mant & (u64(1)<<info.mantbits - 1)
+	end :: proc "contextless" (d: ^decimal.Decimal, mant: u64, exp: int, info: ^Float_Info) -> (bits: u64) {
+		bits = mant & (u64(1)<<info.mantbits - 1)
 		bits |= u64((exp-info.bias) & (1<<info.expbits - 1)) << info.mantbits
 		if d.neg {
 			bits |= 1<< info.mantbits << info.expbits
 		}
-		return bits
+		return
 	}
 	set_overflow :: proc "contextless" (mant: ^u64, exp: ^int, info: ^Float_Info) -> bool {
 		mant^ = 0
@@ -303,7 +303,7 @@ decimal_to_float_bits :: proc(d: ^decimal.Decimal, info: ^Float_Info) -> (b: u64
 
 	mant: u64
 	exp: int
-	if d.decimal_point == 0 {
+	if d.count == 0 {
 		mant = 0
 		exp = info.bias
 		b = end(d, mant, exp, info)
@@ -326,7 +326,7 @@ decimal_to_float_bits :: proc(d: ^decimal.Decimal, info: ^Float_Info) -> (b: u64
 	exp = 0
 	for d.decimal_point > 0 {
 		n := 27 if d.decimal_point >= len(power_table) else power_table[d.decimal_point]
-		decimal.shift(d, n)
+		decimal.shift(d, -n)
 		exp += n
 	}
 	for d.decimal_point < 0 || d.decimal_point == 0 && d.digits[0] < '5' {

core/strconv/integers.odin

@@ -3,7 +3,6 @@ package strconv
 Int_Flag :: enum {
 	Prefix,
 	Plus,
-	Space,
 }
 Int_Flags :: bit_set[Int_Flag]
 
@@ -73,8 +72,6 @@ append_bits :: proc(buf: []byte, x: u64, base: int, is_signed: bool, bit_size: i
 		i-=1; a[i] = '-'
 	case .Plus in flags:
 		i-=1; a[i] = '+'
-	case .Space in flags:
-		i-=1; a[i] = ' '
 	}
 
 	out := a[i:]
@@ -157,8 +154,6 @@ append_bits_128 :: proc(buf: []byte, x: u128, base: int, is_signed: bool, bit_si
 		i-=1; a[i] = '-'
 	case .Plus in flags:
 		i-=1; a[i] = '+'
-	case .Space in flags:
-		i-=1; a[i] = ' '
 	}
 
 	out := a[i:]

core/strconv/strconv.odin

@@ -819,7 +819,7 @@ parse_f64 :: proc(str: string, n: ^int = nil) -> (value: f64, ok: bool) {
 		}
 
 		if mantissa>>_f64_info.mantbits != 0 {
-			return
+			break trunc_block
 		}
 		f := f64(mantissa)
 		if neg {
@@ -841,7 +841,6 @@ parse_f64 :: proc(str: string, n: ^int = nil) -> (value: f64, ok: bool) {
 			return f / pow10[-exp], true
 		}
 	}
-
 	d: decimal.Decimal
 	decimal.set(&d, str[:nr])
 	b, overflow := decimal_to_float_bits(&d, &_f64_info)

core/strings/strings.odin

@@ -5,6 +5,7 @@ import "core:io"
 import "core:mem"
 import "core:slice"
 import "core:unicode"
+import "core:runtime"
 import "core:unicode/utf8"
 
 // returns a clone of the string `s` allocated using the `allocator`
@@ -1425,7 +1426,9 @@ split_multi :: proc(s: string, substrs: []string, allocator := context.allocator
 
 	// TODO maybe remove duplicate substrs
 	// sort substrings by string size, largest to smallest
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	temp_substrs := slice.clone(substrs, context.temp_allocator)
+	defer delete(temp_substrs)
 	slice.sort_by(temp_substrs, proc(a, b: string) -> bool {
 		return len(a) > len(b)	
 	})

core/sync/extended.odin

@@ -1,6 +1,8 @@
 package sync
 
 import "core:time"
+import vg "core:sys/valgrind"
+_ :: vg
 
 // A Wait_Group waits for a collection of threads to finish
 //
@@ -11,7 +13,7 @@ Wait_Group :: struct {
 	cond:    Cond,
 }
 
-wait_group_add :: proc(wg: ^Wait_Group, delta: int) {
+wait_group_add :: proc "contextless" (wg: ^Wait_Group, delta: int) {
 	if delta == 0 {
 		return
 	}
@@ -20,32 +22,32 @@ wait_group_add :: proc(wg: ^Wait_Group, delta: int) {
 
 	atomic_add(&wg.counter, delta)
 	if wg.counter < 0 {
-		panic("sync.Wait_Group negative counter")
+		_panic("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("sync.Wait_Group misuse: sync.wait_group_add called concurrently with sync.wait_group_wait")
 		}
 	}
 }
 
-wait_group_done :: proc(wg: ^Wait_Group) {
+wait_group_done :: proc "contextless" (wg: ^Wait_Group) {
 	wait_group_add(wg, -1)
 }
 
-wait_group_wait :: proc(wg: ^Wait_Group) {
+wait_group_wait :: proc "contextless" (wg: ^Wait_Group) {
 	guard(&wg.mutex)
 
 	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("sync.Wait_Group misuse: sync.wait_group_add called concurrently with sync.wait_group_wait")
 		}
 	}
 }
 
-wait_group_wait_with_timeout :: proc(wg: ^Wait_Group, duration: time.Duration) -> bool {
+wait_group_wait_with_timeout :: proc "contextless" (wg: ^Wait_Group, duration: time.Duration) -> bool {
 	if duration <= 0 {
 		return false
 	}
@@ -56,7 +58,7 @@ wait_group_wait_with_timeout :: proc(wg: ^Wait_Group, duration: time.Duration) -
 			return false
 		}
 		if wg.counter != 0 {
-			panic("sync.Wait_Group misuse: sync.wait_group_add called concurrently with sync.wait_group_wait")
+			_panic("sync.Wait_Group misuse: sync.wait_group_add called concurrently with sync.wait_group_wait")
 		}
 	}
 	return true
@@ -76,7 +78,7 @@ Example:
 
 	barrier := &sync.Barrier{}
 
-	main :: proc() {
+	main :: proc "contextless" () {
 		fmt.println("Start")
 
 		THREAD_COUNT :: 4
@@ -107,7 +109,10 @@ Barrier :: struct {
 	thread_count:  int,
 }
 
-barrier_init :: proc(b: ^Barrier, thread_count: int) {
+barrier_init :: proc "contextless" (b: ^Barrier, thread_count: int) {
+	when ODIN_VALGRIND_SUPPORT {
+		vg.helgrind_barrier_resize_pre(b, uint(thread_count))
+	}
 	b.index = 0
 	b.generation_id = 0
 	b.thread_count = thread_count
@@ -115,7 +120,10 @@ barrier_init :: proc(b: ^Barrier, thread_count: int) {
 
 // Block the current thread until all threads have rendezvoused
 // Barrier can be reused after all threads rendezvoused once, and can be used continuously
-barrier_wait :: proc(b: ^Barrier) -> (is_leader: bool) {
+barrier_wait :: proc "contextless" (b: ^Barrier) -> (is_leader: bool) {
+	when ODIN_VALGRIND_SUPPORT {
+		vg.helgrind_barrier_wait_pre(b)
+	}
 	guard(&b.mutex)
 	local_gen := b.generation_id
 	b.index += 1
@@ -141,7 +149,7 @@ Auto_Reset_Event :: struct {
 	sema:   Sema,
 }
 
-auto_reset_event_signal :: proc(e: ^Auto_Reset_Event) {
+auto_reset_event_signal :: proc "contextless" (e: ^Auto_Reset_Event) {
 	old_status := atomic_load_explicit(&e.status, .Relaxed)
 	for {
 		new_status := old_status + 1 if old_status < 1 else 1
@@ -155,7 +163,7 @@ auto_reset_event_signal :: proc(e: ^Auto_Reset_Event) {
 	}
 }
 
-auto_reset_event_wait :: proc(e: ^Auto_Reset_Event) {
+auto_reset_event_wait :: proc "contextless" (e: ^Auto_Reset_Event) {
 	old_status := atomic_sub_explicit(&e.status, 1, .Acquire)
 	if old_status < 1 {
 		sema_wait(&e.sema)
@@ -169,18 +177,18 @@ Ticket_Mutex :: struct {
 	serving: uint,
 }
 
-ticket_mutex_lock :: #force_inline proc(m: ^Ticket_Mutex) {
+ticket_mutex_lock :: #force_inline proc "contextless" (m: ^Ticket_Mutex) {
 	ticket := atomic_add_explicit(&m.ticket, 1, .Relaxed)
 	for ticket != atomic_load_explicit(&m.serving, .Acquire) {
 		cpu_relax()
 	}
 }
 
-ticket_mutex_unlock :: #force_inline proc(m: ^Ticket_Mutex) {
+ticket_mutex_unlock :: #force_inline proc "contextless" (m: ^Ticket_Mutex) {
 	atomic_add_explicit(&m.serving, 1, .Relaxed)
 }
 @(deferred_in=ticket_mutex_unlock)
-ticket_mutex_guard :: proc(m: ^Ticket_Mutex) -> bool {
+ticket_mutex_guard :: proc "contextless" (m: ^Ticket_Mutex) -> bool {
 	ticket_mutex_lock(m)
 	return true
 }
@@ -191,25 +199,25 @@ Benaphore :: struct {
 	sema:    Sema,
 }
 
-benaphore_lock :: proc(b: ^Benaphore) {
+benaphore_lock :: proc "contextless" (b: ^Benaphore) {
 	if atomic_add_explicit(&b.counter, 1, .Acquire) > 1 {
 		sema_wait(&b.sema)
 	}
 }
 
-benaphore_try_lock :: proc(b: ^Benaphore) -> bool {
+benaphore_try_lock :: proc "contextless" (b: ^Benaphore) -> bool {
 	v, _ := atomic_compare_exchange_strong_explicit(&b.counter, 0, 1, .Acquire, .Acquire)
 	return v == 0
 }
 
-benaphore_unlock :: proc(b: ^Benaphore) {
+benaphore_unlock :: proc "contextless" (b: ^Benaphore) {
 	if atomic_sub_explicit(&b.counter, 1, .Release) > 0 {
 		sema_post(&b.sema)
 	}
 }
 
 @(deferred_in=benaphore_unlock)
-benaphore_guard :: proc(m: ^Benaphore) -> bool {
+benaphore_guard :: proc "contextless" (m: ^Benaphore) -> bool {
 	benaphore_lock(m)
 	return true
 }
@@ -221,7 +229,7 @@ Recursive_Benaphore :: struct {
 	sema:      Sema,
 }
 
-recursive_benaphore_lock :: proc(b: ^Recursive_Benaphore) {
+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 {
@@ -233,7 +241,7 @@ recursive_benaphore_lock :: proc(b: ^Recursive_Benaphore) {
 	b.recursion += 1
 }
 
-recursive_benaphore_try_lock :: proc(b: ^Recursive_Benaphore) -> bool {
+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)
@@ -248,9 +256,9 @@ recursive_benaphore_try_lock :: proc(b: ^Recursive_Benaphore) -> bool {
 	return true
 }
 
-recursive_benaphore_unlock :: proc(b: ^Recursive_Benaphore) {
+recursive_benaphore_unlock :: proc "contextless" (b: ^Recursive_Benaphore) {
 	tid := current_thread_id()
-	assert(tid == b.owner)
+	_assert(tid == b.owner, "tid != b.owner")
 	b.recursion -= 1
 	recursion := b.recursion
 	if recursion == 0 {
@@ -265,7 +273,7 @@ recursive_benaphore_unlock :: proc(b: ^Recursive_Benaphore) {
 }
 
 @(deferred_in=recursive_benaphore_unlock)
-recursive_benaphore_guard :: proc(m: ^Recursive_Benaphore) -> bool {
+recursive_benaphore_guard :: proc "contextless" (m: ^Recursive_Benaphore) -> bool {
 	recursive_benaphore_lock(m)
 	return true
 }
@@ -282,7 +290,15 @@ Once :: struct {
 }
 
 // once_do calls the procedure fn if and only if once_do is being called for the first for this instance of Once.
-once_do :: proc(o: ^Once, fn: proc()) {
+once_do :: proc{
+	once_do_without_data,
+	once_do_without_data_contextless,
+	once_do_with_data,
+	once_do_with_data_contextless,
+}
+
+// once_do_without_data calls the procedure fn if and only if once_do_without_data is being called for the first for this instance of Once.
+once_do_without_data :: proc(o: ^Once, fn: proc()) {
 	@(cold)
 	do_slow :: proc(o: ^Once, fn: proc()) {
 		guard(&o.m)
@@ -292,12 +308,61 @@ once_do :: proc(o: ^Once, fn: proc()) {
 		}
 	}
 
-	
 	if atomic_load_explicit(&o.done, .Acquire) == false {
 		do_slow(o, fn)
 	}
 }
 
+// once_do_without_data calls the procedure fn if and only if once_do_without_data is being called for the first for this instance of Once.
+once_do_without_data_contextless :: proc(o: ^Once, fn: proc "contextless" ()) {
+	@(cold)
+	do_slow :: proc(o: ^Once, fn: proc "contextless" ()) {
+		guard(&o.m)
+		if !o.done {
+			fn()
+			atomic_store_explicit(&o.done, true, .Release)
+		}
+	}
+
+	if atomic_load_explicit(&o.done, .Acquire) == false {
+		do_slow(o, fn)
+	}
+}
+
+// once_do_with_data calls the procedure fn if and only if once_do_with_data is being called for the first for this instance of Once.
+once_do_with_data :: proc(o: ^Once, fn: proc(data: rawptr), data: rawptr) {
+	@(cold)
+	do_slow :: proc(o: ^Once, fn: proc(data: rawptr), data: rawptr) {
+		guard(&o.m)
+		if !o.done {
+			fn(data)
+			atomic_store_explicit(&o.done, true, .Release)
+		}
+	}
+
+	if atomic_load_explicit(&o.done, .Acquire) == false {
+		do_slow(o, fn, data)
+	}
+}
+
+// once_do_with_data_contextless calls the procedure fn if and only if once_do_with_data_contextless is being called for the first for this instance of Once.
+once_do_with_data_contextless :: proc "contextless" (o: ^Once, fn: proc "contextless" (data: rawptr), data: rawptr) {
+	@(cold)
+	do_slow :: proc "contextless" (o: ^Once, fn: proc "contextless" (data: rawptr), data: rawptr) {
+		guard(&o.m)
+		if !o.done {
+			fn(data)
+			atomic_store_explicit(&o.done, true, .Release)
+		}
+	}
+
+	if atomic_load_explicit(&o.done, .Acquire) == false {
+		do_slow(o, fn, data)
+	}
+}
+
+
+
 
 
 // A Parker is an associated token which is initially not present:
@@ -314,7 +379,7 @@ Parker :: struct {
 // Blocks the current thread until the token is made available.
 //
 // Assumes this is only called by the thread that owns the Parker.
-park :: proc(p: ^Parker) {
+park :: proc "contextless" (p: ^Parker) {
 	EMPTY    :: 0
 	NOTIFIED :: 1
 	PARKED   :: max(u32)
@@ -333,7 +398,7 @@ park :: proc(p: ^Parker) {
 // for a limited duration.
 //
 // Assumes this is only called by the thread that owns the Parker
-park_with_timeout :: proc(p: ^Parker, duration: time.Duration) {
+park_with_timeout :: proc "contextless" (p: ^Parker, duration: time.Duration) {
 	EMPTY    :: 0
 	NOTIFIED :: 1
 	PARKED   :: max(u32)
@@ -345,7 +410,7 @@ park_with_timeout :: proc(p: ^Parker, duration: time.Duration) {
 }
 
 // Automatically makes thee token available if it was not already.
-unpark :: proc(p: ^Parker)  {
+unpark :: proc "contextless" (p: ^Parker)  {
 	EMPTY    :: 0
 	NOTIFIED :: 1
 	PARKED   :: max(Futex)

core/sync/futex_darwin.odin

@@ -24,11 +24,11 @@ EINTR     :: -4
 EFAULT    :: -14
 ETIMEDOUT :: -60
 
-_futex_wait :: proc(f: ^Futex, expected: u32) -> bool {
+_futex_wait :: proc "contextless" (f: ^Futex, expected: u32) -> bool {
 	return _futex_wait_with_timeout(f, expected, 0)
 }
 
-_futex_wait_with_timeout :: proc(f: ^Futex, expected: u32, duration: time.Duration) -> bool {
+_futex_wait_with_timeout :: proc "contextless" (f: ^Futex, expected: u32, duration: time.Duration) -> bool {
 	timeout_ns := u32(duration) * 1000
 	
 	s := __ulock_wait(UL_COMPARE_AND_WAIT | ULF_NO_ERRNO, f, u64(expected), timeout_ns)
@@ -41,13 +41,13 @@ _futex_wait_with_timeout :: proc(f: ^Futex, expected: u32, duration: time.Durati
 	case ETIMEDOUT:
 		return false
 	case:
-		panic("futex_wait failure")
+		_panic("futex_wait failure")
 	}
 	return true
 
 }
 
-_futex_signal :: proc(f: ^Futex) {
+_futex_signal :: proc "contextless" (f: ^Futex) {
 	loop: for {
 		s := __ulock_wake(UL_COMPARE_AND_WAIT | ULF_NO_ERRNO, f, 0)
 		if s >= 0 {
@@ -59,12 +59,12 @@ _futex_signal :: proc(f: ^Futex) {
 		case ENOENT:
 			return
 		case:
-			panic("futex_wake_single failure")
+			_panic("futex_wake_single failure")
 		}
 	}
 }
 
-_futex_broadcast :: proc(f: ^Futex) {
+_futex_broadcast :: proc "contextless" (f: ^Futex) {
 	loop: for {
 		s := __ulock_wake(UL_COMPARE_AND_WAIT | ULF_NO_ERRNO | ULF_WAKE_ALL, f, 0)
 		if s >= 0 {
@@ -76,7 +76,7 @@ _futex_broadcast :: proc(f: ^Futex) {
 		case ENOENT:
 			return
 		case:
-			panic("futex_wake_all failure")
+			_panic("futex_wake_all failure")
 		}
 	}
 }

core/sync/futex_freebsd.odin

@@ -17,7 +17,7 @@ foreign libc {
 	__error :: proc "c" () -> ^c.int ---
 }
 
-_futex_wait :: proc(f: ^Futex, expected: u32) -> bool {
+_futex_wait :: proc "contextless" (f: ^Futex, expected: u32) -> bool {
 	timeout := [2]i64{14400, 0} // 4 hours
 	for {
 		res := _umtx_op(f, UMTX_OP_WAIT, c.ulong(expected), nil, &timeout)
@@ -30,12 +30,12 @@ _futex_wait :: proc(f: ^Futex, expected: u32) -> bool {
 			continue
 		}
 
-		panic("_futex_wait failure")
+		_panic("_futex_wait failure")
 	}
 	unreachable()
 }
 
-_futex_wait_with_timeout :: proc(f: ^Futex, expected: u32, duration: time.Duration) -> bool {
+_futex_wait_with_timeout :: proc "contextless" (f: ^Futex, expected: u32, duration: time.Duration) -> bool {
 	if duration <= 0 {
 		return false
 	}
@@ -51,21 +51,21 @@ _futex_wait_with_timeout :: proc(f: ^Futex, expected: u32, duration: time.Durati
 		return false
 	}
 
-	panic("_futex_wait_with_timeout failure")
+	_panic("_futex_wait_with_timeout failure")
 }
 
-_futex_signal :: proc(f: ^Futex) {
+_futex_signal :: proc "contextless" (f: ^Futex) {
 	res := _umtx_op(f, UMTX_OP_WAKE, 1, nil, nil)
 
 	if res == -1 {
-		panic("_futex_signal failure")
+		_panic("_futex_signal failure")
 	}
 }
 
-_futex_broadcast :: proc(f: ^Futex)  {
+_futex_broadcast :: proc "contextless" (f: ^Futex)  {
 	res := _umtx_op(f, UMTX_OP_WAKE, c.ulong(max(i32)), nil, nil)
 
 	if res == -1 {
-		panic("_futex_broadcast failure")
+		_panic("_futex_broadcast failure")
 	}
 }

core/sync/futex_linux.odin

@@ -21,20 +21,20 @@ EFAULT    :: -14
 EINVAL    :: -22
 ETIMEDOUT :: -110
 
-get_errno :: proc(r: int) -> int {
+get_errno :: proc "contextless" (r: int) -> int {
 	if -4096 < r && r < 0 {
 		return r
 	}
 	return 0
 }
 
-internal_futex :: proc(f: ^Futex, op: c.int, val: u32, timeout: rawptr) -> int {
+internal_futex :: proc "contextless" (f: ^Futex, op: c.int, val: u32, timeout: rawptr) -> int {
 	code := int(intrinsics.syscall(unix.SYS_futex, uintptr(f), uintptr(op), uintptr(val), uintptr(timeout), 0, 0))
 	return get_errno(code)
 }
 
 
-_futex_wait :: proc(f: ^Futex, expected: u32) -> bool {
+_futex_wait :: proc "contextless" (f: ^Futex, expected: u32) -> bool {
 	err := internal_futex(f, FUTEX_WAIT_PRIVATE | FUTEX_WAIT, expected, nil)
 	switch err {
 	case ESUCCESS, EINTR, EAGAIN, EINVAL:
@@ -44,12 +44,12 @@ _futex_wait :: proc(f: ^Futex, expected: u32) -> bool {
 	case EFAULT: 
 		fallthrough
 	case:
-		panic("futex_wait failure")
+		_panic("futex_wait failure")
 	}
 	return true
 }
 
-_futex_wait_with_timeout :: proc(f: ^Futex, expected: u32, duration: time.Duration) -> bool {
+_futex_wait_with_timeout :: proc "contextless" (f: ^Futex, expected: u32, duration: time.Duration) -> bool {
 	if duration <= 0 {
 		return false
 	}
@@ -71,27 +71,27 @@ _futex_wait_with_timeout :: proc(f: ^Futex, expected: u32, duration: time.Durati
 	case EFAULT: 
 		fallthrough
 	case:
-		panic("futex_wait_with_timeout failure")
+		_panic("futex_wait_with_timeout failure")
 	}
 	return true
 }
 
 
-_futex_signal :: proc(f: ^Futex) {
+_futex_signal :: proc "contextless" (f: ^Futex) {
 	err := internal_futex(f, FUTEX_WAKE_PRIVATE | FUTEX_WAKE, 1, nil)
 	switch err {
 	case ESUCCESS, EINVAL, EFAULT:
 		// okay
 	case:
-		panic("futex_wake_single failure")
+		_panic("futex_wake_single failure")
 	}
 }
-_futex_broadcast :: proc(f: ^Futex)  {
+_futex_broadcast :: proc "contextless" (f: ^Futex)  {
 	err := internal_futex(f, FUTEX_WAKE_PRIVATE | FUTEX_WAKE, u32(max(i32)), nil)
 	switch err {
 	case ESUCCESS, EINVAL, EFAULT:
 		// okay
 	case:
-		panic("_futex_wake_all failure")
+		_panic("_futex_wake_all failure")
 	}
 }

core/sync/futex_openbsd.odin

@@ -21,7 +21,7 @@ foreign libc {
 	_unix_futex :: proc "c" (f: ^Futex, op: c.int, val: u32, timeout: rawptr) -> c.int ---
 }
 
-_futex_wait :: proc(f: ^Futex, expected: u32) -> bool {
+_futex_wait :: proc "contextless" (f: ^Futex, expected: u32) -> bool {
 	res := _unix_futex(f, FUTEX_WAIT_PRIVATE, expected, nil)
 
 	if res != -1 {
@@ -32,10 +32,10 @@ _futex_wait :: proc(f: ^Futex, expected: u32) -> bool {
 		return false
 	}
 
-	panic("futex_wait failure")
+	_panic("futex_wait failure")
 }
 
-_futex_wait_with_timeout :: proc(f: ^Futex, expected: u32, duration: time.Duration) -> bool {
+_futex_wait_with_timeout :: proc "contextless" (f: ^Futex, expected: u32, duration: time.Duration) -> bool {
 	if duration <= 0 {
 		return false
 	}
@@ -58,21 +58,21 @@ _futex_wait_with_timeout :: proc(f: ^Futex, expected: u32, duration: time.Durati
 		return false
 	}
 
-	panic("futex_wait_with_timeout failure")
+	_panic("futex_wait_with_timeout failure")
 }
 
-_futex_signal :: proc(f: ^Futex) {
+_futex_signal :: proc "contextless" (f: ^Futex) {
 	res := _unix_futex(f, FUTEX_WAKE_PRIVATE, 1, nil)
 
 	if res == -1 {
-		panic("futex_wake_single failure")
+		_panic("futex_wake_single failure")
 	}
 }
 
-_futex_broadcast :: proc(f: ^Futex)  {
+_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("_futex_wake_all failure")
 	}
 }

core/sync/futex_wasm.odin

@@ -5,18 +5,18 @@ package sync
 import "core:intrinsics"
 import "core:time"
 
-_futex_wait :: proc(f: ^Futex, expected: u32) -> bool {
+_futex_wait :: proc "contextless" (f: ^Futex, expected: u32) -> bool {
 	s := intrinsics.wasm_memory_atomic_wait32((^u32)(f), expected, -1)
 	return s != 0
 }
 
-_futex_wait_with_timeout :: proc(f: ^Futex, expected: u32, duration: time.Duration) -> bool {
+_futex_wait_with_timeout :: proc "contextless" (f: ^Futex, expected: u32, duration: time.Duration) -> bool {
 	s := intrinsics.wasm_memory_atomic_wait32((^u32)(f), expected, i64(duration))
 	return s != 0
 
 }
 
-_futex_signal :: proc(f: ^Futex) {
+_futex_signal :: proc "contextless" (f: ^Futex) {
 	loop: for {
 		s := intrinsics.wasm_memory_atomic_notify32((^u32)(f), 1)
 		if s >= 1 {
@@ -25,7 +25,7 @@ _futex_signal :: proc(f: ^Futex) {
 	}
 }
 
-_futex_broadcast :: proc(f: ^Futex) {
+_futex_broadcast :: proc "contextless" (f: ^Futex) {
 	loop: for {
 		s := intrinsics.wasm_memory_atomic_notify32((^u32)(f), ~u32(0))
 		if s >= 0 {

core/sync/futex_windows.odin

@@ -39,22 +39,22 @@ CustomWaitOnAddress :: proc "stdcall" (Address: rawptr, CompareAddress: rawptr,
 }
 
 
-_futex_wait :: proc(f: ^Futex, expect: u32) -> bool {
+_futex_wait :: proc "contextless" (f: ^Futex, expect: u32) -> bool {
 	expect := expect
 	return CustomWaitOnAddress(f, &expect, size_of(expect), nil)
 }
 
-_futex_wait_with_timeout :: proc(f: ^Futex, expect: u32, duration: time.Duration) -> bool {
+_futex_wait_with_timeout :: proc "contextless" (f: ^Futex, expect: u32, duration: time.Duration) -> bool {
 	expect := expect
 	// NOTE(bill): for some bizarre reason, this has be a negative number
 	timeout := -i64(duration / 100)
 	return CustomWaitOnAddress(f, &expect, size_of(expect), &timeout)
 }
 
-_futex_signal :: proc(f: ^Futex) {
+_futex_signal :: proc "contextless" (f: ^Futex) {
 	WakeByAddressSingle(f)
 }
 
-_futex_broadcast :: proc(f: ^Futex) {
+_futex_broadcast :: proc "contextless" (f: ^Futex) {
 	WakeByAddressAll(f)
 }

core/sync/primitives.odin

@@ -1,5 +1,6 @@
 package sync
 
+import "core:runtime"
 import "core:time"
 
 current_thread_id :: proc "contextless" () -> int {
@@ -15,17 +16,17 @@ Mutex :: struct {
 }
 
 // mutex_lock locks m
-mutex_lock :: proc(m: ^Mutex) {
+mutex_lock :: proc "contextless" (m: ^Mutex) {
 	_mutex_lock(m)
 }
 
 // mutex_unlock unlocks m
-mutex_unlock :: proc(m: ^Mutex) {
+mutex_unlock :: proc "contextless" (m: ^Mutex) {
 	_mutex_unlock(m)
 }
 
 // mutex_try_lock tries to lock m, will return true on success, and false on failure
-mutex_try_lock :: proc(m: ^Mutex) -> bool {
+mutex_try_lock :: proc "contextless" (m: ^Mutex) -> bool {
 	return _mutex_try_lock(m)
 }
 
@@ -36,7 +37,7 @@ Example:
 	}
 */
 @(deferred_in=mutex_unlock)
-mutex_guard :: proc(m: ^Mutex) -> bool {
+mutex_guard :: proc "contextless" (m: ^Mutex) -> bool {
 	mutex_lock(m)
 	return true
 }
@@ -52,32 +53,32 @@ RW_Mutex :: struct {
 
 // rw_mutex_lock locks rw for writing (with a single writer)
 // If the mutex is already locked for reading or writing, the mutex blocks until the mutex is available.
-rw_mutex_lock :: proc(rw: ^RW_Mutex) {
+rw_mutex_lock :: proc "contextless" (rw: ^RW_Mutex) {
 	_rw_mutex_lock(rw)
 }
 
 // rw_mutex_unlock unlocks rw for writing (with a single writer)
-rw_mutex_unlock :: proc(rw: ^RW_Mutex) {
+rw_mutex_unlock :: proc "contextless" (rw: ^RW_Mutex) {
 	_rw_mutex_unlock(rw)
 }
 
 // rw_mutex_try_lock tries to lock rw for writing (with a single writer)
-rw_mutex_try_lock :: proc(rw: ^RW_Mutex) -> bool {
+rw_mutex_try_lock :: proc "contextless" (rw: ^RW_Mutex) -> bool {
 	return _rw_mutex_try_lock(rw)
 }
 
 // rw_mutex_shared_lock locks rw for reading (with arbitrary number of readers)
-rw_mutex_shared_lock :: proc(rw: ^RW_Mutex) {
+rw_mutex_shared_lock :: proc "contextless" (rw: ^RW_Mutex) {
 	_rw_mutex_shared_lock(rw)
 }
 
 // rw_mutex_shared_unlock unlocks rw for reading (with arbitrary number of readers)
-rw_mutex_shared_unlock :: proc(rw: ^RW_Mutex) {
+rw_mutex_shared_unlock :: proc "contextless" (rw: ^RW_Mutex) {
 	_rw_mutex_shared_unlock(rw)
 }
 
 // rw_mutex_try_shared_lock tries to lock rw for reading (with arbitrary number of readers)
-rw_mutex_try_shared_lock :: proc(rw: ^RW_Mutex) -> bool {
+rw_mutex_try_shared_lock :: proc "contextless" (rw: ^RW_Mutex) -> bool {
 	return _rw_mutex_try_shared_lock(rw)
 }
 /*
@@ -87,7 +88,7 @@ Example:
 	}
 */
 @(deferred_in=rw_mutex_unlock)
-rw_mutex_guard :: proc(m: ^RW_Mutex) -> bool {
+rw_mutex_guard :: proc "contextless" (m: ^RW_Mutex) -> bool {
 	rw_mutex_lock(m)
 	return true
 }
@@ -99,7 +100,7 @@ Example:
 	}
 */
 @(deferred_in=rw_mutex_shared_unlock)
-rw_mutex_shared_guard :: proc(m: ^RW_Mutex) -> bool {
+rw_mutex_shared_guard :: proc "contextless" (m: ^RW_Mutex) -> bool {
 	rw_mutex_shared_lock(m)
 	return true
 }
@@ -114,15 +115,15 @@ Recursive_Mutex :: struct {
 	impl: _Recursive_Mutex,
 }
 
-recursive_mutex_lock :: proc(m: ^Recursive_Mutex) {
+recursive_mutex_lock :: proc "contextless" (m: ^Recursive_Mutex) {
 	_recursive_mutex_lock(m)
 }
 
-recursive_mutex_unlock :: proc(m: ^Recursive_Mutex) {
+recursive_mutex_unlock :: proc "contextless" (m: ^Recursive_Mutex) {
 	_recursive_mutex_unlock(m)
 }
 
-recursive_mutex_try_lock :: proc(m: ^Recursive_Mutex) -> bool {
+recursive_mutex_try_lock :: proc "contextless" (m: ^Recursive_Mutex) -> bool {
 	return _recursive_mutex_try_lock(m)
 }
 
@@ -133,7 +134,7 @@ Example:
 	}
 */
 @(deferred_in=recursive_mutex_unlock)
-recursive_mutex_guard :: proc(m: ^Recursive_Mutex) -> bool {
+recursive_mutex_guard :: proc "contextless" (m: ^Recursive_Mutex) -> bool {
 	recursive_mutex_lock(m)
 	return true
 }
@@ -147,22 +148,22 @@ Cond :: struct {
 	impl: _Cond,
 }
 
-cond_wait :: proc(c: ^Cond, m: ^Mutex) {
+cond_wait :: proc "contextless" (c: ^Cond, m: ^Mutex) {
 	_cond_wait(c, m)
 }
 
-cond_wait_with_timeout :: proc(c: ^Cond, m: ^Mutex, duration: time.Duration) -> bool {
+cond_wait_with_timeout :: proc "contextless" (c: ^Cond, m: ^Mutex, duration: time.Duration) -> bool {
 	if duration <= 0 {
 		return false
 	}
 	return _cond_wait_with_timeout(c, m, duration)
 }
 
-cond_signal :: proc(c: ^Cond) {
+cond_signal :: proc "contextless" (c: ^Cond) {
 	_cond_signal(c)
 }
 
-cond_broadcast :: proc(c: ^Cond) {
+cond_broadcast :: proc "contextless" (c: ^Cond) {
 	_cond_broadcast(c)
 }
 
@@ -175,15 +176,15 @@ Sema :: struct {
 	impl: _Sema,
 }
 
-sema_post :: proc(s: ^Sema, count := 1) {
+sema_post :: proc "contextless" (s: ^Sema, count := 1) {
 	_sema_post(s, count)
 }
 
-sema_wait :: proc(s: ^Sema) {
+sema_wait :: proc "contextless" (s: ^Sema) {
 	_sema_wait(s)
 }
 
-sema_wait_with_timeout :: proc(s: ^Sema, duration: time.Duration) -> bool {
+sema_wait_with_timeout :: proc "contextless" (s: ^Sema, duration: time.Duration) -> bool {
 	return _sema_wait_with_timeout(s, duration)
 }
 
@@ -194,16 +195,16 @@ sema_wait_with_timeout :: proc(s: ^Sema, duration: time.Duration) -> bool {
 // An Futex must not be copied after first use
 Futex :: distinct u32
 
-futex_wait :: proc(f: ^Futex, expected: u32) {
+futex_wait :: proc "contextless" (f: ^Futex, expected: u32) {
 	if u32(atomic_load_explicit(f, .Acquire)) != expected {
 		return
 	}
 	
-	assert(_futex_wait(f, expected), "futex_wait failure")
+	_assert(_futex_wait(f, expected), "futex_wait failure")
 }
 
 // returns true if the wait happened within the duration, false if it exceeded the time duration
-futex_wait_with_timeout :: proc(f: ^Futex, expected: u32, duration: time.Duration) -> bool {
+futex_wait_with_timeout :: proc "contextless" (f: ^Futex, expected: u32, duration: time.Duration) -> bool {
 	if u32(atomic_load_explicit(f, .Acquire)) != expected {
 		return true
 	}
@@ -214,10 +215,25 @@ futex_wait_with_timeout :: proc(f: ^Futex, expected: u32, duration: time.Duratio
 	return _futex_wait_with_timeout(f, expected, duration)
 }
 
-futex_signal :: proc(f: ^Futex) {
+futex_signal :: proc "contextless" (f: ^Futex) {
 	_futex_signal(f)
 }
 
-futex_broadcast :: proc(f: ^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()
+}

core/sync/primitives_atomic.odin

@@ -18,9 +18,9 @@ Atomic_Mutex :: struct {
 }
 
 // atomic_mutex_lock locks m
-atomic_mutex_lock :: proc(m: ^Atomic_Mutex) {
+atomic_mutex_lock :: proc "contextless" (m: ^Atomic_Mutex) {
 	@(cold)
-	lock_slow :: proc(m: ^Atomic_Mutex, curr_state: Atomic_Mutex_State) {
+	lock_slow :: proc "contextless" (m: ^Atomic_Mutex, curr_state: Atomic_Mutex_State) {
 		new_state := curr_state // Make a copy of it
 
 		spin_lock: for spin in 0..<i32(100) {
@@ -58,9 +58,9 @@ atomic_mutex_lock :: proc(m: ^Atomic_Mutex) {
 }
 
 // atomic_mutex_unlock unlocks m
-atomic_mutex_unlock :: proc(m: ^Atomic_Mutex) {
+atomic_mutex_unlock :: proc "contextless" (m: ^Atomic_Mutex) {
 	@(cold)
-	unlock_slow :: proc(m: ^Atomic_Mutex) {
+	unlock_slow :: proc "contextless" (m: ^Atomic_Mutex) {
 		futex_signal((^Futex)(&m.state))
 	}
 
@@ -76,7 +76,7 @@ atomic_mutex_unlock :: proc(m: ^Atomic_Mutex) {
 }
 
 // atomic_mutex_try_lock tries to lock m, will return true on success, and false on failure
-atomic_mutex_try_lock :: proc(m: ^Atomic_Mutex) -> bool {
+atomic_mutex_try_lock :: proc "contextless" (m: ^Atomic_Mutex) -> bool {
 	_, ok := atomic_compare_exchange_strong_explicit(&m.state, .Unlocked, .Locked, .Acquire, .Consume)
 	return ok
 }
@@ -88,7 +88,7 @@ Example:
 	}
 */
 @(deferred_in=atomic_mutex_unlock)
-atomic_mutex_guard :: proc(m: ^Atomic_Mutex) -> bool {
+atomic_mutex_guard :: proc "contextless" (m: ^Atomic_Mutex) -> bool {
 	atomic_mutex_lock(m)
 	return true
 }
@@ -117,7 +117,7 @@ Atomic_RW_Mutex :: struct {
 
 // atomic_rw_mutex_lock locks rw for writing (with a single writer)
 // If the mutex is already locked for reading or writing, the mutex blocks until the mutex is available.
-atomic_rw_mutex_lock :: proc(rw: ^Atomic_RW_Mutex) {
+atomic_rw_mutex_lock :: proc "contextless" (rw: ^Atomic_RW_Mutex) {
 	_ = atomic_add(&rw.state, Atomic_RW_Mutex_State_Writer)
 	atomic_mutex_lock(&rw.mutex)
 
@@ -128,13 +128,13 @@ atomic_rw_mutex_lock :: proc(rw: ^Atomic_RW_Mutex) {
 }
 
 // atomic_rw_mutex_unlock unlocks rw for writing (with a single writer)
-atomic_rw_mutex_unlock :: proc(rw: ^Atomic_RW_Mutex) {
+atomic_rw_mutex_unlock :: proc "contextless" (rw: ^Atomic_RW_Mutex) {
 	_ = atomic_and(&rw.state, ~Atomic_RW_Mutex_State_Is_Writing)
 	atomic_mutex_unlock(&rw.mutex)
 }
 
 // atomic_rw_mutex_try_lock tries to lock rw for writing (with a single writer)
-atomic_rw_mutex_try_lock :: proc(rw: ^Atomic_RW_Mutex) -> bool {
+atomic_rw_mutex_try_lock :: proc "contextless" (rw: ^Atomic_RW_Mutex) -> bool {
 	if atomic_mutex_try_lock(&rw.mutex) {
 		state := atomic_load(&rw.state)
 		if state & Atomic_RW_Mutex_State_Reader_Mask == 0 {
@@ -148,7 +148,7 @@ atomic_rw_mutex_try_lock :: proc(rw: ^Atomic_RW_Mutex) -> bool {
 }
 
 // atomic_rw_mutex_shared_lock locks rw for reading (with arbitrary number of readers)
-atomic_rw_mutex_shared_lock :: proc(rw: ^Atomic_RW_Mutex) {
+atomic_rw_mutex_shared_lock :: proc "contextless" (rw: ^Atomic_RW_Mutex) {
 	state := atomic_load(&rw.state)
 	for state & (Atomic_RW_Mutex_State_Is_Writing|Atomic_RW_Mutex_State_Writer_Mask) == 0 {
 		ok: bool
@@ -164,7 +164,7 @@ atomic_rw_mutex_shared_lock :: proc(rw: ^Atomic_RW_Mutex) {
 }
 
 // atomic_rw_mutex_shared_unlock unlocks rw for reading (with arbitrary number of readers)
-atomic_rw_mutex_shared_unlock :: proc(rw: ^Atomic_RW_Mutex) {
+atomic_rw_mutex_shared_unlock :: proc "contextless" (rw: ^Atomic_RW_Mutex) {
 	state := atomic_sub(&rw.state, Atomic_RW_Mutex_State_Reader)
 
 	if (state & Atomic_RW_Mutex_State_Reader_Mask == Atomic_RW_Mutex_State_Reader) &&
@@ -174,7 +174,7 @@ atomic_rw_mutex_shared_unlock :: proc(rw: ^Atomic_RW_Mutex) {
 }
 
 // atomic_rw_mutex_try_shared_lock tries to lock rw for reading (with arbitrary number of readers)
-atomic_rw_mutex_try_shared_lock :: proc(rw: ^Atomic_RW_Mutex) -> bool {
+atomic_rw_mutex_try_shared_lock :: proc "contextless" (rw: ^Atomic_RW_Mutex) -> bool {
 	state := atomic_load(&rw.state)
 	if state & (Atomic_RW_Mutex_State_Is_Writing|Atomic_RW_Mutex_State_Writer_Mask) == 0 {
 		_, ok := atomic_compare_exchange_strong(&rw.state, state, state + Atomic_RW_Mutex_State_Reader)
@@ -198,7 +198,7 @@ Example:
 	}
 */
 @(deferred_in=atomic_rw_mutex_unlock)
-atomic_rw_mutex_guard :: proc(m: ^Atomic_RW_Mutex) -> bool {
+atomic_rw_mutex_guard :: proc "contextless" (m: ^Atomic_RW_Mutex) -> bool {
 	atomic_rw_mutex_lock(m)
 	return true
 }
@@ -210,7 +210,7 @@ Example:
 	}
 */
 @(deferred_in=atomic_rw_mutex_shared_unlock)
-atomic_rw_mutex_shared_guard :: proc(m: ^Atomic_RW_Mutex) -> bool {
+atomic_rw_mutex_shared_guard :: proc "contextless" (m: ^Atomic_RW_Mutex) -> bool {
 	atomic_rw_mutex_shared_lock(m)
 	return true
 }
@@ -228,7 +228,7 @@ Atomic_Recursive_Mutex :: struct {
 	mutex: Mutex,
 }
 
-atomic_recursive_mutex_lock :: proc(m: ^Atomic_Recursive_Mutex) {
+atomic_recursive_mutex_lock :: proc "contextless" (m: ^Atomic_Recursive_Mutex) {
 	tid := current_thread_id()
 	if tid != m.owner {
 		mutex_lock(&m.mutex)
@@ -238,9 +238,9 @@ atomic_recursive_mutex_lock :: proc(m: ^Atomic_Recursive_Mutex) {
 	m.recursion += 1
 }
 
-atomic_recursive_mutex_unlock :: proc(m: ^Atomic_Recursive_Mutex) {
+atomic_recursive_mutex_unlock :: proc "contextless" (m: ^Atomic_Recursive_Mutex) {
 	tid := current_thread_id()
-	assert(tid == m.owner)
+	_assert(tid == m.owner, "tid != m.owner")
 	m.recursion -= 1
 	recursion := m.recursion
 	if recursion == 0 {
@@ -253,7 +253,7 @@ atomic_recursive_mutex_unlock :: proc(m: ^Atomic_Recursive_Mutex) {
 
 }
 
-atomic_recursive_mutex_try_lock :: proc(m: ^Atomic_Recursive_Mutex) -> bool {
+atomic_recursive_mutex_try_lock :: proc "contextless" (m: ^Atomic_Recursive_Mutex) -> bool {
 	tid := current_thread_id()
 	if m.owner == tid {
 		return mutex_try_lock(&m.mutex)
@@ -274,7 +274,7 @@ Example:
 	}
 */
 @(deferred_in=atomic_recursive_mutex_unlock)
-atomic_recursive_mutex_guard :: proc(m: ^Atomic_Recursive_Mutex) -> bool {
+atomic_recursive_mutex_guard :: proc "contextless" (m: ^Atomic_Recursive_Mutex) -> bool {
 	atomic_recursive_mutex_lock(m)
 	return true
 }
@@ -289,7 +289,7 @@ Atomic_Cond :: struct {
 	state: Futex,
 }
 
-atomic_cond_wait :: proc(c: ^Atomic_Cond, m: ^Atomic_Mutex) {
+atomic_cond_wait :: proc "contextless" (c: ^Atomic_Cond, m: ^Atomic_Mutex) {
 	state := u32(atomic_load_explicit(&c.state, .Relaxed))
 	unlock(m)
 	futex_wait(&c.state, state)
@@ -297,7 +297,7 @@ atomic_cond_wait :: proc(c: ^Atomic_Cond, m: ^Atomic_Mutex) {
 
 }
 
-atomic_cond_wait_with_timeout :: proc(c: ^Atomic_Cond, m: ^Atomic_Mutex, duration: time.Duration) -> (ok: bool) {
+atomic_cond_wait_with_timeout :: proc "contextless" (c: ^Atomic_Cond, m: ^Atomic_Mutex, duration: time.Duration) -> (ok: bool) {
 	state := u32(atomic_load_explicit(&c.state, .Relaxed))
 	unlock(m)
 	ok = futex_wait_with_timeout(&c.state, state, duration)
@@ -306,12 +306,12 @@ atomic_cond_wait_with_timeout :: proc(c: ^Atomic_Cond, m: ^Atomic_Mutex, duratio
 }
 
 
-atomic_cond_signal :: proc(c: ^Atomic_Cond) {
+atomic_cond_signal :: proc "contextless" (c: ^Atomic_Cond) {
 	atomic_add_explicit(&c.state, 1, .Release)
 	futex_signal(&c.state)
 }
 
-atomic_cond_broadcast :: proc(c: ^Atomic_Cond) {
+atomic_cond_broadcast :: proc "contextless" (c: ^Atomic_Cond) {
 	atomic_add_explicit(&c.state, 1, .Release)
 	futex_broadcast(&c.state)
 }
@@ -324,7 +324,7 @@ Atomic_Sema :: struct {
 	count: Futex,
 }
 
-atomic_sema_post :: proc(s: ^Atomic_Sema, count := 1) {
+atomic_sema_post :: proc "contextless" (s: ^Atomic_Sema, count := 1) {
 	atomic_add_explicit(&s.count, Futex(count), .Release)
 	if count == 1 {
 		futex_signal(&s.count)
@@ -333,7 +333,7 @@ atomic_sema_post :: proc(s: ^Atomic_Sema, count := 1) {
 	}
 }
 
-atomic_sema_wait :: proc(s: ^Atomic_Sema) {
+atomic_sema_wait :: proc "contextless" (s: ^Atomic_Sema) {
 	for {
 		original_count := atomic_load_explicit(&s.count, .Relaxed)
 		for original_count == 0 {
@@ -346,7 +346,7 @@ atomic_sema_wait :: proc(s: ^Atomic_Sema) {
 	}
 }
 
-atomic_sema_wait_with_timeout :: proc(s: ^Atomic_Sema, duration: time.Duration) -> bool {
+atomic_sema_wait_with_timeout :: proc "contextless" (s: ^Atomic_Sema, duration: time.Duration) -> bool {
 	if duration <= 0 {
 		return false
 	}

core/sync/primitives_internal.odin

@@ -2,20 +2,31 @@
 package sync
 
 import "core:time"
+import vg "core:sys/valgrind"
+_ :: vg
 
 _Sema :: struct {
 	atomic: Atomic_Sema,
 }
 
-_sema_post :: proc(s: ^Sema, count := 1) {
+_sema_post :: proc "contextless" (s: ^Sema, count := 1) {
+	when ODIN_VALGRIND_SUPPORT {
+		vg.helgrind_sem_post_pre(s)
+	}
 	atomic_sema_post(&s.impl.atomic, count)
 }
 
-_sema_wait :: proc(s: ^Sema) {
+_sema_wait :: proc "contextless" (s: ^Sema) {
 	atomic_sema_wait(&s.impl.atomic)
+	when ODIN_VALGRIND_SUPPORT {
+		vg.helgrind_sem_wait_post(s)
+	}
 }
 
-_sema_wait_with_timeout :: proc(s: ^Sema, duration: time.Duration) -> bool {
+_sema_wait_with_timeout :: proc "contextless" (s: ^Sema, duration: time.Duration) -> bool {
+	when ODIN_VALGRIND_SUPPORT {
+		defer vg.helgrind_sem_wait_post(s)
+	}
 	return atomic_sema_wait_with_timeout(&s.impl.atomic, duration)
 }
 
@@ -25,7 +36,12 @@ _Recursive_Mutex :: struct {
 	recursion: i32,
 }
 
-_recursive_mutex_lock :: proc(m: ^Recursive_Mutex) {
+_recursive_mutex_lock :: proc "contextless" (m: ^Recursive_Mutex) {
+	when ODIN_VALGRIND_SUPPORT {
+		vg.helgrind_mutex_lock_pre(m, false)
+		defer vg.helgrind_mutex_lock_post(m)
+	}
+
 	tid := Futex(current_thread_id())
 	for {
 		prev_owner := atomic_compare_exchange_strong_explicit(&m.impl.owner, 0, tid, .Acquire, .Acquire)
@@ -40,7 +56,12 @@ _recursive_mutex_lock :: proc(m: ^Recursive_Mutex) {
 	}
 }
 
-_recursive_mutex_unlock :: proc(m: ^Recursive_Mutex) {
+_recursive_mutex_unlock :: proc "contextless" (m: ^Recursive_Mutex) {
+	when ODIN_VALGRIND_SUPPORT {
+		vg.helgrind_mutex_unlock_pre(m)
+		defer vg.helgrind_mutex_unlock_post(m)
+	}
+
 	m.impl.recursion -= 1
 	if m.impl.recursion != 0 {
 		return
@@ -52,7 +73,7 @@ _recursive_mutex_unlock :: proc(m: ^Recursive_Mutex) {
 
 }
 
-_recursive_mutex_try_lock :: proc(m: ^Recursive_Mutex) -> bool {
+_recursive_mutex_try_lock :: proc "contextless" (m: ^Recursive_Mutex) -> bool {
 	tid := Futex(current_thread_id())
 	prev_owner := atomic_compare_exchange_strong_explicit(&m.impl.owner, 0, tid, .Acquire, .Acquire)
 	switch prev_owner {
@@ -70,15 +91,27 @@ when ODIN_OS != .Windows {
 		mutex: Atomic_Mutex,
 	}
 
-	_mutex_lock :: proc(m: ^Mutex) {
+	_mutex_lock :: proc "contextless" (m: ^Mutex) {
+		when ODIN_VALGRIND_SUPPORT {
+			vg.helgrind_mutex_lock_pre(m, false)
+			defer vg.helgrind_mutex_lock_post(m)
+		}
 		atomic_mutex_lock(&m.impl.mutex)
 	}
 
-	_mutex_unlock :: proc(m: ^Mutex) {
+	_mutex_unlock :: proc "contextless" (m: ^Mutex) {
+		when ODIN_VALGRIND_SUPPORT {
+			vg.helgrind_mutex_unlock_pre(m)
+			defer vg.helgrind_mutex_unlock_post(m)
+		}
 		atomic_mutex_unlock(&m.impl.mutex)
 	}
 
-	_mutex_try_lock :: proc(m: ^Mutex) -> bool {
+	_mutex_try_lock :: proc "contextless" (m: ^Mutex) -> bool {
+		when ODIN_VALGRIND_SUPPORT {
+			vg.helgrind_mutex_lock_pre(m, true)
+			defer vg.helgrind_mutex_lock_post(m)
+		}
 		return atomic_mutex_try_lock(&m.impl.mutex)
 	}
 
@@ -86,19 +119,33 @@ when ODIN_OS != .Windows {
 		cond: Atomic_Cond,
 	}
 
-	_cond_wait :: proc(c: ^Cond, m: ^Mutex) {
+	_cond_wait :: proc "contextless" (c: ^Cond, m: ^Mutex) {
+		when ODIN_VALGRIND_SUPPORT {
+			_ = vg.helgrind_cond_wait_pre(c, m)
+			defer _ = vg.helgrind_cond_wait_post(c, m)
+		}
 		atomic_cond_wait(&c.impl.cond, &m.impl.mutex)
 	}
 
-	_cond_wait_with_timeout :: proc(c: ^Cond, m: ^Mutex, duration: time.Duration) -> bool {
+	_cond_wait_with_timeout :: proc "contextless" (c: ^Cond, m: ^Mutex, duration: time.Duration) -> bool {
+		when ODIN_VALGRIND_SUPPORT {
+			_ = vg.helgrind_cond_wait_pre(c, m)
+			defer _ = vg.helgrind_cond_wait_post(c, m)
+		}
 		return atomic_cond_wait_with_timeout(&c.impl.cond, &m.impl.mutex, duration)
 	}
 
-	_cond_signal :: proc(c: ^Cond) {
+	_cond_signal :: proc "contextless" (c: ^Cond) {
+		when ODIN_VALGRIND_SUPPORT {
+			vg.helgrind_cond_signal_pre(c)
+		}
 		atomic_cond_signal(&c.impl.cond)
 	}
 
-	_cond_broadcast :: proc(c: ^Cond) {
+	_cond_broadcast :: proc "contextless" (c: ^Cond) {
+		when ODIN_VALGRIND_SUPPORT {
+			vg.helgrind_cond_broadcast_pre(c)
+		}
 		atomic_cond_broadcast(&c.impl.cond)
 	}
 
@@ -107,27 +154,39 @@ when ODIN_OS != .Windows {
 		mutex: Atomic_RW_Mutex,
 	}
 
-	_rw_mutex_lock :: proc(rw: ^RW_Mutex) {
+	_rw_mutex_lock :: proc "contextless" (rw: ^RW_Mutex) {
+		when ODIN_VALGRIND_SUPPORT {
+			vg.helgrind_rwlock_lock_pre(rw, true)
+		}
 		atomic_rw_mutex_lock(&rw.impl.mutex)
 	}
 
-	_rw_mutex_unlock :: proc(rw: ^RW_Mutex) {
+	_rw_mutex_unlock :: proc "contextless" (rw: ^RW_Mutex) {
 		atomic_rw_mutex_unlock(&rw.impl.mutex)
+		when ODIN_VALGRIND_SUPPORT {
+			vg.helgrind_rwlock_unlock_post(rw, true)
+		}
 	}
 
-	_rw_mutex_try_lock :: proc(rw: ^RW_Mutex) -> bool {
+	_rw_mutex_try_lock :: proc "contextless" (rw: ^RW_Mutex) -> bool {
 		return atomic_rw_mutex_try_lock(&rw.impl.mutex)
 	}
 
-	_rw_mutex_shared_lock :: proc(rw: ^RW_Mutex) {
+	_rw_mutex_shared_lock :: proc "contextless" (rw: ^RW_Mutex) {
+		when ODIN_VALGRIND_SUPPORT {
+			vg.helgrind_rwlock_lock_pre(rw, false)
+		}
 		atomic_rw_mutex_shared_lock(&rw.impl.mutex)
 	}
 
-	_rw_mutex_shared_unlock :: proc(rw: ^RW_Mutex) {
+	_rw_mutex_shared_unlock :: proc "contextless" (rw: ^RW_Mutex) {
 		atomic_rw_mutex_shared_unlock(&rw.impl.mutex)
+		when ODIN_VALGRIND_SUPPORT {
+			vg.helgrind_rwlock_unlock_post(rw, false)
+		}
 	}
 
-	_rw_mutex_try_shared_lock :: proc(rw: ^RW_Mutex) -> bool {
+	_rw_mutex_try_shared_lock :: proc "contextless" (rw: ^RW_Mutex) -> bool {
 		return atomic_rw_mutex_try_shared_lock(&rw.impl.mutex)
 	}
 }

core/sync/primitives_windows.odin

@@ -13,15 +13,15 @@ _Mutex :: struct {
 	srwlock: win32.SRWLOCK,
 }
 
-_mutex_lock :: proc(m: ^Mutex) {
+_mutex_lock :: proc "contextless" (m: ^Mutex) {
 	win32.AcquireSRWLockExclusive(&m.impl.srwlock)
 }
 
-_mutex_unlock :: proc(m: ^Mutex) {
+_mutex_unlock :: proc "contextless" (m: ^Mutex) {
 	win32.ReleaseSRWLockExclusive(&m.impl.srwlock)
 }
 
-_mutex_try_lock :: proc(m: ^Mutex) -> bool {
+_mutex_try_lock :: proc "contextless" (m: ^Mutex) -> bool {
 	return bool(win32.TryAcquireSRWLockExclusive(&m.impl.srwlock))
 }
 
@@ -29,27 +29,27 @@ _RW_Mutex :: struct {
 	srwlock: win32.SRWLOCK,
 }
 
-_rw_mutex_lock :: proc(rw: ^RW_Mutex) {
+_rw_mutex_lock :: proc "contextless" (rw: ^RW_Mutex) {
 	win32.AcquireSRWLockExclusive(&rw.impl.srwlock)
 }
 
-_rw_mutex_unlock :: proc(rw: ^RW_Mutex) {
+_rw_mutex_unlock :: proc "contextless" (rw: ^RW_Mutex) {
 	win32.ReleaseSRWLockExclusive(&rw.impl.srwlock)
 }
 
-_rw_mutex_try_lock :: proc(rw: ^RW_Mutex) -> bool {
+_rw_mutex_try_lock :: proc "contextless" (rw: ^RW_Mutex) -> bool {
 	return bool(win32.TryAcquireSRWLockExclusive(&rw.impl.srwlock))
 }
 
-_rw_mutex_shared_lock :: proc(rw: ^RW_Mutex) {
+_rw_mutex_shared_lock :: proc "contextless" (rw: ^RW_Mutex) {
 	win32.AcquireSRWLockShared(&rw.impl.srwlock)
 }
 
-_rw_mutex_shared_unlock :: proc(rw: ^RW_Mutex) {
+_rw_mutex_shared_unlock :: proc "contextless" (rw: ^RW_Mutex) {
 	win32.ReleaseSRWLockShared(&rw.impl.srwlock)
 }
 
-_rw_mutex_try_shared_lock :: proc(rw: ^RW_Mutex) -> bool {
+_rw_mutex_try_shared_lock :: proc "contextless" (rw: ^RW_Mutex) -> bool {
 	return bool(win32.TryAcquireSRWLockShared(&rw.impl.srwlock))
 }
 
@@ -58,22 +58,22 @@ _Cond :: struct {
 	cond: win32.CONDITION_VARIABLE,
 }
 
-_cond_wait :: proc(c: ^Cond, m: ^Mutex) {
+_cond_wait :: proc "contextless" (c: ^Cond, m: ^Mutex) {
 	_ = win32.SleepConditionVariableSRW(&c.impl.cond, &m.impl.srwlock, win32.INFINITE, 0)
 }
 
-_cond_wait_with_timeout :: proc(c: ^Cond, m: ^Mutex, duration: time.Duration) -> bool {
+_cond_wait_with_timeout :: proc "contextless" (c: ^Cond, m: ^Mutex, duration: time.Duration) -> bool {
 	duration := u32(duration / time.Millisecond)
 	ok := win32.SleepConditionVariableSRW(&c.impl.cond, &m.impl.srwlock, duration, 0)
 	return bool(ok)
 }
 
 
-_cond_signal :: proc(c: ^Cond) {
+_cond_signal :: proc "contextless" (c: ^Cond) {
 	win32.WakeConditionVariable(&c.impl.cond)
 }
 
-_cond_broadcast :: proc(c: ^Cond) {
+_cond_broadcast :: proc "contextless" (c: ^Cond) {
 	win32.WakeAllConditionVariable(&c.impl.cond)
 }
 

core/sys/darwin/xnu_system_call_helpers.odin

@@ -97,6 +97,7 @@ clone_to_cstring :: proc(s: string, allocator: runtime.Allocator, loc := #caller
 
 
 sys_open :: proc(path: string, oflag: Open_Flags, mode: Permission) -> (c.int, bool) {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	
 	cmode: u32 = 0
 	cflags: u32 = 0
@@ -132,30 +133,35 @@ sys_open :: proc(path: string, oflag: Open_Flags, mode: Permission) -> (c.int, b
 }
 
 sys_mkdir :: proc(path: string, mode: Permission) -> bool {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	cpath: cstring = clone_to_cstring(path, context.temp_allocator)
 	cflags := _sys_permission_mode(mode)
 	return syscall_mkdir(cpath, cflags) != -1
 }
 
 sys_mkdir_at :: proc(fd: c.int, path: string, mode: Permission) -> bool {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	cpath: cstring = clone_to_cstring(path, context.temp_allocator)
 	cflags := _sys_permission_mode(mode)
 	return syscall_mkdir_at(fd, cpath, cflags) != -1
 }
 
 sys_rmdir :: proc(path: string, mode: Permission) -> bool {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	cpath: cstring = clone_to_cstring(path, context.temp_allocator)
 	cflags := _sys_permission_mode(mode)
 	return syscall_rmdir(cpath, cflags) != -1
 }
 
 sys_rename :: proc(path: string, new_path: string) -> bool {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	cpath: cstring = clone_to_cstring(path, context.temp_allocator)
 	cnpath: cstring = clone_to_cstring(new_path, context.temp_allocator)
 	return syscall_rename(cpath, cnpath) != -1
 }
 
 sys_rename_at :: proc(fd: c.int, path: string, to_fd: c.int, new_path: string) -> bool {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	cpath: cstring = clone_to_cstring(path, context.temp_allocator)
 	cnpath: cstring = clone_to_cstring(new_path, context.temp_allocator)
 	return syscall_rename_at(fd, cpath, to_fd, cnpath) != -1
@@ -166,12 +172,14 @@ sys_lseek :: proc(fd: c.int, offset: i64, whence: Offset_From) -> i64 {
 }
 
 sys_chmod :: proc(path: string, mode: Permission) -> bool {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	cpath: cstring = clone_to_cstring(path, context.temp_allocator)
 	cmode := _sys_permission_mode(mode)
 	return syscall_chmod(cpath, cmode) != -1
 }
 
 sys_lstat :: proc(path: string, status: ^stat) -> bool {
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
 	cpath: cstring = clone_to_cstring(path, context.temp_allocator)
 	return syscall_lstat(cpath, status) != -1
 }

core/sys/darwin/xnu_system_call_numbers.odin

@@ -1,6 +1,6 @@
 package darwin
 
-unix_offset_syscall :: proc(number: System_Call_Number) -> uintptr {
+unix_offset_syscall :: proc "contextless" (number: System_Call_Number) -> uintptr {
     return uintptr(number) + uintptr(0x2000000)
 }
 

core/sys/darwin/xnu_system_call_wrappers.odin

@@ -229,191 +229,191 @@ _Proc_Bsdinfo :: struct {
 
 /*--==========================================================================--*/
 
-syscall_fsync :: #force_inline proc(fildes: c.int) -> bool {
+syscall_fsync :: #force_inline proc "contextless" (fildes: c.int) -> bool {
 	return !(cast(bool)intrinsics.syscall(unix_offset_syscall(.fsync), uintptr(fildes)))
 }
 
-syscall_write :: #force_inline proc (fildes: c.int, buf: ^byte, nbyte: u64) -> bool {
+syscall_write :: #force_inline proc "contextless" (fildes: c.int, buf: ^byte, nbyte: u64) -> bool {
 	return !(cast(bool)intrinsics.syscall(unix_offset_syscall(.write),  uintptr(fildes), uintptr(buf), uintptr(nbyte)))
 }
  
-syscall_read :: #force_inline proc(fildes: c.int, buf: ^byte, nbyte: u64) -> i64 {
+syscall_read :: #force_inline proc "contextless" (fildes: c.int, buf: ^byte, nbyte: u64) -> i64 {
 	return cast(i64)intrinsics.syscall(unix_offset_syscall(.read), uintptr(fildes), uintptr(buf), uintptr(nbyte))
 }
 
-syscall_open :: #force_inline proc(path: cstring, oflag: u32, mode: u32) -> c.int {
+syscall_open :: #force_inline proc "contextless" (path: cstring, oflag: u32, mode: u32) -> c.int {
 	return cast(c.int)intrinsics.syscall(unix_offset_syscall(.open), transmute(uintptr)path, uintptr(oflag), uintptr(mode))
 }
 
-syscall_close :: #force_inline proc(fd: c.int) -> bool {
+syscall_close :: #force_inline proc "contextless" (fd: c.int) -> bool {
 	return !(cast(bool)intrinsics.syscall(unix_offset_syscall(.close), uintptr(fd)))
 }
 
-syscall_fchmod :: #force_inline proc(fildes: c.int, mode: u32) -> c.int {
+syscall_fchmod :: #force_inline proc "contextless" (fildes: c.int, mode: u32) -> c.int {
 	return (cast(c.int)intrinsics.syscall(unix_offset_syscall(.fchmod), uintptr(fildes), uintptr(mode)))
 }
 
-syscall_chmod :: #force_inline proc(path: cstring, mode: u32) -> c.int {
+syscall_chmod :: #force_inline proc "contextless" (path: cstring, mode: u32) -> c.int {
 	return (cast(c.int)intrinsics.syscall(unix_offset_syscall(.chmod), transmute(uintptr)path, uintptr(mode)))
 }
 
-syscall_mkdir :: #force_inline proc(path: cstring, mode: u32) -> c.int {
+syscall_mkdir :: #force_inline proc "contextless" (path: cstring, mode: u32) -> c.int {
 	return (cast(c.int)intrinsics.syscall(unix_offset_syscall(.mkdir), transmute(uintptr)path, uintptr(mode)))
 }
 
-syscall_mkdir_at :: #force_inline proc(fd: c.int, path: cstring, mode: u32) -> c.int {
+syscall_mkdir_at :: #force_inline proc "contextless" (fd: c.int, path: cstring, mode: u32) -> c.int {
 	return (cast(c.int)intrinsics.syscall(unix_offset_syscall(.mkdir), uintptr(fd), transmute(uintptr)path, uintptr(mode)))
 }
 
-syscall_rmdir :: #force_inline proc(path: cstring, mode: u32) -> c.int {
+syscall_rmdir :: #force_inline proc "contextless" (path: cstring, mode: u32) -> c.int {
 	return (cast(c.int)intrinsics.syscall(unix_offset_syscall(.rmdir), transmute(uintptr)path, uintptr(mode)))
 }
 
-syscall_rename :: #force_inline proc(path_old: cstring, path_new: cstring) -> c.int {
+syscall_rename :: #force_inline proc "contextless" (path_old: cstring, path_new: cstring) -> c.int {
 	return (cast(c.int)intrinsics.syscall(unix_offset_syscall(.rename), transmute(uintptr)path_old, transmute(uintptr)path_new))
 }
 
-syscall_rename_at :: #force_inline proc(from_fd: c.int, from: cstring, to_fd: c.int, to: cstring) -> c.int {
+syscall_rename_at :: #force_inline proc "contextless" (from_fd: c.int, from: cstring, to_fd: c.int, to: cstring) -> c.int {
 	return (cast(c.int)intrinsics.syscall(unix_offset_syscall(.renameat), uintptr(from_fd), transmute(uintptr)from, uintptr(to_fd), transmute(uintptr)to))
 }
 
-syscall_lseek :: #force_inline proc(fd: c.int, offset: i64, whence: c.int) -> i64 {
+syscall_lseek :: #force_inline proc "contextless" (fd: c.int, offset: i64, whence: c.int) -> i64 {
 	return cast(i64)intrinsics.syscall(unix_offset_syscall(.lseek), uintptr(fd), uintptr(offset), uintptr(whence))
 }
 
-syscall_gettid :: #force_inline proc() -> u64 {
+syscall_gettid :: #force_inline proc "contextless" () -> u64 {
 	return cast(u64)intrinsics.syscall(unix_offset_syscall(.gettid))
 }
 
-syscall_fstat :: #force_inline proc(fd: c.int, status: ^stat) -> c.int {
+syscall_fstat :: #force_inline proc "contextless" (fd: c.int, status: ^stat) -> c.int {
 	return cast(c.int)intrinsics.syscall(unix_offset_syscall(.fstat), uintptr(fd), uintptr(status))
 }
 
-syscall_lstat :: #force_inline proc(path: cstring, status: ^stat) -> c.int {
+syscall_lstat :: #force_inline proc "contextless" (path: cstring, status: ^stat) -> c.int {
 	return cast(c.int)intrinsics.syscall(unix_offset_syscall(.lstat), transmute(uintptr)path, uintptr(status))
 }
 
-syscall_stat :: #force_inline proc(path: cstring, status: ^stat) -> c.int {
+syscall_stat :: #force_inline proc "contextless" (path: cstring, status: ^stat) -> c.int {
 	return cast(c.int)intrinsics.syscall(unix_offset_syscall(.stat), transmute(uintptr)path, uintptr(status))
 }
 
-syscall_fstatat :: #force_inline proc(fd: c.int, path: cstring, status: ^stat) -> c.int {
+syscall_fstatat :: #force_inline proc "contextless" (fd: c.int, path: cstring, status: ^stat) -> c.int {
 	return cast(c.int)intrinsics.syscall(unix_offset_syscall(.fstatat), uintptr(fd), transmute(uintptr)path, uintptr(status))
 }
 
-syscall_link :: #force_inline proc(path: cstring, to_link: cstring) -> c.int {
+syscall_link :: #force_inline proc "contextless" (path: cstring, to_link: cstring) -> c.int {
 	return cast(c.int)intrinsics.syscall(unix_offset_syscall(.link), transmute(uintptr)path, transmute(uintptr)to_link)
 }
 
-syscall_linkat :: #force_inline proc(fd: c.int, path: cstring, fd2: c.int, to_link: cstring) -> c.int {
+syscall_linkat :: #force_inline proc "contextless" (fd: c.int, path: cstring, fd2: c.int, to_link: cstring) -> c.int {
 	return cast(c.int)intrinsics.syscall(unix_offset_syscall(.linkat), uintptr(fd), transmute(uintptr)path, uintptr(fd2), transmute(uintptr)to_link)
 }
 
-syscall_readlink :: #force_inline proc(path: cstring, buf: ^u8, buf_size: u64) -> i64 {
+syscall_readlink :: #force_inline proc "contextless" (path: cstring, buf: ^u8, buf_size: u64) -> i64 {
 	return cast(i64)intrinsics.syscall(unix_offset_syscall(.readlink), transmute(uintptr)path, uintptr(buf), uintptr(buf_size))
 }
 
-syscall_readlinkat :: #force_inline proc(fd: c.int, path: cstring, buf: ^u8, buf_size: u64) -> i64 {
+syscall_readlinkat :: #force_inline proc "contextless" (fd: c.int, path: cstring, buf: ^u8, buf_size: u64) -> i64 {
 	return cast(i64)intrinsics.syscall(unix_offset_syscall(.readlinkat), uintptr(fd), transmute(uintptr)path, uintptr(buf), uintptr(buf_size))
 }
 
-syscall_access :: #force_inline proc(path: cstring, mode: c.int) -> c.int {
+syscall_access :: #force_inline proc "contextless" (path: cstring, mode: c.int) -> c.int {
 	return cast(c.int)intrinsics.syscall(unix_offset_syscall(.access), transmute(uintptr)path, uintptr(mode))
 }
 
-syscall_faccessat :: #force_inline proc(fd: c.int, path: cstring, mode: c.int, flag: c.int) -> c.int {
+syscall_faccessat :: #force_inline proc "contextless" (fd: c.int, path: cstring, mode: c.int, flag: c.int) -> c.int {
 	return cast(c.int)intrinsics.syscall(unix_offset_syscall(.faccessat), uintptr(fd), transmute(uintptr)path, uintptr(mode), uintptr(flag))
 }
 
-syscall_getdirentries :: #force_inline proc(fd: c.int, buf: ^u8, nbytes: c.int, base_pointer: ^u32) -> c.int {
+syscall_getdirentries :: #force_inline proc "contextless" (fd: c.int, buf: ^u8, nbytes: c.int, base_pointer: ^u32) -> c.int {
 	return cast(c.int)intrinsics.syscall(unix_offset_syscall(.getdirentries), uintptr(fd), uintptr(buf), uintptr(nbytes), uintptr(base_pointer))
 }
 
-syscall_truncate :: #force_inline proc (path: cstring, length: off_t) -> c.int {
+syscall_truncate :: #force_inline proc "contextless" (path: cstring, length: off_t) -> c.int {
 	return cast(c.int)intrinsics.syscall(unix_offset_syscall(.truncate), transmute(uintptr)path, uintptr(length))
 }
 
-syscall_ftruncate :: #force_inline proc (fd: c.int, length: off_t) -> c.int {
+syscall_ftruncate :: #force_inline proc "contextless" (fd: c.int, length: off_t) -> c.int {
 	return cast(c.int)intrinsics.syscall(unix_offset_syscall(.ftruncate), uintptr(fd), uintptr(length))
 }
 
-syscall_sysctl :: #force_inline proc (name: ^c.int, namelen: c.uint, oldp: rawptr, oldlenp: ^i64, newp: ^i8, newlen: i64) -> c.int {
+syscall_sysctl :: #force_inline proc "contextless" (name: ^c.int, namelen: c.uint, oldp: rawptr, oldlenp: ^i64, newp: ^i8, newlen: i64) -> c.int {
 	return cast(c.int)intrinsics.syscall(unix_offset_syscall(.sysctl), uintptr(name), uintptr(namelen), uintptr(oldp), uintptr(oldlenp), uintptr(newp), uintptr(newlen))
 }
 
-syscall_copyfile ::  #force_inline proc(from: cstring, to: cstring, state: rawptr, flags: u32) -> c.int {
+syscall_copyfile ::  #force_inline proc "contextless" (from: cstring, to: cstring, state: rawptr, flags: u32) -> c.int {
 	return cast(c.int)intrinsics.syscall(unix_offset_syscall(.copyfile), transmute(uintptr)from, transmute(uintptr)to, uintptr(state), uintptr(flags))
 } 
 
 // think about this? last arg should be more than one
-syscall_fcntl :: #force_inline proc(fd: c.int, cmd: c.int, other: rawptr) -> c.int {
+syscall_fcntl :: #force_inline proc "contextless" (fd: c.int, cmd: c.int, other: rawptr) -> c.int {
 	return cast(c.int)intrinsics.syscall(unix_offset_syscall(.fsctl), uintptr(fd), uintptr(cmd), uintptr(other))
 }
 
-syscall_exit :: #force_inline proc(code: c.int) {
+syscall_exit :: #force_inline proc "contextless" (code: c.int) {
 	intrinsics.syscall(unix_offset_syscall(.exit), uintptr(code))
 }
 
-syscall_kill :: #force_inline proc(pid: pid_t, sig: c.int) -> c.int {
+syscall_kill :: #force_inline proc "contextless" (pid: pid_t, sig: c.int) -> c.int {
 	return cast(c.int)intrinsics.syscall(unix_offset_syscall(.kill), uintptr(pid), uintptr(sig))
 }
 
-syscall_dup :: #force_inline proc(fd: c.int) -> c.int {
+syscall_dup :: #force_inline proc "contextless" (fd: c.int) -> c.int {
 	return cast(c.int)intrinsics.syscall(unix_offset_syscall(.dup), uintptr(fd))
 }
 
-syscall_execve :: #force_inline proc(path: cstring, argv: [^]cstring, env: [^]cstring) -> c.int {
+syscall_execve :: #force_inline proc "contextless" (path: cstring, argv: [^]cstring, env: [^]cstring) -> c.int {
 	return cast(c.int)intrinsics.syscall(unix_offset_syscall(.execve), transmute(uintptr)path, transmute(uintptr)argv, transmute(uintptr)env)
 }
 
-syscall_munmap :: #force_inline proc(addr: rawptr, len: u64) -> c.int {
+syscall_munmap :: #force_inline proc "contextless" (addr: rawptr, len: u64) -> c.int {
 	return cast(c.int)intrinsics.syscall(unix_offset_syscall(.mmap), uintptr(addr), uintptr(len))
 }
 
-syscall_mmap :: #force_inline proc(addr: ^u8, len: u64, port: c.int, flags: c.int, fd: int, offset: off_t) -> ^u8 {
+syscall_mmap :: #force_inline proc "contextless" (addr: ^u8, len: u64, port: c.int, flags: c.int, fd: int, offset: off_t) -> ^u8 {
 	return cast(^u8)intrinsics.syscall(unix_offset_syscall(.mmap), uintptr(addr), uintptr(len), uintptr(port), uintptr(flags), uintptr(fd), uintptr(offset))
 }
 
-syscall_flock :: #force_inline proc(fd: c.int, operation: c.int) -> c.int {
+syscall_flock :: #force_inline proc "contextless" (fd: c.int, operation: c.int) -> c.int {
 	return cast(c.int)intrinsics.syscall(unix_offset_syscall(.flock), uintptr(fd), uintptr(operation)) 
 }
 
-syscall_utimes :: #force_inline proc(path: cstring, times: ^timeval) -> c.int {
+syscall_utimes :: #force_inline proc "contextless" (path: cstring, times: ^timeval) -> c.int {
 	return cast(c.int)intrinsics.syscall(unix_offset_syscall(.utimes), transmute(uintptr)path, uintptr(times))
 }
 
-syscall_futimes :: #force_inline proc(fd: c.int, times: ^timeval) -> c.int {
+syscall_futimes :: #force_inline proc "contextless" (fd: c.int, times: ^timeval) -> c.int {
 	return cast(c.int)intrinsics.syscall(unix_offset_syscall(.futimes), uintptr(fd), uintptr(times))
 }
 
-syscall_adjtime :: #force_inline proc(delta: ^timeval, old_delta: ^timeval) -> c.int {
+syscall_adjtime :: #force_inline proc "contextless" (delta: ^timeval, old_delta: ^timeval) -> c.int {
 	return cast(c.int)intrinsics.syscall(unix_offset_syscall(.adjtime), uintptr(delta), uintptr(old_delta))
 }
 
-syscall_sysctlbyname :: #force_inline proc(name: cstring, oldp: rawptr, oldlenp: ^i64, newp: rawptr, newlen: i64) -> c.int {
+syscall_sysctlbyname :: #force_inline proc "contextless" (name: cstring, oldp: rawptr, oldlenp: ^i64, newp: rawptr, newlen: i64) -> c.int {
 	return cast(c.int)intrinsics.syscall(unix_offset_syscall(.sysctlbyname), transmute(uintptr)name, uintptr(oldp), uintptr(oldlenp), uintptr(newp), uintptr(newlen))
 }
 
-syscall_proc_info :: #force_inline proc(num: c.int, pid: u32, flavor: c.int, arg: u64, buffer: rawptr, buffer_size: c.int) -> c.int {
+syscall_proc_info :: #force_inline proc "contextless" (num: c.int, pid: u32, flavor: c.int, arg: u64, buffer: rawptr, buffer_size: c.int) -> c.int {
 	return cast(c.int)intrinsics.syscall(unix_offset_syscall(.proc_info), uintptr(num), uintptr(pid), uintptr(flavor), uintptr(arg), uintptr(buffer), uintptr(buffer_size))
 }
 
-syscall_openat :: #force_inline proc(fd: int, path: cstring, oflag: u32, mode: u32) -> c.int {
+syscall_openat :: #force_inline proc "contextless" (fd: int, path: cstring, oflag: u32, mode: u32) -> c.int {
 	return cast(c.int)intrinsics.syscall(unix_offset_syscall(.openat), uintptr(fd), transmute(uintptr)path, uintptr(oflag), uintptr(mode))
 } 
 
-syscall_getentropy :: #force_inline proc(buf: [^]u8, buflen: u64) -> c.int {
+syscall_getentropy :: #force_inline proc "contextless" (buf: [^]u8, buflen: u64) -> c.int {
 	return cast(c.int)intrinsics.syscall(unix_offset_syscall(.getentropy), uintptr(buf), uintptr(buflen))
 }
 
-syscall_pipe :: #force_inline proc(fds: [^]c.int) -> c.int {
+syscall_pipe :: #force_inline proc "contextless" (fds: [^]c.int) -> c.int {
 	return cast(c.int)intrinsics.syscall(unix_offset_syscall(.getentropy), uintptr(&fds[0]), uintptr(&fds[1]))
 }
 
-syscall_chdir :: #force_inline proc(path: cstring) -> c.int {
+syscall_chdir :: #force_inline proc "contextless" (path: cstring) -> c.int {
 	return cast(c.int)intrinsics.syscall(unix_offset_syscall(.getentropy), transmute(uintptr)path)
 }
 
-syscall_fchdir :: #force_inline proc(fd: c.int, path: cstring) -> c.int {
+syscall_fchdir :: #force_inline proc "contextless" (fd: c.int, path: cstring) -> c.int {
 	return cast(c.int)intrinsics.syscall(unix_offset_syscall(.getentropy), uintptr(fd), transmute(uintptr)path)
-}
+}

core/sys/info/platform_darwin.odin

@@ -4,6 +4,7 @@ package sysinfo
 import sys "core:sys/unix"
 import "core:strconv"
 import "core:strings"
+import "core:runtime"
 
 @(private)
 version_string_buf: [1024]u8
@@ -41,6 +42,8 @@ init_os_version :: proc () {
 
 		major_ok, minor_ok, patch_ok: bool
 
+		tmp := runtime.default_temp_allocator_temp_begin()
+
 		triplet := strings.split(string(cstring(&version_bits[0])), ".", context.temp_allocator)
 		if len(triplet) != 3 {
 			have_kernel_version = false
@@ -54,6 +57,8 @@ init_os_version :: proc () {
 			}
 		}
 
+		runtime.default_temp_allocator_temp_end(tmp)
+
 		if !have_kernel_version {
 			// We don't know the kernel version, but we do know the build
 			strings.write_string(&b, "macOS Unknown (build ")

core/sys/info/platform_freebsd.odin

@@ -4,6 +4,7 @@ package sysinfo
 import sys "core:sys/unix"
 import "core:strings"
 import "core:strconv"
+import "core:runtime"
 
 @(private)
 version_string_buf: [1024]u8
@@ -47,6 +48,8 @@ init_os_version :: proc () {
 		return
 	}
 
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
+
 	// Parse kernel version
 	release := string(cstring(raw_data(kernel_version_buf[:])))
 	version_bits := strings.split_n(release, "-", 2, context.temp_allocator)

core/sys/info/platform_linux.odin

@@ -4,6 +4,7 @@ package sysinfo
 import "core:c"
 import sys "core:sys/unix"
 import "core:intrinsics"
+import "core:runtime"
 import "core:os"
 import "core:strings"
 import "core:strconv"
@@ -69,6 +70,8 @@ init_os_version :: proc () {
 	l := strings.builder_len(b)
 	strings.write_string(&b, string(cstring(&uts.release[0])))
 
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
+
 	// Parse kernel version, as substrings of the version info in `version_string_buf`
 	version_bits := strings.split_n(strings.to_string(b)[l:], "-", 2, context.temp_allocator)
 	if len(version_bits) > 1 {

core/sys/info/platform_openbsd.odin

@@ -4,6 +4,7 @@ package sysinfo
 import sys "core:sys/unix"
 import "core:strings"
 import "core:strconv"
+import "core:runtime"
 
 @(private)
 version_string_buf: [1024]u8
@@ -32,7 +33,9 @@ init_os_version :: proc () {
 	version := string(cstring(raw_data(kernel_version_buf[:])))
 	strings.write_string(&b, version)
 
-	// // Parse kernel version
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
+
+	// Parse kernel version
 	triplet := strings.split(version, ".", context.temp_allocator)
 	if len(triplet) == 2 {
 		major, major_ok := strconv.parse_int(triplet[0])

core/sys/info/platform_windows.odin

@@ -7,6 +7,7 @@ import "core:strings"
 import "core:unicode/utf16"
 
 import "core:fmt"
+import "core:runtime"
 
 @(private)
 version_string_buf: [1024]u8
@@ -314,6 +315,8 @@ read_reg :: proc(hkey: sys.HKEY, subkey, val: string, $T: typeid) -> (res: T, ok
 		return {}, false
 	}
 
+	runtime.DEFAULT_TEMP_ALLOCATOR_TEMP_GUARD()
+
 	key_name_wide := make([]u16, BUF_SIZE, context.temp_allocator)
 	val_name_wide := make([]u16, BUF_SIZE, context.temp_allocator)
 

core/sys/unix/syscalls_linux.odin

@@ -1563,15 +1563,167 @@ MADV_WIPEONFORK  :: 18
 MADV_KEEPONFORK  :: 19
 MADV_HWPOISON    :: 100
 
+// pipe2 flags
+O_CLOEXEC :: 0o2000000
+
+// perf event data
+Perf_Sample :: struct #raw_union {
+	period:    u64,
+	frequency: u64,
+}
+Perf_Wakeup :: struct #raw_union {
+	events:    u32,
+	watermark: u32,
+}
+Perf_Field1 :: struct #raw_union {
+	breakpoint_addr: u64,
+	kprobe_func:     u64,
+	uprobe_path:     u64,
+	config1:         u64,
+}
+Perf_Field2 :: struct #raw_union {
+	breakpoint_len: u64,
+	kprobe_addr:    u64,
+	uprobe_offset:  u64,
+	config2:        u64,
+}
+Perf_Event_Attr :: struct #packed {
+	type:   u32,
+	size:   u32,
+	config: u64,
+	sample: Perf_Sample,
+	sample_type: u64,
+	read_format: u64,
+	flags:       Perf_Flags,
+	wakeup: Perf_Wakeup,
+	breakpoint_type: u32,
+	field1: Perf_Field1,
+	field2: Perf_Field2,
+	branch_sample_type: u64,
+	sample_regs_user:   u64,
+	sample_stack_user:  u32,
+	clock_id:           i32,
+	sample_regs_intr:   u64,
+	aux_watermark:      u32,
+	sample_max_stack:   u16,
+	_padding:           u16,
+}
+
+Perf_Event_Flags :: distinct bit_set[Perf_Event_Flag; u64]
+Perf_Event_Flag :: enum u64 {
+	Bit0               = 0,
+	Bit0_Is_Deprecated = 1,
+	User_Rdpmc         = 2,
+	User_Time          = 3,
+	User_Time_Zero     = 4,
+	User_Time_Short    = 5,
+}
+Perf_Capabilities :: struct #raw_union {
+	capabilities: u64,
+	flags: Perf_Event_Flags,
+}
+Perf_Event_mmap_Page :: struct #packed {
+	version:        u32,
+	compat_version: u32,
+	lock:           u32,
+	index:          u32,
+	offset:         i64,
+	time_enabled:   u64,
+	time_running:   u64,
+	cap: Perf_Capabilities,
+	pmc_width:      u16,
+	time_shift:     u16,
+	time_mult:      u32,
+	time_offset:    u64,
+	time_zero:      u64,
+	size:           u32,
+	reserved1:      u32,
+	time_cycles:    u64,
+	time_mask:      u64,
+	reserved2:      [116*8]u8,
+	data_head:      u64,
+	data_tail:      u64,
+	data_offset:    u64,
+	data_size:      u64,
+	aux_head:       u64,
+	aux_tail:       u64,
+	aux_offset:     u64,
+	aux_size:       u64,
+}
+
+Perf_Type_Id :: enum u32 {
+	Hardware   = 0,
+	Software   = 1,
+	Tracepoint = 2,
+	HW_Cache   = 3,
+	Raw        = 4,
+	Breakpoint = 5,
+}
+
+Perf_Hardware_Id :: enum u64 {
+	CPU_Cycles              = 0,
+	Instructions            = 1,
+	Cache_References        = 2,
+	Cache_Misses            = 3,
+	Branch_Instructions     = 4,
+	Branch_Misses           = 5,
+	Bus_Cycles              = 6,
+	Stalled_Cycles_Frontend = 7,
+	Stalled_Cycles_Backend  = 8,
+	Ref_CPU_Cycles          = 9,
+}
+
+Perf_Flags :: distinct bit_set[Perf_Flag; u64]
+Perf_Flag :: enum u64 {
+	Disabled       = 0,
+	Inherit        = 1,
+	Pinned         = 2,
+	Exclusive      = 3,
+	Exclude_User   = 4,
+	Exclude_Kernel = 5,
+	Exclude_HV     = 6,
+	Exclude_Idle   = 7,
+	mmap           = 8,
+	Comm           = 9,
+	Freq           = 10,
+	Inherit_Stat   = 11,
+	Enable_On_Exec = 12,
+	Task           = 13,
+	Watermark      = 14,
+	Precise_IP_0   = 15,
+	Precise_IP_1   = 16,
+	mmap_Data      = 17,
+	Sample_Id_All  = 18,
+	Exclude_Host   = 19,
+	Exclude_Guest  = 20,
+	Exclude_Callchain_Kernel = 21,
+	Exclude_Callchain_User   = 22,
+	mmap2          = 23,
+	Comm_Exec      = 24,
+	Use_Clockid    = 25,
+	Context_Switch = 26,
+	Write_Backward = 27,
+	Namespaces     = 28,
+	KSymbol        = 29,
+	BPF_Event      = 30,
+	Aux_Output     = 31,
+	CGroup         = 32,
+	Text_Poke      = 33,
+	Build_Id       = 34,
+	Inherit_Thread = 35,
+	Remove_On_Exec = 36,
+	Sigtrap        = 37,
+}
+
 sys_gettid :: proc "contextless" () -> int {
-	return cast(int)intrinsics.syscall(SYS_gettid)
+	return int(intrinsics.syscall(SYS_gettid))
 }
 
-sys_getrandom :: proc "contextless" (buf: [^]byte, buflen: int, flags: uint) -> int {
-	return cast(int)intrinsics.syscall(SYS_getrandom, uintptr(buf), uintptr(buflen), uintptr(flags))
+sys_getrandom :: proc "contextless" (buf: [^]byte, buflen: uint, flags: int) -> int {
+	return int(intrinsics.syscall(SYS_getrandom, uintptr(buf), uintptr(buflen), uintptr(flags)))
 }
 
-sys_open :: proc "contextless" (path: cstring, flags: int, mode: int = 0o000) -> int {
+sys_open :: proc "contextless" (path: cstring, flags: int, mode: uint = 0o000) -> int {
 	when ODIN_ARCH != .arm64 {
 		return int(intrinsics.syscall(SYS_open, uintptr(rawptr(path)), uintptr(flags), uintptr(mode)))
 	} else { // NOTE: arm64 does not have open
@@ -1579,7 +1731,7 @@ sys_open :: proc "contextless" (path: cstring, flags: int, mode: int = 0o000) ->
 	}
 }
 
-sys_openat :: proc "contextless" (dfd: int, path: cstring, flags: int, mode: int = 0o000) -> int {
+sys_openat :: proc "contextless" (dfd: int, path: cstring, flags: int, mode: uint = 0o000) -> int {
 	return int(intrinsics.syscall(SYS_openat, uintptr(dfd), uintptr(rawptr(path)), uintptr(flags), uintptr(mode)))
 }
 
@@ -1691,7 +1843,7 @@ sys_fchdir :: proc "contextless" (fd: int) -> int {
 	return int(intrinsics.syscall(SYS_fchdir, uintptr(fd)))
 }
 
-sys_chmod :: proc "contextless" (path: cstring, mode: int) -> int {
+sys_chmod :: proc "contextless" (path: cstring, mode: uint) -> int {
 	when ODIN_ARCH != .arm64 {
 		return int(intrinsics.syscall(SYS_chmod, uintptr(rawptr(path)), uintptr(mode)))
 	} else { // NOTE: arm64 does not have chmod
@@ -1699,7 +1851,7 @@ sys_chmod :: proc "contextless" (path: cstring, mode: int) -> int {
 	}
 }
 
-sys_fchmod :: proc "contextless" (fd: int, mode: int) -> int {
+sys_fchmod :: proc "contextless" (fd: int, mode: uint) -> int {
 	return int(intrinsics.syscall(SYS_fchmod, uintptr(fd), uintptr(mode)))
 }
 
@@ -1759,7 +1911,7 @@ sys_rmdir :: proc "contextless" (path: cstring) -> int {
 	}
 }
 
-sys_mkdir :: proc "contextless" (path: cstring, mode: int) -> int {
+sys_mkdir :: proc "contextless" (path: cstring, mode: uint) -> int {
 	when ODIN_ARCH != .arm64 {
 		return int(intrinsics.syscall(SYS_mkdir, uintptr(rawptr(path)), uintptr(mode)))
 	} else { // NOTE: arm64 does not have mkdir
@@ -1767,11 +1919,11 @@ sys_mkdir :: proc "contextless" (path: cstring, mode: int) -> int {
 	}
 }
 
-sys_mkdirat :: proc "contextless" (dfd: int, path: cstring, mode: int) -> int {
+sys_mkdirat :: proc "contextless" (dfd: int, path: cstring, mode: uint) -> int {
 	return int(intrinsics.syscall(SYS_mkdirat, uintptr(dfd), uintptr(rawptr(path)), uintptr(mode)))
 }
 
-sys_mknod :: proc "contextless" (path: cstring, mode: int, dev: int) -> int {
+sys_mknod :: proc "contextless" (path: cstring, mode: uint, dev: int) -> int {
 	when ODIN_ARCH != .arm64 {
 		return int(intrinsics.syscall(SYS_mknod, uintptr(rawptr(path)), uintptr(mode), uintptr(dev)))
 	} else { // NOTE: arm64 does not have mknod
@@ -1779,7 +1931,7 @@ sys_mknod :: proc "contextless" (path: cstring, mode: int, dev: int) -> int {
 	}
 }
 
-sys_mknodat :: proc "contextless" (dfd: int, path: cstring, mode: int, dev: int) -> int {
+sys_mknodat :: proc "contextless" (dfd: int, path: cstring, mode: uint, dev: int) -> int {
 	return int(intrinsics.syscall(SYS_mknodat, uintptr(dfd), uintptr(rawptr(path)), uintptr(mode), uintptr(dev)))
 }
 
@@ -1818,6 +1970,16 @@ sys_fork :: proc "contextless" () -> int {
 		return int(intrinsics.syscall(SYS_clone, SIGCHLD))
 	}
 }
+sys_pipe2 :: proc "contextless" (fds: rawptr, flags: int) -> int {
+	return int(intrinsics.syscall(SYS_pipe2, uintptr(fds), uintptr(flags)))
+}
+sys_dup2 :: proc "contextless" (oldfd: int, newfd: int) -> int {
+	when ODIN_ARCH != .arm64 {
+		return int(intrinsics.syscall(SYS_dup2, uintptr(oldfd), uintptr(newfd)))
+	} else {
+		return int(intrinsics.syscall(SYS_dup3, uintptr(oldfd), uintptr(newfd), 0))
+	}
+}
 
 sys_mmap :: proc "contextless" (addr: rawptr, length: uint, prot, flags, fd: int, offset: uintptr) -> int {
 	return int(intrinsics.syscall(SYS_mmap, uintptr(addr), uintptr(length), uintptr(prot), uintptr(flags), uintptr(fd), offset))
@@ -1846,6 +2008,14 @@ sys_utimensat :: proc "contextless" (dfd: int, path: cstring, times: rawptr, fla
 	return int(intrinsics.syscall(SYS_utimensat, uintptr(dfd), uintptr(rawptr(path)), uintptr(times), uintptr(flags)))
 }
 
+sys_perf_event_open :: proc "contextless" (event_attr: rawptr, pid: i32, cpu: i32, group_fd: i32, flags: u32) -> int {
+	return int(intrinsics.syscall(SYS_perf_event_open, uintptr(event_attr), uintptr(pid), uintptr(cpu), uintptr(group_fd), uintptr(flags)))
+}
+
+sys_personality :: proc(persona: u64) -> int {
+	return int(intrinsics.syscall(SYS_personality, uintptr(persona)))
+}
+
 get_errno :: proc "contextless" (res: int) -> i32 {
 	if res < 0 && res > -4096 {
 		return i32(-res)

core/sys/windows/dwmapi.odin

@@ -3,7 +3,45 @@ package sys_windows
 
 foreign import dwmapi "system:Dwmapi.lib"
 
+DWMWINDOWATTRIBUTE :: enum {
+	DWMWA_NCRENDERING_ENABLED,
+	DWMWA_NCRENDERING_POLICY,
+	DWMWA_TRANSITIONS_FORCEDISABLED,
+	DWMWA_ALLOW_NCPAINT,
+	DWMWA_CAPTION_BUTTON_BOUNDS,
+	DWMWA_NONCLIENT_RTL_LAYOUT,
+	DWMWA_FORCE_ICONIC_REPRESENTATION,
+	DWMWA_FLIP3D_POLICY,
+	DWMWA_EXTENDED_FRAME_BOUNDS,
+	DWMWA_HAS_ICONIC_BITMAP,
+	DWMWA_DISALLOW_PEEK,
+	DWMWA_EXCLUDED_FROM_PEEK,
+	DWMWA_CLOAK,
+	DWMWA_CLOAKED,
+	DWMWA_FREEZE_REPRESENTATION,
+	DWMWA_PASSIVE_UPDATE_MODE,
+	DWMWA_USE_HOSTBACKDROPBRUSH,
+	DWMWA_USE_IMMERSIVE_DARK_MODE = 20,
+	DWMWA_WINDOW_CORNER_PREFERENCE = 33,
+	DWMWA_BORDER_COLOR,
+	DWMWA_CAPTION_COLOR,
+	DWMWA_TEXT_COLOR,
+	DWMWA_VISIBLE_FRAME_BORDER_THICKNESS,
+	DWMWA_SYSTEMBACKDROP_TYPE,
+  	DWMWA_LAST,
+}
+
+DWMNCRENDERINGPOLICY :: enum {
+	DWMNCRP_USEWINDOWSTYLE,
+	DWMNCRP_DISABLED,
+	DWMNCRP_ENABLED,
+	DWMNCRP_LAST,
+}
+
 @(default_calling_convention="stdcall")
 foreign dwmapi {
 	DwmFlush :: proc() -> HRESULT ---
+	DwmIsCompositionEnabled :: proc(pfEnabled: ^BOOL) -> HRESULT ---
+	DwmExtendFrameIntoClientArea :: proc(hWnd: HWND, pMarInset: PMARGINS) -> HRESULT ---
+	DwmSetWindowAttribute :: proc(hWnd: HWND, dwAttribute: DWORD, pvAttribute: LPCVOID, cbAttribute: DWORD) -> HRESULT ---
 }

core/sys/windows/gdi32.odin

@@ -62,6 +62,7 @@ foreign gdi32 {
 
 	SetPixelFormat :: proc(hdc: HDC, format: c_int, ppfd: ^PIXELFORMATDESCRIPTOR) -> BOOL ---
 	ChoosePixelFormat :: proc(hdc: HDC, ppfd: ^PIXELFORMATDESCRIPTOR) -> c_int ---
+	DescribePixelFormat :: proc(hdc: HDC, iPixelFormat: c_int, nBytes: UINT, ppfd: ^PIXELFORMATDESCRIPTOR) -> c_int ---
 	SwapBuffers :: proc(HDC) -> BOOL ---
 
 	SetDCBrushColor :: proc(hdc: HDC, color: COLORREF) -> COLORREF ---
@@ -78,6 +79,8 @@ foreign gdi32 {
 	TextOutW :: proc(hdc: HDC, x, y: c_int, lpString: LPCWSTR, c: c_int) -> BOOL ---
 	GetTextExtentPoint32W :: proc(hdc: HDC, lpString: LPCWSTR, c: c_int, psizl: LPSIZE) -> BOOL ---
 	GetTextMetricsW :: proc(hdc: HDC, lptm: LPTEXTMETRICW) -> BOOL ---
+
+	CreateSolidBrush :: proc(color: COLORREF) -> HBRUSH ---
 }
 
 RGB :: #force_inline proc "contextless" (r, g, b: u8) -> COLORREF {

core/sys/windows/kernel32.odin

@@ -315,6 +315,13 @@ foreign kernel32 {
 		lpOverlapped: LPOVERLAPPED,
 		lpCompletionRoutine: LPOVERLAPPED_COMPLETION_ROUTINE,
 	) -> BOOL ---
+	FindFirstChangeNotificationW :: proc(
+		lpPathName: LPWSTR,
+		bWatchSubtree: BOOL,
+		dwNotifyFilter: DWORD,
+	) -> HANDLE ---
+	FindNextChangeNotification :: proc(hChangeHandle: HANDLE) -> BOOL ---
+	FindCloseChangeNotification :: proc(hChangeHandle: HANDLE) -> BOOL ---
 
 	InitializeSRWLock          :: proc(SRWLock: ^SRWLOCK) ---
 	AcquireSRWLockExclusive    :: proc(SRWLock: ^SRWLOCK) ---
@@ -363,6 +370,9 @@ foreign kernel32 {
 	GenerateConsoleCtrlEvent :: proc(dwCtrlEvent: DWORD, dwProcessGroupId: DWORD) -> BOOL ---
 	FreeConsole :: proc() -> BOOL ---
 	GetConsoleWindow :: proc() -> HWND ---
+	GetConsoleScreenBufferInfo :: proc(hConsoleOutput: HANDLE, lpConsoleScreenBufferInfo: PCONSOLE_SCREEN_BUFFER_INFO) -> BOOL ---
+	SetConsoleScreenBufferSize :: proc(hConsoleOutput: HANDLE, dwSize: COORD) -> BOOL ---
+	SetConsoleWindowInfo :: proc(hConsoleOutput: HANDLE, bAbsolute : BOOL, lpConsoleWindow: ^SMALL_RECT) -> BOOL ---
 
 	GetDiskFreeSpaceExW :: proc(
 		lpDirectoryName: LPCWSTR,
@@ -370,6 +380,8 @@ foreign kernel32 {
 		lpTotalNumberOfBytes: PULARGE_INTEGER,
 		lpTotalNumberOfFreeBytes: PULARGE_INTEGER,
 	) -> BOOL ---
+
+	GetLogicalProcessorInformation :: proc(buffer: ^SYSTEM_LOGICAL_PROCESSOR_INFORMATION, returnedLength: PDWORD) -> BOOL ---
 }
 
 
@@ -999,3 +1011,49 @@ foreign kernel32 {
 	ConvertThreadToFiber :: proc(lpParameter: LPVOID) -> LPVOID ---
 	SwitchToFiber :: proc(lpFiber: LPVOID) ---
 }
+
+LOGICAL_PROCESSOR_RELATIONSHIP :: enum c_int {
+	RelationProcessorCore,
+	RelationNumaNode,
+	RelationCache,
+	RelationProcessorPackage,
+	RelationGroup,
+	RelationProcessorDie,
+	RelationNumaNodeEx,
+	RelationProcessorModule,
+	RelationAll = 0xffff,
+}
+
+PROCESSOR_CACHE_TYPE :: enum c_int {
+	CacheUnified,
+	CacheInstruction,
+	CacheData,
+	CacheTrace,
+}
+
+CACHE_DESCRIPTOR :: struct {
+	Level: BYTE,
+	Associativity: BYTE,
+	LineSize: WORD,
+	Size: DWORD,
+	Type: PROCESSOR_CACHE_TYPE,
+}
+
+ProcessorCore :: struct {
+	Flags: BYTE,
+}
+NumaNode :: struct {
+	NodeNumber: DWORD,
+}
+DUMMYUNIONNAME_u :: struct #raw_union {
+	Core: ProcessorCore,
+	Node: NumaNode,
+	Cache: CACHE_DESCRIPTOR,
+	Reserved: [2]ULONGLONG,
+}
+
+SYSTEM_LOGICAL_PROCESSOR_INFORMATION :: struct {
+	ProcessorMask: ULONG_PTR,
+	Relationship: LOGICAL_PROCESSOR_RELATIONSHIP,
+	DummyUnion: DUMMYUNIONNAME_u,
+}

core/sys/windows/shell32.odin

@@ -22,4 +22,37 @@ foreign shell32 {
 	) -> c_int ---
 	SHFileOperationW :: proc(lpFileOp: LPSHFILEOPSTRUCTW) -> c_int ---
 	SHGetFolderPathW :: proc(hwnd: HWND, csidl: c_int, hToken: HANDLE, dwFlags: DWORD, pszPath: LPWSTR) -> HRESULT ---
+	SHAppBarMessage :: proc(dwMessage: DWORD, pData: PAPPBARDATA) -> UINT_PTR --- 
 }
+
+APPBARDATA :: struct {
+	cbSize: DWORD,
+	hWnd: HWND,
+	uCallbackMessage: UINT,
+	uEdge: UINT,
+	rc: RECT,
+	lParam: LPARAM,
+}
+PAPPBARDATA :: ^APPBARDATA
+ 
+ABM_NEW              :: 0x00000000
+ABM_REMOVE           :: 0x00000001
+ABM_QUERYPOS         :: 0x00000002
+ABM_SETPOS           :: 0x00000003
+ABM_GETSTATE         :: 0x00000004
+ABM_GETTASKBARPOS    :: 0x00000005
+ABM_ACTIVATE         :: 0x00000006 
+ABM_GETAUTOHIDEBAR   :: 0x00000007
+ABM_SETAUTOHIDEBAR   :: 0x00000008 
+ABM_WINDOWPOSCHANGED :: 0x0000009
+ABM_SETSTATE         :: 0x0000000a
+ABN_STATECHANGE      :: 0x0000000
+ABN_POSCHANGED       :: 0x0000001
+ABN_FULLSCREENAPP    :: 0x0000002
+ABN_WINDOWARRANGE    :: 0x0000003
+ABS_AUTOHIDE         :: 0x0000001
+ABS_ALWAYSONTOP      :: 0x0000002
+ABE_LEFT             :: 0
+ABE_TOP              :: 1
+ABE_RIGHT            :: 2
+ABE_BOTTOM           :: 3

core/sys/windows/types.odin

@@ -38,6 +38,7 @@ HHOOK :: distinct HANDLE
 HKEY :: distinct HANDLE
 HDESK :: distinct HANDLE
 HFONT :: distinct HANDLE
+HRGN :: distinct HANDLE
 BOOL :: distinct b32
 BYTE :: distinct u8
 BOOLEAN :: distinct b8
@@ -1554,6 +1555,25 @@ WA_INACTIVE    :: 0
 WA_ACTIVE      :: 1
 WA_CLICKACTIVE :: 2
 
+// Struct pointed to by WM_GETMINMAXINFO lParam
+MINMAXINFO :: struct {
+	ptReserved: POINT,
+	ptMaxSize: POINT,
+	ptMaxPosition: POINT,
+	ptMinTrackSize: POINT,
+	ptMaxTrackSize: POINT,
+}
+PMINMAXINFO  :: ^MINMAXINFO
+LPMINMAXINFO :: PMINMAXINFO
+
+MONITORINFO :: struct {
+	cbSize: DWORD,
+	rcMonitor: RECT,
+	rcWork: RECT,
+	dwFlags: DWORD,
+}
+LPMONITORINFO :: ^MONITORINFO
+
 // SetWindowsHook() codes
 WH_MIN             :: -1
 WH_MSGFILTER       :: -1
@@ -3864,3 +3884,21 @@ COORD :: struct {
 	X: SHORT,
 	Y: SHORT,
 }
+
+SMALL_RECT :: struct {
+	Left: SHORT,
+	Top: SHORT,
+	Right: SHORT,
+	Bottom: SHORT,
+}
+
+CONSOLE_SCREEN_BUFFER_INFO :: struct {
+	dwSize: COORD,
+	dwCursorPosition: COORD,
+	wAttributes: WORD,
+	srWindow: SMALL_RECT,
+	dwMaximumWindowSize: COORD,
+}
+
+
+PCONSOLE_SCREEN_BUFFER_INFO :: ^CONSOLE_SCREEN_BUFFER_INFO

core/sys/windows/user32.odin

@@ -100,6 +100,7 @@ foreign user32 {
 	AdjustWindowRectExForDpi :: proc(lpRect: LPRECT, dwStyle: DWORD, bMenu: BOOL, dwExStyle: DWORD, dpi: UINT) -> BOOL ---
 
 	SystemParametersInfoW :: proc(uiAction, uiParam: UINT, pvParam: PVOID, fWinIni: UINT) -> BOOL ---
+	GetMonitorInfoW :: proc(hMonitor: HMONITOR, lpmi: LPMONITORINFO) -> BOOL ---
 
 	GetWindowDC :: proc(hWnd: HWND) -> HDC ---
 	GetDC :: proc(hWnd: HWND) -> HDC ---
@@ -122,6 +123,8 @@ foreign user32 {
 
 	GetKeyState :: proc(nVirtKey: c_int) -> SHORT ---
 	GetAsyncKeyState :: proc(vKey: c_int) -> SHORT ---
+	
+	GetKeyboardState :: proc(lpKeyState: PBYTE) -> BOOL ---
 
 	MapVirtualKeyW :: proc(uCode: UINT, uMapType: UINT) -> UINT ---
 
@@ -202,6 +205,17 @@ foreign user32 {
 	GetRawInputDeviceList :: proc(pRawInputDeviceList: PRAWINPUTDEVICELIST, puiNumDevices: PUINT, cbSize: UINT) -> UINT ---
 	GetRegisteredRawInputDevices :: proc(pRawInputDevices: PRAWINPUTDEVICE, puiNumDevices: PUINT, cbSize: UINT) -> UINT ---
 	RegisterRawInputDevices :: proc(pRawInputDevices: PCRAWINPUTDEVICE, uiNumDevices: UINT, cbSize: UINT) -> BOOL ---
+
+	SetLayeredWindowAttributes  :: proc(hWnd: HWND, crKey: COLORREF, bAlpha: BYTE, dwFlags: DWORD) -> BOOL ---
+
+	FillRect :: proc(hDC: HDC, lprc: ^RECT, hbr: HBRUSH) -> int ---
+	EqualRect :: proc(lprc1: ^RECT, lprc2: ^RECT) -> BOOL ---
+
+	GetWindowInfo :: proc(hwnd: HWND, pwi: PWINDOWINFO) -> BOOL ---
+	GetWindowPlacement :: proc(hWnd: HWND, lpwndpl: ^WINDOWPLACEMENT) -> BOOL ---
+	SetWindowRgn :: proc(hWnd: HWND, hRgn: HRGN, bRedraw: BOOL) -> int ---
+	CreateRectRgnIndirect :: proc(lprect: ^RECT) -> HRGN ---
+	GetSystemMetricsForDpi :: proc(nIndex: int, dpi: UINT) -> int ---
 }
 
 CreateWindowW :: #force_inline proc "stdcall" (
@@ -432,3 +446,27 @@ RI_MOUSE_BUTTON_5_DOWN :: 0x0100
 RI_MOUSE_BUTTON_5_UP :: 0x0200
 RI_MOUSE_WHEEL :: 0x0400
 RI_MOUSE_HWHEEL :: 0x0800
+
+WINDOWPLACEMENT :: struct {
+	length: UINT,
+	flags: UINT,
+	showCmd: UINT,
+	ptMinPosition: POINT,
+  	ptMaxPosition: POINT,
+  	rcNormalPosition: RECT,
+  	rcDevice: RECT,
+}
+
+WINDOWINFO :: struct {
+	cbSize: DWORD,
+	rcWindow: RECT,
+	rcClient: RECT,
+	dwStyle: DWORD,
+	dwExStyle: DWORD,
+	dwWindowStatus: DWORD,
+	cxWindowBorders: UINT,
+	cyWindowBorders: UINT,
+	atomWindowType: ATOM,
+	wCreatorVersion: WORD,
+}
+PWINDOWINFO :: ^WINDOWINFO

core/sys/windows/ux_theme.odin

@@ -0,0 +1,12 @@
+// +build windows
+package sys_windows
+
+foreign import uxtheme "system:UxTheme.lib"
+
+MARGINS :: distinct [4]int
+PMARGINS :: ^MARGINS
+
+@(default_calling_convention="stdcall")
+foreign uxtheme {
+    IsThemeActive :: proc() -> BOOL ---
+}

core/text/match/strlib.odin

@@ -0,0 +1,959 @@
+package text_match
+
+import "core:runtime"
+import "core:unicode"
+import "core:unicode/utf8"
+import "core:strings"
+
+MAX_CAPTURES :: 32
+
+Capture :: struct {
+	init: int,
+	len: int,
+}
+
+Match :: struct {
+	byte_start, byte_end: int,
+}
+
+Error :: enum {
+	OK,
+	OOB,
+	Invalid_Capture_Index,
+	Invalid_Pattern_Capture,
+	Unfinished_Capture,
+	Malformed_Pattern,
+	Rune_Error,
+	Match_Invalid,
+}
+
+L_ESC :: '%'
+CAP_POSITION :: -2
+CAP_UNFINISHED :: -1
+INVALID :: -1
+
+Match_State :: struct {
+	src: string,
+	pattern: string,
+	level: int,
+	capture: [MAX_CAPTURES]Capture,
+}
+
+match_class :: proc(c: rune, cl: rune) -> (res: bool) {
+	switch unicode.to_lower(cl) {
+	case 'a': res = is_alpha(c)
+	case 'c': res = is_cntrl(c)
+	case 'd': res = is_digit(c)
+	case 'g': res = is_graph(c)
+	case 'l': res = is_lower(c)
+	case 'p': res = is_punct(c)
+	case 's': res = is_space(c)
+	case 'u': res = is_upper(c)
+	case 'w': res = is_alnum(c)
+	case 'x': res = is_xdigit(c)
+	case: return cl == c
+	}
+
+	return is_lower(cl) ? res : !res
+}
+
+is_alpha :: unicode.is_alpha
+is_digit :: unicode.is_digit
+is_lower :: unicode.is_lower
+is_upper :: unicode.is_upper
+is_punct :: unicode.is_punct
+is_space :: unicode.is_space
+is_cntrl :: unicode.is_control
+
+is_alnum :: proc(c: rune) -> bool {
+	return unicode.is_alpha(c) || unicode.is_digit(c)
+}
+
+is_graph :: proc(c: rune) -> bool {
+	return (c >= 'a' && c <= 'f') || (c >= 'A' && c <= 'F') || unicode.is_digit(c)
+}
+
+is_xdigit :: proc(c: rune) -> bool {
+	return (c >= 'a' && c <= 'f') || (c >= 'A' && c <= 'F') || unicode.is_digit(c)
+}
+
+// find the first utf8 charater and its size, return an error if the character is an error
+utf8_peek :: proc(bytes: string) -> (c: rune, size: int, err: Error) {
+	c, size = utf8.decode_rune_in_string(bytes)
+
+	if c == utf8.RUNE_ERROR {
+		err = .Rune_Error
+	}
+
+	return
+}
+
+// find the first utf8 charater and its size and advance the index
+// return an error if the character is an error
+utf8_advance :: proc(bytes: string, index: ^int) -> (c: rune, err: Error) {
+	size: int
+	c, size = utf8.decode_rune_in_string(bytes[index^:])
+
+	if c == utf8.RUNE_ERROR {
+		err = .Rune_Error
+	}
+
+	index^ += size
+	return
+}
+
+// continuation byte?
+is_cont :: proc(b: byte) -> bool {
+	return b & 0xc0 == 0x80
+}
+
+utf8_prev :: proc(bytes: string, a, b: int) -> int {
+	b := b
+
+	for a < b && is_cont(bytes[b - 1]) {
+		b -= 1
+	}
+
+	return a < b ? b - 1 : a
+}
+
+utf8_next :: proc(bytes: string, a: int) -> int {
+	a := a
+	b := len(bytes)
+
+	for a < b - 1 && is_cont(bytes[a + 1]) {
+		a += 1
+	}
+
+	return a < b ? a + 1 : b
+}
+
+check_capture :: proc(ms: ^Match_State, l: rune) -> (int, Error) {
+	l := int(l - '1')
+	
+	if l < 0 || l >= ms.level || ms.capture[l].len == CAP_UNFINISHED {
+		return 0, .Invalid_Capture_Index
+	}
+
+	return l, .OK
+}
+
+capture_to_close :: proc(ms: ^Match_State) -> (int, Error) {
+	level := ms.level - 1
+
+	for level >= 0 {
+		if ms.capture[level].len == CAP_UNFINISHED {
+			return level, .OK
+		}
+
+		level -= 1
+	}
+
+	return 0, .Invalid_Pattern_Capture
+}
+
+class_end :: proc(ms: ^Match_State, p: int) -> (step: int, err: Error) {
+	step = p
+	ch := utf8_advance(ms.pattern, &step) or_return
+
+	switch ch {
+	case L_ESC: 
+		if step == len(ms.pattern) {
+			err = .Malformed_Pattern
+			return
+		}
+
+		utf8_advance(ms.pattern, &step) or_return
+
+	case '[': 
+		// fine with step by 1
+		if step + 1 < len(ms.pattern) && ms.pattern[step] == '^' {
+			step += 1
+		}
+
+		// run till end is reached
+		for {
+			if step == len(ms.pattern) {
+				err = .Malformed_Pattern
+				return
+			}
+
+			if ms.pattern[step] == ']' {
+				break
+			}
+
+			// dont care about utf8 here
+			step += 1
+
+			if step < len(ms.pattern) && ms.pattern[step] == L_ESC {
+				// skip escapes like '%'
+				step += 1
+			}
+		}
+
+		// advance last time
+		step += 1
+	}
+
+	return
+}
+
+match_bracket_class :: proc(ms: ^Match_State, c: rune, p, ec: int) -> (sig: bool, err: Error) {
+	sig = true
+	p := p
+
+	if ms.pattern[p + 1] == '^' {
+		p += 1
+		sig = false
+	}
+
+	// while inside of class range
+	for p < ec {
+		char := utf8_advance(ms.pattern, &p) or_return
+
+		// e.g. %a
+		if char == L_ESC { 
+			next := utf8_advance(ms.pattern, &p) or_return
+
+			if match_class(c, next) {
+				return
+			}
+		} else {
+			next, next_size := utf8_peek(ms.pattern[p:]) or_return
+
+			// TODO test case for [a-???] where ??? is missing
+			if next == '-' && p + next_size < len(ms.pattern) {
+				// advance 2 codepoints
+				p += next_size
+				last := utf8_advance(ms.pattern, &p) or_return
+
+				if char <= c && c <= last {
+					return
+				}
+			} else if char == c {
+				return
+			}
+		}
+	}
+
+	sig = !sig
+	return
+}
+
+single_match :: proc(ms: ^Match_State, s, p, ep: int) -> (matched: bool, schar_size: int, err: Error) {
+	if s >= len(ms.src) {
+		return
+	}
+
+	pchar, psize := utf8_peek(ms.pattern[p:]) or_return
+	schar, ssize := utf8_peek(ms.src[s:]) or_return
+	schar_size = ssize
+
+	switch pchar {
+	case '.': matched = true
+	case L_ESC: 
+		pchar_next, _ := utf8_peek(ms.pattern[p + psize:]) or_return
+		matched = match_class(schar, pchar_next)
+	case '[': matched = match_bracket_class(ms, schar, p, ep - 1) or_return
+	case: matched = schar == pchar
+	}
+
+	return
+}
+
+match_balance :: proc(ms: ^Match_State, s, p: int) -> (unused: int, err: Error) {
+	if p >= len(ms.pattern) - 1 {
+		return INVALID, .Invalid_Pattern_Capture
+	}
+
+	schar, ssize := utf8_peek(ms.src[s:]) or_return
+	pchar, psize := utf8_peek(ms.pattern[p:]) or_return
+
+	// skip until the src and pattern match
+	if schar != pchar {
+		return INVALID, .OK
+	}
+
+	s_begin := s
+	cont := 1
+	s := s + ssize
+	begin := pchar
+	end, _ := utf8_peek(ms.pattern[p + psize:]) or_return
+
+	for s < len(ms.src) {
+		ch := utf8_advance(ms.src, &s) or_return
+
+		switch ch{
+		case end:
+			cont -= 1
+
+			if cont == 0 {
+				return s, .OK
+			}
+
+		case begin:
+			cont += 1
+		}
+	}
+
+	return INVALID, .OK
+}
+
+max_expand :: proc(ms: ^Match_State, s, p, ep: int) -> (res: int, err: Error) {
+	m := s
+
+	// count up matches
+	for {
+		matched, size := single_match(ms, m, p, ep) or_return
+		
+		if !matched {
+			break
+		}
+
+		m += size
+	}
+
+	for s <= m {
+		result := match(ms, m, ep + 1) or_return
+
+		if result != INVALID {
+			return result, .OK
+		}
+
+		if s == m {
+			break
+		}
+
+		m = utf8_prev(ms.src, s, m)
+	}
+
+	return INVALID, .OK
+}
+
+min_expand :: proc(ms: ^Match_State, s, p, ep: int) -> (res: int, err: Error) {
+	s := s
+
+	for {
+		result := match(ms, s, ep + 1) or_return
+
+		if result != INVALID {
+			return result, .OK
+		} else {
+			// TODO receive next step maybe?
+			matched, rune_size := single_match(ms, s, p, ep) or_return
+
+			if matched {
+				s += rune_size
+			} else {
+				return INVALID, .OK
+			}
+		}
+	}
+}
+
+start_capture :: proc(ms: ^Match_State, s, p, what: int) -> (res: int, err: Error) {
+	level := ms.level
+
+	ms.capture[level].init = s
+	ms.capture[level].len = what
+	ms.level += 1
+
+	res = match(ms, s, p) or_return
+	if res == INVALID {
+		ms.level -= 1
+	}
+	return
+}
+
+end_capture :: proc(ms: ^Match_State, s, p: int) -> (res: int, err: Error) {
+	l := capture_to_close(ms) or_return
+	
+	// TODO double check, could do string as int index
+	ms.capture[l].len = s - ms.capture[l].init
+
+	res = match(ms, s, p) or_return
+	if res == INVALID {
+		ms.capture[l].len = CAP_UNFINISHED
+	}
+	return
+}
+
+match_capture :: proc(ms: ^Match_State, s: int, char: rune) -> (res: int, err: Error) {
+	index := check_capture(ms, char) or_return
+	length := ms.capture[index].len
+
+	if len(ms.src) - s >= length {
+		return s + length, .OK
+	}
+
+	return INVALID, .OK
+}
+
+match :: proc(ms: ^Match_State, s, p: int) -> (unused: int, err: Error) {
+	s := s
+	p := p
+
+	if p == len(ms.pattern) {
+		return s, .OK
+	}
+
+	// NOTE we can walk by ascii steps if we know the characters are ascii
+	char, _ := utf8_peek(ms.pattern[p:]) or_return
+	switch char {
+	case '(': 
+		if p + 1 < len(ms.pattern) && ms.pattern[p + 1] == ')' {
+			s = start_capture(ms, s, p + 2, CAP_POSITION) or_return
+		} else {
+			s = start_capture(ms, s, p + 1, CAP_UNFINISHED) or_return
+		}
+
+	case ')': 
+		s = end_capture(ms, s, p + 1) or_return
+
+	case '$': 
+		if p + 1 != len(ms.pattern) {
+			return match_default(ms, s, p)
+		} 
+
+		if len(ms.src) != s {
+			s = INVALID
+		}
+
+	case L_ESC: 
+		// stop short patterns like "%" only
+		if p + 1 >= len(ms.pattern) {
+			err = .OOB
+			return
+		}
+
+		switch ms.pattern[p + 1] {
+		// balanced string
+		case 'b': 
+			s = match_balance(ms, s, p + 2) or_return
+
+			if s != INVALID {
+				// eg after %b()
+				return match(ms, s, p + 4)
+			}
+
+		// frontier
+		case 'f':
+			p += 2
+			
+			if ms.pattern[p] != '[' {
+				return INVALID, .Invalid_Pattern_Capture
+			}
+
+			ep := class_end(ms, p) or_return
+			previous, current: rune
+
+			// get previous
+			if s != 0 {
+				temp := utf8_prev(ms.src, 0, s)
+				previous, _ = utf8_peek(ms.src[temp:]) or_return
+			}
+
+			// get current
+			if s != len(ms.src) {
+				current, _ = utf8_peek(ms.src[s:]) or_return
+			}
+
+			m1 := match_bracket_class(ms, previous, p, ep - 1) or_return
+			m2 := match_bracket_class(ms, current, p, ep - 1) or_return
+
+			if !m1 && m2 {
+				return match(ms, s, ep)
+			}
+
+			s = INVALID
+
+		// capture group
+		case '0'..<'9':
+			s = match_capture(ms, s, rune(ms.pattern[p + 1])) or_return
+
+			if s != INVALID {
+				return match(ms, s, p + 2)
+			}
+
+		case: return match_default(ms, s, p)
+	}
+
+	case: 
+		return match_default(ms, s, p)
+	}
+
+	return s, .OK
+}
+
+match_default :: proc(ms: ^Match_State, s, p: int) -> (unused: int, err: Error) {
+	s := s
+	ep := class_end(ms, p) or_return
+	single_matched, ssize := single_match(ms, s, p, ep) or_return
+
+	if !single_matched {
+		epc := ep < len(ms.pattern) ? ms.pattern[ep] : 0
+
+		switch epc {
+		case '*', '?', '-': return match(ms, s, ep + 1)
+		case: s = INVALID
+		}
+	} else {
+		epc := ep < len(ms.pattern) ? ms.pattern[ep] : 0
+
+		switch epc {
+		case '?':
+			result := match(ms, s + ssize, ep + 1) or_return
+			
+			if result != INVALID {
+				s = result
+			} else {
+				return match(ms, s, ep + 1)
+			}
+
+		case '+': s = max_expand(ms, s + ssize, p, ep) or_return
+		case '*': s = max_expand(ms, s, p, ep) or_return
+		case '-': s = min_expand(ms, s, p, ep) or_return
+		case: return match(ms, s + ssize, ep)
+		}
+	}
+
+	return s, .OK
+}
+
+push_onecapture :: proc(ms: ^Match_State,  i: int,  s: int, e: int, matches: []Match) -> (err: Error) {
+	if i >= ms.level {
+		if i == 0 {
+			matches[0] = { 0, e - s }
+		} else {
+			err = .Invalid_Capture_Index
+		}
+	} else {
+		init := ms.capture[i].init 
+		length := ms.capture[i].len
+
+		switch length {
+		case CAP_UNFINISHED: err = .Unfinished_Capture
+		case CAP_POSITION: matches[i] = { init, init + 1 }
+		case: matches[i] = { init, init + length }
+		}
+	}
+
+	return
+}
+
+push_captures :: proc(
+	ms: ^Match_State,
+	s: int,
+	e: int,
+	matches: []Match,
+) -> (nlevels: int, err: Error) {
+	nlevels = 1 if ms.level == 0 && s != -1 else ms.level
+
+	for i in 0..<nlevels {
+		push_onecapture(ms, i, s, e, matches) or_return
+	}
+
+	return
+}
+
+// SPECIALS := "^$*+?.([%-"
+// all special characters inside a small ascii array
+SPECIALS_TABLE := [256]bool {
+	'^' = true,
+	'$' = true,
+	'*' = true,
+	'+' = true,
+	'?' = true,
+	'.' = true,
+	'(' = true,
+	'[' = true,
+	'%' = true,
+	'-' = true,
+}
+
+// helper call to quick search for special characters
+index_special :: proc(text: string) -> int {
+	for i in 0..<len(text) {
+		if SPECIALS_TABLE[text[i]] {
+			return i
+		}
+	}
+
+	return -1
+}
+
+lmem_find :: proc(s1, s2: string) -> int {
+	l1 := len(s1)
+	l2 := len(s2)
+
+	if l2 == 0 {
+		return 0
+	} else if l2 > l1 {
+		return -1
+	} else {
+		init := strings.index_byte(s1, s2[0])
+		end := init + l2
+
+		for end <= l1 && init != -1 {
+			init += 1
+
+			if s1[init - 1:end] == s2 {
+				return init - 1
+			} else {
+				next := strings.index_byte(s1[init:], s2[0])
+
+				if next == -1 {
+					return -1
+				} else {
+					init = init + next
+					end = init + l2
+				}
+			}
+		}
+	}
+
+	return -1
+}
+
+// find a pattern with in a haystack with an offset
+// allow_memfind will speed up simple searches
+find_aux :: proc(
+	haystack: string, 
+	pattern: string, 
+	offset: int,
+	allow_memfind: bool,
+	matches: ^[MAX_CAPTURES]Match,
+) -> (captures: int, err: Error) {
+	s := offset
+	p := 0
+
+	specials_idx := index_special(pattern)
+	if allow_memfind && specials_idx == -1 {
+		if index := lmem_find(haystack[s:], pattern); index != -1 {
+			matches[0] = { index + s, index + s + len(pattern) }
+			captures = 1
+			return
+		} else {
+			return
+		}
+	}
+
+	pattern := pattern
+	anchor: bool
+	if len(pattern) > 0 && pattern[0] == '^' {
+		anchor = true
+		pattern = pattern[1:]
+	}
+
+	ms := Match_State {
+		src = haystack,
+		pattern = pattern,
+	}
+
+	for {
+		res := match(&ms, s, p) or_return
+
+		if res != INVALID {
+			// disallow non advancing match
+			if s == res {
+				err = .Match_Invalid
+			} 
+			
+			// NOTE(Skytrias): first result is reserved for a full match
+			matches[0] = { s, res }
+
+			// rest are the actual captures
+			captures = push_captures(&ms, -1, -1, matches[1:]) or_return
+			captures += 1
+
+			return
+		}
+
+		s += 1
+
+		if !(s < len(ms.src) && !anchor) {
+			break
+		}
+	}
+
+	return
+}
+
+// iterative matching which returns the 0th/1st match
+// rest has to be used from captures
+gmatch :: proc(
+	haystack: ^string,
+	pattern: string,
+	captures: ^[MAX_CAPTURES]Match,
+) -> (res: string, ok: bool) {
+	if len(haystack) > 0 {
+		length, err := find_aux(haystack^, pattern, 0, false, captures)
+
+		if length != 0 && err == .OK {
+			ok = true
+			first := length > 1 ? 1 : 0
+			cap := captures[first]
+			res = haystack[cap.byte_start:cap.byte_end]
+			haystack^ = haystack[cap.byte_end:]
+		}
+	} 
+
+	return
+}
+
+// gsub with builder, replace patterns found with the replace content
+gsub_builder :: proc(
+	builder: ^strings.Builder,
+	haystack: string,
+	pattern: string,
+	replace: string,
+) -> string {
+	// find matches
+	captures: [MAX_CAPTURES]Match
+	haystack := haystack
+
+	for {
+		length, err := find_aux(haystack, pattern, 0, false, &captures)
+
+		// done
+		if length == 0 {
+			break
+		}
+
+		if err != .OK {
+			return {}
+		}
+
+		cap := captures[0]
+
+		// write front till capture
+		strings.write_string(builder, haystack[:cap.byte_start])
+
+		// write replacements
+		strings.write_string(builder, replace)
+
+		// advance string till end
+		haystack = haystack[cap.byte_end:]
+	}
+
+	strings.write_string(builder, haystack[:])
+	return strings.to_string(builder^)
+}
+
+// uses temp builder to build initial string - then allocates the result
+gsub_allocator :: proc(
+	haystack: string,
+	pattern: string,
+	replace: string,
+	allocator := context.allocator,
+) -> string {
+	builder := strings.builder_make(0, 256, context.temp_allocator)
+	return gsub_builder(&builder, haystack, pattern, replace)
+}
+
+Gsub_Proc :: proc(
+	// optional passed data
+	data: rawptr, 
+	// word match found
+	word: string, 
+	// current haystack for found captures
+	haystack: string, 
+	// found captures - empty for no captures
+	captures: []Match,
+)
+
+// call a procedure on every match in the haystack
+gsub_with :: proc(
+	haystack: string,
+	pattern: string,
+	data: rawptr,
+	call: Gsub_Proc,
+) {
+	// find matches
+	captures: [MAX_CAPTURES]Match
+	haystack := haystack
+
+	for {
+		length, err := find_aux(haystack, pattern, 0, false, &captures)
+
+		// done
+		if length == 0 || err != .OK {
+			break
+		}
+
+		cap := captures[0]
+
+		word := haystack[cap.byte_start:cap.byte_end]
+		call(data, word, haystack, captures[1:length])
+
+		// advance string till end
+		haystack = haystack[cap.byte_end:]
+	}
+}
+
+gsub :: proc { gsub_builder, gsub_allocator }
+
+// iterative find with zeroth capture only
+gfind :: proc(
+	haystack: ^string,
+	pattern: string,
+	captures: ^[MAX_CAPTURES]Match,
+) -> (res: string, ok: bool) {
+	if len(haystack) > 0 {
+		length, err := find_aux(haystack^, pattern, 0, true, captures)
+
+		if length != 0 && err == .OK {
+			ok = true
+			cap := captures[0]
+			res = haystack[cap.byte_start:cap.byte_end]
+			haystack^ = haystack[cap.byte_end:]
+		}
+	} 
+
+	return
+}
+
+// rebuilds a pattern into a case insensitive pattern
+pattern_case_insensitive_builder :: proc(
+	builder: ^strings.Builder, 
+	pattern: string,
+) -> (res: string) {
+	p := pattern
+	last_percent: bool
+
+	for len(p) > 0 {
+		char, size := utf8.decode_rune_in_string(p)
+
+		if unicode.is_alpha(char) && !last_percent {
+			// write character class in manually
+			strings.write_byte(builder, '[')
+			strings.write_rune(builder, unicode.to_lower(char))
+			strings.write_rune(builder, unicode.to_upper(char))
+			strings.write_byte(builder, ']')
+		} else {
+			strings.write_rune(builder, char)
+		}
+
+		last_percent = char == L_ESC 
+		p = p[size:]
+	}
+
+	return strings.to_string(builder^)
+}
+
+pattern_case_insensitive_allocator :: proc(
+	pattern: string, 
+	cap: int = 256,
+	allocator := context.allocator,
+) -> (res: string) {
+	builder := strings.builder_make(0, cap, context.temp_allocator)
+	return pattern_case_insensitive_builder(&builder, pattern)	
+}
+
+pattern_case_insensitive :: proc { pattern_case_insensitive_builder, pattern_case_insensitive_allocator }
+
+// Matcher helper struct that stores optional data you might want to use or not
+// as lua is far more dynamic this helps dealing with too much data
+// this also allows use of find/match/gmatch at through one struct
+Matcher :: struct {
+	haystack: string,
+	pattern: string,
+	captures: [MAX_CAPTURES]Match,
+	captures_length: int,
+	offset: int,
+	err: Error,
+
+	// changing content for iterators
+	iter: string,
+	iter_index: int,
+}
+
+// init using haystack & pattern and an optional byte offset
+matcher_init :: proc(haystack, pattern: string, offset: int = 0) -> (res: Matcher) {
+	res.haystack = haystack
+	res.pattern = pattern
+	res.offset = offset
+	res.iter = haystack
+	return
+}
+
+// find the first match and return the byte start / end position in the string, true on success
+matcher_find :: proc(matcher: ^Matcher) -> (start, end: int, ok: bool) #no_bounds_check {
+	matcher.captures_length, matcher.err = find_aux(
+		matcher.haystack, 
+		matcher.pattern, 
+		matcher.offset, 
+		true, 
+		&matcher.captures,
+	)
+	ok = matcher.captures_length > 0 && matcher.err == .OK
+	match := matcher.captures[0]
+	start = match.byte_start
+	end = match.byte_end
+	return
+}
+
+// find the first match and return the matched word, true on success
+matcher_match :: proc(matcher: ^Matcher) -> (word: string, ok: bool) #no_bounds_check {
+	matcher.captures_length, matcher.err = find_aux(
+		matcher.haystack, 
+		matcher.pattern, 
+		matcher.offset, 
+		false, 
+		&matcher.captures,
+	)
+	ok = matcher.captures_length > 0 && matcher.err == .OK
+	match := matcher.captures[0]
+	word = matcher.haystack[match.byte_start:match.byte_end]
+	return
+}
+
+// get the capture at the "correct" spot, as spot 0 is reserved for the first match
+matcher_capture :: proc(matcher: ^Matcher, index: int, loc := #caller_location) -> string #no_bounds_check {
+	runtime.bounds_check_error_loc(loc, index + 1, MAX_CAPTURES - 1)
+	cap := matcher.captures[index + 1]
+	return matcher.haystack[cap.byte_start:cap.byte_end]
+}
+
+// get the raw match out of the captures, skipping spot 0
+matcher_capture_raw :: proc(matcher: ^Matcher, index: int, loc := #caller_location) -> Match #no_bounds_check {
+	runtime.bounds_check_error_loc(loc, index + 1, MAX_CAPTURES - 1)
+	return matcher.captures[index + 1]
+}
+
+// alias
+matcher_gmatch :: matcher_match_iter
+
+// iteratively match the haystack till it cant find any matches
+matcher_match_iter :: proc(matcher: ^Matcher) -> (res: string, index: int, ok: bool) {
+	if len(matcher.iter) > 0 {
+		matcher.captures_length, matcher.err = find_aux(
+			matcher.iter, 
+			matcher.pattern, 
+			matcher.offset, 
+			false, 
+			&matcher.captures,
+		)
+
+		if matcher.captures_length != 0 && matcher.err == .OK {
+			ok = true
+			first := matcher.captures_length > 1 ? 1 : 0
+			match := matcher.captures[first]
+			
+			// output
+			res = matcher.iter[match.byte_start:match.byte_end]
+			index = matcher.iter_index
+			
+			// advance
+			matcher.iter_index += 1
+			matcher.iter = matcher.iter[match.byte_end:]
+		}
+	}
+
+	return
+}
+
+// get a slice of all valid captures above the first match
+matcher_captures_slice :: proc(matcher: ^Matcher) -> []Match {
+	return matcher.captures[1:matcher.captures_length]
+}