Switch order of __clz and __ctz

Author: winner245

libcxx/include/__algorithm/fill_n.h

@@ -41,7 +41,7 @@ __fill_n_bool(__bit_iterator<_Cp, false> __first, typename __size_difference_typ
   if (__first.__ctz_ != 0) {
     __storage_type __clz_f = static_cast<__storage_type>(__bits_per_word - __first.__ctz_);
     __storage_type __dn    = std::min(__clz_f, __n);
-    std::__fill_masked_range(std::__to_address(__first.__seg_), __first.__ctz_, __clz_f - __dn, _FillVal);
+    std::__fill_masked_range(std::__to_address(__first.__seg_), __clz_f - __dn, __first.__ctz_, _FillVal);
     __n -= __dn;
     ++__first.__seg_;
   }
@@ -52,7 +52,7 @@ __fill_n_bool(__bit_iterator<_Cp, false> __first, typename __size_difference_typ
   // do last partial word
   if (__n > 0) {
     __first.__seg_ += __nw;
-    std::__fill_masked_range(std::__to_address(__first.__seg_), 0u, __bits_per_word - __n, _FillVal);
+    std::__fill_masked_range(std::__to_address(__first.__seg_), __bits_per_word - __n, 0u, _FillVal);
   }
 }
 

libcxx/include/__bit_reference

@@ -62,16 +62,16 @@ struct __size_difference_type_traits<_Cp, __void_t<typename _Cp::difference_type
 // This function is designed to operate correctly even for smaller integral types like `uint8_t`, `uint16_t`,
 // or `unsigned short`. Casting back to _StorageType is crucial to prevent undefined behavior that can arise
 // from integral promotions.
-// See https://github.com/llvm/llvm-project/pull/122410
+// See https://github.com/llvm/llvm-project/pull/122410.
 template <class _StoragePointer,
           __enable_if_t<is_unsigned<typename pointer_traits<_StoragePointer>::element_type>::value, int> >
 _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX14 void
-__fill_masked_range(_StoragePointer __word, unsigned __ctz, unsigned __clz, bool __fill_val) {
+__fill_masked_range(_StoragePointer __word, unsigned __clz, unsigned __ctz, bool __fill_val) {
   using _StorageType = typename pointer_traits<_StoragePointer>::element_type;
   _LIBCPP_ASSERT_VALID_INPUT_RANGE(
       __ctz + __clz < sizeof(_StorageType) * CHAR_BIT, "__fill_masked_range called with invalid range");
-  _StorageType __m = static_cast<_StorageType>(static_cast<_StorageType>(~static_cast<_StorageType>(0)) << __ctz) &
-                     static_cast<_StorageType>(static_cast<_StorageType>(~static_cast<_StorageType>(0)) >> __clz);
+  _StorageType __m = static_cast<_StorageType>(static_cast<_StorageType>(~static_cast<_StorageType>(0)) >> __clz) &
+                     static_cast<_StorageType>(static_cast<_StorageType>(~static_cast<_StorageType>(0)) << __ctz);
   if (__fill_val)
     *__word |= __m;
   else

libcxx/test/std/algorithms/alg.modifying.operations/alg.fill/ranges.fill.pass.cpp

@@ -78,8 +78,12 @@ constexpr void test_iterators() {
   }
 }
 
-// The `ranges::{fill, fill_n}` algorithms require `vector<bool, Alloc>::iterator` to satisfy the
-// `std::indirectly_writable` concept when used with `vector<bool, Alloc>`, which is only met since C++23.
+// Make sure we behave properly with std::vector<bool> iterators with custom size types, see
+// https://github.com/llvm/llvm-project/pull/122410.
+//
+// The `ranges::{fill, fill_n}` algorithms require `vector<bool, Alloc>::iterator` to satisfy
+// the `std::indirectly_writable` concept when used with `vector<bool, Alloc>`, which is only
+// satisfied since C++23.
 #if TEST_STD_VER >= 23
 TEST_CONSTEXPR_CXX20 void test_bititer_with_custom_sized_types() {
   {

libcxx/test/std/algorithms/alg.modifying.operations/alg.fill/ranges.fill_n.pass.cpp

@@ -51,8 +51,12 @@ constexpr void test_iterators() {
   }
 }
 
-// The `ranges::{fill, fill_n}` algorithms require `vector<bool, Alloc>::iterator` to satisfy the
-// `std::indirectly_writable` concept when used with `vector<bool, Alloc>`, which is only met since C++23.
+// Make sure we behave properly with std::vector<bool> iterators with custom size types, see
+// https://github.com/llvm/llvm-project/pull/122410.
+//
+// The `ranges::{fill, fill_n}` algorithms require `vector<bool, Alloc>::iterator` to satisfy
+// the `std::indirectly_writable` concept when used with `vector<bool, Alloc>`, which is only
+// satisfied since C++23.
 #if TEST_STD_VER >= 23
 TEST_CONSTEXPR_CXX20 void test_bititer_with_custom_sized_types() {
   {

libcxx/test/support/sized_allocator.h

@@ -16,6 +16,8 @@
 
 #include "test_macros.h"
 
+// Allocator with a provided size_type and difference_type, used to test corner cases
+// like arithmetic on Allocator::size_type in generic code.
 template <typename T, typename Size = std::size_t, typename Difference = std::ptrdiff_t>
 class sized_allocator {
   template <typename U, typename Sz, typename Diff>