diff --git a/ext/json/ext/generator/generator.c b/ext/json/ext/generator/generator.c
index 6ece9f05..35d543a7 100644
--- a/ext/json/ext/generator/generator.c
+++ b/ext/json/ext/generator/generator.c
@@ -128,7 +128,7 @@ typedef struct _search_state {
 #endif /* HAVE_SIMD */
 } search_state;
 
-static ALWAYS_INLINE() void search_flush(search_state *search)
+ALWAYS_INLINE(static) void search_flush(search_state *search)
 {
     // Do not remove this conditional without profiling, specifically escape-heavy text.
     // escape_UTF8_char_basic will advance search->ptr and search->cursor (effectively a search_flush).
@@ -171,7 +171,7 @@ static inline unsigned char search_escape_basic(search_state *search)
     return 0;
 }
 
-static ALWAYS_INLINE() void escape_UTF8_char_basic(search_state *search)
+ALWAYS_INLINE(static) void escape_UTF8_char_basic(search_state *search)
 {
     const unsigned char ch = (unsigned char)*search->ptr;
     switch (ch) {
@@ -258,7 +258,7 @@ static inline void escape_UTF8_char(search_state *search, unsigned char ch_len)
 
 #ifdef HAVE_SIMD
 
-static ALWAYS_INLINE() char *copy_remaining_bytes(search_state *search, unsigned long vec_len, unsigned long len)
+ALWAYS_INLINE(static) char *copy_remaining_bytes(search_state *search, unsigned long vec_len, unsigned long len)
 {
     // Flush the buffer so everything up until the last 'len' characters are unflushed.
     search_flush(search);
@@ -281,7 +281,7 @@ static ALWAYS_INLINE() char *copy_remaining_bytes(search_state *search, unsigned
 
 #ifdef HAVE_SIMD_NEON
 
-static ALWAYS_INLINE() unsigned char neon_next_match(search_state *search)
+ALWAYS_INLINE(static) unsigned char neon_next_match(search_state *search)
 {
     uint64_t mask = search->matches_mask;
     uint32_t index = trailing_zeros64(mask) >> 2;
@@ -395,7 +395,7 @@ static inline unsigned char search_escape_basic_neon(search_state *search)
 
 #ifdef HAVE_SIMD_SSE2
 
-static ALWAYS_INLINE() unsigned char sse2_next_match(search_state *search)
+ALWAYS_INLINE(static) unsigned char sse2_next_match(search_state *search)
 {
     int mask = search->matches_mask;
     int index = trailing_zeros(mask);
@@ -419,7 +419,7 @@ static ALWAYS_INLINE() unsigned char sse2_next_match(search_state *search)
 #define TARGET_SSE2
 #endif
 
-static TARGET_SSE2 ALWAYS_INLINE() unsigned char search_escape_basic_sse2(search_state *search)
+ALWAYS_INLINE(static) TARGET_SSE2 unsigned char search_escape_basic_sse2(search_state *search)
 {
     if (RB_UNLIKELY(search->has_matches)) {
         // There are more matches if search->matches_mask > 0.
@@ -1123,8 +1123,7 @@ struct hash_foreach_arg {
     bool mixed_keys_encountered;
 };
 
-NOINLINE()
-static void
+NOINLINE(static) void
 json_inspect_hash_with_mixed_keys(struct hash_foreach_arg *arg)
 {
     if (arg->mixed_keys_encountered) {
diff --git a/ext/json/ext/parser/parser.c b/ext/json/ext/parser/parser.c
index 0216eef9..5b7cd835 100644
--- a/ext/json/ext/parser/parser.c
+++ b/ext/json/ext/parser/parser.c
@@ -88,7 +88,7 @@ static void rvalue_cache_insert_at(rvalue_cache *cache, int index, VALUE rstring
 #if JSON_CPU_LITTLE_ENDIAN_64BITS
 #if __has_builtin(__builtin_bswap64)
 #undef rstring_cache_memcmp
-static ALWAYS_INLINE() int rstring_cache_memcmp(const char *str, const char *rptr, const long length)
+ALWAYS_INLINE(static) int rstring_cache_memcmp(const char *str, const char *rptr, const long length)
 {
     // The libc memcmp has numerous complex optimizations, but in this particular case,
     // we know the string is small (JSON_RVALUE_CACHE_MAX_ENTRY_LENGTH), so being able to
@@ -117,7 +117,7 @@ static ALWAYS_INLINE() int rstring_cache_memcmp(const char *str, const char *rpt
 #endif
 #endif
 
-static ALWAYS_INLINE() int rstring_cache_cmp(const char *str, const long length, VALUE rstring)
+ALWAYS_INLINE(static) int rstring_cache_cmp(const char *str, const long length, VALUE rstring)
 {
     const char *rstring_ptr;
     long rstring_length;
@@ -131,7 +131,7 @@ static ALWAYS_INLINE() int rstring_cache_cmp(const char *str, const long length,
     }
 }
 
-static ALWAYS_INLINE() VALUE rstring_cache_fetch(rvalue_cache *cache, const char *str, const long length)
+ALWAYS_INLINE(static) VALUE rstring_cache_fetch(rvalue_cache *cache, const char *str, const long length)
 {
     int low = 0;
     int high = cache->length - 1;
@@ -540,7 +540,7 @@ json_eat_comments(JSON_ParserState *state)
     }
 }
 
-static ALWAYS_INLINE() void
+ALWAYS_INLINE(static) void
 json_eat_whitespace(JSON_ParserState *state)
 {
     while (true) {
@@ -661,7 +661,7 @@ static inline const char *json_next_backslash(const char *pe, const char *string
     return NULL;
 }
 
-static NOINLINE() VALUE json_string_unescape(JSON_ParserState *state, JSON_ParserConfig *config, const char *string, const char *stringEnd, bool is_name, JSON_UnescapePositions *positions)
+NOINLINE(static) VALUE json_string_unescape(JSON_ParserState *state, JSON_ParserConfig *config, const char *string, const char *stringEnd, bool is_name, JSON_UnescapePositions *positions)
 {
     bool intern = is_name || config->freeze;
     bool symbolize = is_name && config->symbolize_names;
@@ -946,7 +946,7 @@ static const bool string_scan_table[256] = {
 static SIMD_Implementation simd_impl = SIMD_NONE;
 #endif /* HAVE_SIMD */
 
-static ALWAYS_INLINE() bool string_scan(JSON_ParserState *state)
+ALWAYS_INLINE(static) bool string_scan(JSON_ParserState *state)
 {
 #ifdef HAVE_SIMD
 #if defined(HAVE_SIMD_NEON)
@@ -1015,7 +1015,7 @@ static VALUE json_parse_escaped_string(JSON_ParserState *state, JSON_ParserConfi
     return Qfalse;
 }
 
-static ALWAYS_INLINE() VALUE json_parse_string(JSON_ParserState *state, JSON_ParserConfig *config, bool is_name)
+ALWAYS_INLINE(static) VALUE json_parse_string(JSON_ParserState *state, JSON_ParserConfig *config, bool is_name)
 {
     state->cursor++;
     const char *start = state->cursor;
diff --git a/ext/json/ext/simd/simd.h b/ext/json/ext/simd/simd.h
index c9e3b3ec..f8e5ee18 100644
--- a/ext/json/ext/simd/simd.h
+++ b/ext/json/ext/simd/simd.h
@@ -73,14 +73,14 @@ static inline SIMD_Implementation find_simd_implementation(void)
 #define HAVE_SIMD_NEON 1
 
 // See: https://community.arm.com/arm-community-blogs/b/servers-and-cloud-computing-blog/posts/porting-x86-vector-bitmask-optimizations-to-arm-neon
-static ALWAYS_INLINE() uint64_t neon_match_mask(uint8x16_t matches)
+ALWAYS_INLINE(static) uint64_t neon_match_mask(uint8x16_t matches)
 {
     const uint8x8_t res = vshrn_n_u16(vreinterpretq_u16_u8(matches), 4);
     const uint64_t mask = vget_lane_u64(vreinterpret_u64_u8(res), 0);
     return mask & 0x8888888888888888ull;
 }
 
-static ALWAYS_INLINE() uint64_t compute_chunk_mask_neon(const char *ptr)
+ALWAYS_INLINE(static) uint64_t compute_chunk_mask_neon(const char *ptr)
 {
     uint8x16_t chunk = vld1q_u8((const unsigned char *)ptr);
 
@@ -93,7 +93,7 @@ static ALWAYS_INLINE() uint64_t compute_chunk_mask_neon(const char *ptr)
     return neon_match_mask(needs_escape);
 }
 
-static ALWAYS_INLINE() int string_scan_simd_neon(const char **ptr, const char *end, uint64_t *mask)
+ALWAYS_INLINE(static) int string_scan_simd_neon(const char **ptr, const char *end, uint64_t *mask)
 {
     while (*ptr + sizeof(uint8x16_t) <= end) {
         uint64_t chunk_mask = compute_chunk_mask_neon(*ptr);
@@ -140,7 +140,7 @@ static inline uint8x16x4_t load_uint8x16_4(const unsigned char *table)
 #define _mm_cmpgt_epu8(a, b) _mm_xor_si128(_mm_cmple_epu8(a, b), _mm_set1_epi8(-1))
 #define _mm_cmplt_epu8(a, b) _mm_cmpgt_epu8(b, a)
 
-static TARGET_SSE2 ALWAYS_INLINE() int compute_chunk_mask_sse2(const char *ptr)
+ALWAYS_INLINE(static) TARGET_SSE2 int compute_chunk_mask_sse2(const char *ptr)
 {
     __m128i chunk         = _mm_loadu_si128((__m128i const*)ptr);
     // Trick: c < 32 || c == 34 can be factored as c ^ 2 < 33
@@ -151,7 +151,7 @@ static TARGET_SSE2 ALWAYS_INLINE() int compute_chunk_mask_sse2(const char *ptr)
     return _mm_movemask_epi8(needs_escape);
 }
 
-static TARGET_SSE2 ALWAYS_INLINE() int string_scan_simd_sse2(const char **ptr, const char *end, int *mask)
+ALWAYS_INLINE(static) TARGET_SSE2 int string_scan_simd_sse2(const char **ptr, const char *end, int *mask)
 {
     while (*ptr + sizeof(__m128i) <= end) {
         int chunk_mask = compute_chunk_mask_sse2(*ptr);