[Clang] [C23] Implement N2653: u8 strings are char8_t[] (#97208)

Summary:
https://www.open-std.org/jtc1/sc22/wg14/www/docs/n2653.htm

Closes #97202

---------

Co-authored-by: cor3ntin <[email protected]>

Test Plan: 

Reviewers: 

Subscribers: 

Tasks: 

Tags: 


Differential Revision: https://phabricator.intern.facebook.com/D60250894

Author: MitalAshok

clang/docs/ReleaseNotes.rst

@@ -362,6 +362,12 @@ C23 Feature Support
 - Added the ``FLT_NORM_MAX``, ``DBL_NORM_MAX``, and ``LDBL_NORM_MAX`` to the
   freestanding implementation of ``<float.h>`` that ships with Clang.
 
+- Compiler support for `N2653 char8_t: A type for UTF-8 characters and strings`
+  <https://www.open-std.org/jtc1/sc22/wg14/www/docs/n2653.htm>`_: ``u8`` string
+  literals are now of type ``char8_t[N]`` in C23 and expose
+  ``__CLANG_ATOMIC_CHAR8_T_LOCK_FREE``/``__GCC_ATOMIC_CHAR8_T_LOCK_FREE`` to
+  implement the corresponding macro in ``<stdatomic.h>``.
+
 Non-comprehensive list of changes in this release
 -------------------------------------------------
 

clang/include/clang/Basic/DiagnosticSemaKinds.td

@@ -7249,7 +7249,10 @@ def err_array_init_utf8_string_into_char : Error<
 def warn_cxx20_compat_utf8_string : Warning<
   "type of UTF-8 string literal will change from array of const char to "
   "array of const char8_t in C++20">, InGroup<CXX20Compat>, DefaultIgnore;
-def note_cxx20_compat_utf8_string_remove_u8 : Note<
+def warn_c23_compat_utf8_string : Warning<
+  "type of UTF-8 string literal will change from array of char to "
+  "array of char8_t in C23">, InGroup<C23Compat>, DefaultIgnore;
+def note_cxx20_c23_compat_utf8_string_remove_u8 : Note<
   "remove 'u8' prefix to avoid a change of behavior; "
   "Clang encodes unprefixed narrow string literals as UTF-8">;
 def err_array_init_different_type : Error<

clang/lib/Frontend/InitPreprocessor.cpp

@@ -1170,6 +1170,8 @@ static void InitializePredefinedMacros(const TargetInfo &TI,
   DefineType("__WCHAR_TYPE__", TI.getWCharType(), Builder);
   DefineType("__WINT_TYPE__", TI.getWIntType(), Builder);
   DefineTypeSizeAndWidth("__SIG_ATOMIC", TI.getSigAtomicType(), TI, Builder);
+  if (LangOpts.C23)
+    DefineType("__CHAR8_TYPE__", TI.UnsignedChar, Builder);
   DefineType("__CHAR16_TYPE__", TI.getChar16Type(), Builder);
   DefineType("__CHAR32_TYPE__", TI.getChar32Type(), Builder);
 
@@ -1349,8 +1351,10 @@ static void InitializePredefinedMacros(const TargetInfo &TI,
                       getLockFreeValue(TI.get##Type##Width(), TI));
     DEFINE_LOCK_FREE_MACRO(BOOL, Bool);
     DEFINE_LOCK_FREE_MACRO(CHAR, Char);
-    if (LangOpts.Char8)
-      DEFINE_LOCK_FREE_MACRO(CHAR8_T, Char); // Treat char8_t like char.
+    // char8_t has the same representation / width as unsigned
+    // char in C++ and is a typedef for unsigned char in C23
+    if (LangOpts.Char8 || LangOpts.C23)
+      DEFINE_LOCK_FREE_MACRO(CHAR8_T, Char);
     DEFINE_LOCK_FREE_MACRO(CHAR16_T, Char16);
     DEFINE_LOCK_FREE_MACRO(CHAR32_T, Char32);
     DEFINE_LOCK_FREE_MACRO(WCHAR_T, WChar);

clang/lib/Headers/stdatomic.h

@@ -35,6 +35,9 @@ extern "C" {
 
 #define ATOMIC_BOOL_LOCK_FREE       __CLANG_ATOMIC_BOOL_LOCK_FREE
 #define ATOMIC_CHAR_LOCK_FREE       __CLANG_ATOMIC_CHAR_LOCK_FREE
+#if defined(__STDC_VERSION__) && __STDC_VERSION__ >= 202311L
+#define ATOMIC_CHAR8_T_LOCK_FREE    __CLANG_ATOMIC_CHAR8_T_LOCK_FREE
+#endif
 #define ATOMIC_CHAR16_T_LOCK_FREE   __CLANG_ATOMIC_CHAR16_T_LOCK_FREE
 #define ATOMIC_CHAR32_T_LOCK_FREE   __CLANG_ATOMIC_CHAR32_T_LOCK_FREE
 #define ATOMIC_WCHAR_T_LOCK_FREE    __CLANG_ATOMIC_WCHAR_T_LOCK_FREE
@@ -104,6 +107,9 @@ typedef _Atomic(long)               atomic_long;
 typedef _Atomic(unsigned long)      atomic_ulong;
 typedef _Atomic(long long)          atomic_llong;
 typedef _Atomic(unsigned long long) atomic_ullong;
+#if defined(__STDC_VERSION__) && __STDC_VERSION__ >= 202311L
+typedef _Atomic(unsigned char)      atomic_char8_t;
+#endif
 typedef _Atomic(uint_least16_t)     atomic_char16_t;
 typedef _Atomic(uint_least32_t)     atomic_char32_t;
 typedef _Atomic(wchar_t)            atomic_wchar_t;

clang/lib/Sema/SemaExpr.cpp

@@ -2051,6 +2051,8 @@ Sema::ActOnStringLiteral(ArrayRef<Token> StringToks, Scope *UDLScope) {
   } else if (Literal.isUTF8()) {
     if (getLangOpts().Char8)
       CharTy = Context.Char8Ty;
+    else if (getLangOpts().C23)
+      CharTy = Context.UnsignedCharTy;
     Kind = StringLiteralKind::UTF8;
   } else if (Literal.isUTF16()) {
     CharTy = Context.Char16Ty;
@@ -2062,17 +2064,23 @@ Sema::ActOnStringLiteral(ArrayRef<Token> StringToks, Scope *UDLScope) {
     CharTy = Context.UnsignedCharTy;
   }
 
-  // Warn on initializing an array of char from a u8 string literal; this
-  // becomes ill-formed in C++2a.
-  if (getLangOpts().CPlusPlus && !getLangOpts().CPlusPlus20 &&
-      !getLangOpts().Char8 && Kind == StringLiteralKind::UTF8) {
-    Diag(StringTokLocs.front(), diag::warn_cxx20_compat_utf8_string);
+  // Warn on u8 string literals before C++20 and C23, whose type
+  // was an array of char before but becomes an array of char8_t.
+  // In C++20, it cannot be used where a pointer to char is expected.
+  // In C23, it might have an unexpected value if char was signed.
+  if (Kind == StringLiteralKind::UTF8 &&
+      (getLangOpts().CPlusPlus
+           ? !getLangOpts().CPlusPlus20 && !getLangOpts().Char8
+           : !getLangOpts().C23)) {
+    Diag(StringTokLocs.front(), getLangOpts().CPlusPlus
+                                    ? diag::warn_cxx20_compat_utf8_string
+                                    : diag::warn_c23_compat_utf8_string);
 
     // Create removals for all 'u8' prefixes in the string literal(s). This
-    // ensures C++2a compatibility (but may change the program behavior when
+    // ensures C++20/C23 compatibility (but may change the program behavior when
     // built by non-Clang compilers for which the execution character set is
     // not always UTF-8).
-    auto RemovalDiag = PDiag(diag::note_cxx20_compat_utf8_string_remove_u8);
+    auto RemovalDiag = PDiag(diag::note_cxx20_c23_compat_utf8_string_remove_u8);
     SourceLocation RemovalDiagLoc;
     for (const Token &Tok : StringToks) {
       if (Tok.getKind() == tok::utf8_string_literal) {

clang/test/C/C23/n2653.c

@@ -0,0 +1,34 @@
+// RUN: %clang_cc1 -ffreestanding -verify=c23 -std=c23 %s
+// RUN: %clang_cc1 -ffreestanding -verify=c17 -std=c17 %s
+
+// c23-no-diagnostics
+
+#include <stdatomic.h>
+
+#define __enable_constant_folding(x) (__builtin_constant_p(x) ? (x) : (x))
+#define __is_same(a, b) (__extension__ _Generic(a, b: 1, default: 0) && __extension__ _Generic(b, a: 1, default: 0))
+
+#ifndef ATOMIC_CHAR8_T_LOCK_FREE
+#error missing
+#endif
+// c17-error@-2 {{missing}}
+
+_Static_assert(__is_same(atomic_char8_t, unsigned char _Atomic), "");
+// c17-error@-1 {{use of undeclared identifier 'atomic_char8_t'}}
+// c17-error@-2 {{unknown type name 'atomic_char8_t'}}
+
+_Static_assert(_Generic(u8"", unsigned char*: 1, char*: 0), "");
+// c17-error@-1 {{static assertion failed}}
+
+// -fsigned-char is the default
+#define M(X) __enable_constant_folding((X) >= 0x80)
+
+_Static_assert(M(u8"\U000000E9"[0]), "");
+// c17-error@-1 {{static assertion failed}}
+#if __STDC_VERSION__ >= 202311L
+_Static_assert(M(u8'\xC3'), "");
+#endif
+
+const          char cu8[]  = u8"text";
+const signed   char scu8[] = u8"text";
+const unsigned char ucu8[] = u8"text";

clang/www/c_status.html

@@ -1066,7 +1066,7 @@ <h2 id="c2x">C23 implementation status</h2>
     <tr>
       <td>char8_t: A type for UTF-8 characters and strings</td>
       <td><a href="https://www.open-std.org/jtc1/sc22/wg14/www/docs/n2653.htm">N2653</a></td>
-      <td class="none" align="center">No</td>
+      <td class="unreleased" align="center">Clang 19</td>
     </tr>
     <tr>
       <td>Clarification for max exponent macros-update</td>