Replace call_once usage in libc++ with atomics

This avoid a dependency on pthread API (which are used to implement
call_once under the hood) which makes libc++ much more usable from
Wasm workers (there pthreads are not available).

The downside here is that when thread race to be first one to use a
given facet we would "leak" an ID which could cause the `facets_` vector
to become more spare that it otherwise would be.

IIUC this `facets_` vector can already be sparse so it not clear to
me what impact this would have in practice.

See: #26375

Author: sbc100

system/lib/libcxx/include/__locale

@@ -17,7 +17,6 @@
 #  include <__locale_dir/locale_base_api.h>
 #  include <__memory/addressof.h>
 #  include <__memory/shared_count.h>
-#  include <__mutex/once_flag.h>
 #  include <__type_traits/make_unsigned.h>
 #  include <__utility/no_destroy.h>
 #  include <__utility/private_constructor_tag.h>
@@ -146,7 +145,6 @@ private:
 };
 
 class _LIBCPP_EXPORTED_FROM_ABI locale::id {
-  once_flag __flag_;
   int32_t __id_;
 
   static int32_t __next_id;

system/lib/libcxx/src/locale.cpp

@@ -589,7 +589,14 @@ void locale::facet::__on_zero_shared() noexcept { delete this; }
 constinit int32_t locale::id::__next_id = 0;
 
 long locale::id::__get() {
-  call_once(__flag_, [&] { __id_ = __libcpp_atomic_add(&__next_id, 1); });
+  if (__libcpp_atomic_load(&__id_) == 0) {
+    int32_t proposed_id = __libcpp_atomic_add(&__next_id, 1);
+    int32_t expected = 0;
+    // If we race with another thread here the CAS will fail and
+    // the proposed_id will be leaked, but __id_ will be non-zero
+    // in either case.
+    __libcpp_atomic_compare_exchange(&__id_, &expected, proposed_id);
+  }
   return __id_ - 1;
 }
 

test/codesize/hello_libcxx.cpp

@@ -9,4 +9,3 @@ int main() {
   std::cout << "hello, world!" << std::endl;
   return 0;
 }
-

test/codesize/test_codesize_cxx_ctors1.json

@@ -1,10 +1,10 @@
 {
   "a.out.js": 19214,
   "a.out.js.gz": 7982,
-  "a.out.nodebug.wasm": 132865,
-  "a.out.nodebug.wasm.gz": 49889,
-  "total": 152079,
-  "total_gz": 57871,
+  "a.out.nodebug.wasm": 132617,
+  "a.out.nodebug.wasm.gz": 49801,
+  "total": 151831,
+  "total_gz": 57783,
   "sent": [
     "__cxa_throw",
     "_abort_js",

test/codesize/test_codesize_cxx_ctors2.json

@@ -1,10 +1,10 @@
 {
   "a.out.js": 19191,
   "a.out.js.gz": 7968,
-  "a.out.nodebug.wasm": 132285,
-  "a.out.nodebug.wasm.gz": 49542,
-  "total": 151476,
-  "total_gz": 57510,
+  "a.out.nodebug.wasm": 132036,
+  "a.out.nodebug.wasm.gz": 49456,
+  "total": 151227,
+  "total_gz": 57424,
   "sent": [
     "__cxa_throw",
     "_abort_js",

test/codesize/test_codesize_cxx_except.json

@@ -1,10 +1,10 @@
 {
   "a.out.js": 22875,
   "a.out.js.gz": 8959,
-  "a.out.nodebug.wasm": 172743,
-  "a.out.nodebug.wasm.gz": 57374,
-  "total": 195618,
-  "total_gz": 66333,
+  "a.out.nodebug.wasm": 173885,
+  "a.out.nodebug.wasm.gz": 57439,
+  "total": 196760,
+  "total_gz": 66398,
   "sent": [
     "__cxa_begin_catch",
     "__cxa_end_catch",

test/codesize/test_codesize_cxx_except_wasm.json

@@ -1,10 +1,10 @@
 {
   "a.out.js": 19025,
   "a.out.js.gz": 7906,
-  "a.out.nodebug.wasm": 148138,
-  "a.out.nodebug.wasm.gz": 55261,
-  "total": 167163,
-  "total_gz": 63167,
+  "a.out.nodebug.wasm": 149501,
+  "a.out.nodebug.wasm.gz": 55425,
+  "total": 168526,
+  "total_gz": 63331,
   "sent": [
     "_abort_js",
     "_tzset_js",

test/codesize/test_codesize_cxx_except_wasm_legacy.json

@@ -1,10 +1,10 @@
 {
   "a.out.js": 19099,
   "a.out.js.gz": 7926,
-  "a.out.nodebug.wasm": 145944,
-  "a.out.nodebug.wasm.gz": 54887,
-  "total": 165043,
-  "total_gz": 62813,
+  "a.out.nodebug.wasm": 147317,
+  "a.out.nodebug.wasm.gz": 55080,
+  "total": 166416,
+  "total_gz": 63006,
   "sent": [
     "_abort_js",
     "_tzset_js",

test/codesize/test_codesize_cxx_lto.json

@@ -1,10 +1,10 @@
 {
   "a.out.js": 18562,
   "a.out.js.gz": 7670,
-  "a.out.nodebug.wasm": 102126,
-  "a.out.nodebug.wasm.gz": 39426,
-  "total": 120688,
-  "total_gz": 47096,
+  "a.out.nodebug.wasm": 104147,
+  "a.out.nodebug.wasm.gz": 39808,
+  "total": 122709,
+  "total_gz": 47478,
   "sent": [
     "a (emscripten_resize_heap)",
     "b (_setitimer_js)",

test/codesize/test_codesize_cxx_mangle.json

@@ -1,10 +1,10 @@
 {
   "a.out.js": 22925,
   "a.out.js.gz": 8978,
-  "a.out.nodebug.wasm": 239177,
-  "a.out.nodebug.wasm.gz": 79774,
-  "total": 262102,
-  "total_gz": 88752,
+  "a.out.nodebug.wasm": 240319,
+  "a.out.nodebug.wasm.gz": 79859,
+  "total": 263244,
+  "total_gz": 88837,
   "sent": [
     "__cxa_begin_catch",
     "__cxa_end_catch",