Use -Wcharacter-conversion, add it to -Wconversion, make it a feature for libcxx

Author: cor3ntin

clang/docs/ReleaseNotes.rst

@@ -503,7 +503,7 @@ Improvements to Clang's diagnostics
 - ``-Wreserved-identifier`` now fires on reserved parameter names in a function
   declaration which is not a definition.
 
-- A new ``-Wimplicit-unicode-conversion`` warns where comparing or implicitly converting
+- A new ``-Wcharacter-conversion`` warns where comparing or implicitly converting
   between different Unicode character types (``char8_t``, ``char16_t``, ``char32_t``).
   This warning only triggers in C++ as these types are aliases in C. (#GH138526)
 

clang/include/clang/Basic/DiagnosticGroups.td

@@ -111,7 +111,7 @@ def EnumConversion : DiagGroup<"enum-conversion",
                                 ImplicitEnumEnumCast,
                                 EnumFloatConversion,
                                 EnumCompareConditional]>;
-def ImplicitUnicodeConversion : DiagGroup<"implicit-unicode-conversion">;
+def CharacterConversion : DiagGroup<"character-conversion">;
 def DeprecatedOFast : DiagGroup<"deprecated-ofast">;
 def ObjCSignedCharBoolImplicitIntConversion :
   DiagGroup<"objc-signed-char-bool-implicit-int-conversion">;
@@ -1074,6 +1074,7 @@ def Parentheses : DiagGroup<"parentheses",
 //   - __null-to-integer conversion warnings are on by default
 def Conversion : DiagGroup<"conversion",
                            [BoolConversion,
+                            CharacterConversion,
                             ConstantConversion,
                             EnumConversion,
                             BitFieldEnumConversion,

clang/include/clang/Basic/DiagnosticSemaKinds.td

@@ -4360,25 +4360,25 @@ def warn_address_of_reference_bool_conversion : Warning<
 def warn_impcast_unicode_char_type
     : Warning<"implicit conversion from %0 to %1 may change the meaning of the "
               "represented code unit">,
-      InGroup<ImplicitUnicodeConversion>;
+      InGroup<CharacterConversion>;
 def warn_impcast_unicode_precision
     : Warning<"implicit conversion from %0 to %1 may lose precision and change "
               "the meaning of the represented code unit">,
-      InGroup<ImplicitUnicodeConversion>;
+      InGroup<CharacterConversion>;
 def warn_impcast_unicode_char_type_constant
     : Warning<"implicit conversion from %0 to %1 changes the meaning of the "
               "%select{code unit|codepoint}2 '%3'">,
-      InGroup<ImplicitUnicodeConversion>;
+      InGroup<CharacterConversion>;
 
 def warn_comparison_unicode_mixed_types
     : Warning<"comparing values of different Unicode code unit types %0 and %1 "
               "may compare different codepoints">,
-      InGroup<ImplicitUnicodeConversion>;
+      InGroup<CharacterConversion>;
 
 def warn_comparison_unicode_mixed_types_constant
     : Warning<"comparing values of different Unicode code unit types %0 and %1 "
               "compares unrelated code units '%2' and '%3'">,
-      InGroup<ImplicitUnicodeConversion>;
+      InGroup<CharacterConversion>;
 
 def warn_xor_used_as_pow : Warning<
   "result of '%0' is %1; did you mean exponentiation?">,
@@ -6843,7 +6843,7 @@ def err_counted_by_on_incomplete_type_on_use : Error <
 
 def note_counted_by_consider_completing_pointee_ty : Note<
   "consider providing a complete definition for %0">;
-  
+
 def note_counted_by_consider_using_sized_by : Note<
   "consider using '__sized_by%select{|_or_null}0' instead of "
   "'__counted_by%select{|_or_null}0'">;
@@ -7745,7 +7745,7 @@ def warn_comparison_of_mixed_enum_types_switch : Warning<
 def warn_arith_conv_mixed__unicode_types
     : Warning<"%sub{select_arith_conv_kind}0 "
               "different Unicode character types %1 and %2">,
-      InGroup<ImplicitUnicodeConversion>;
+      InGroup<CharacterConversion>;
 
 def err_typecheck_assign_const : Error<
   "%select{"

libcxx/test/std/algorithms/alg.nonmodifying/alg.equal/equal.pass.cpp

@@ -20,7 +20,7 @@
 
 // We test the cartesian product, so we sometimes compare differently signed types
 // ADDITIONAL_COMPILE_FLAGS(gcc-style-warnings): -Wno-sign-compare
-// ADDITIONAL_COMPILE_FLAGS(clang-21): -Wno-implicit-unicode-conversion
+// ADDITIONAL_COMPILE_FLAGS(character-conversion-warnings): -Wno-character-conversion
 
 // MSVC warning C4242: 'argument': conversion from 'int' to 'const _Ty', possible loss of data
 // MSVC warning C4244: 'argument': conversion from 'wchar_t' to 'const _Ty', possible loss of data

libcxx/test/std/algorithms/alg.nonmodifying/alg.find/find.pass.cpp

@@ -8,7 +8,7 @@
 
 // ADDITIONAL_COMPILE_FLAGS(gcc): -Wno-bool-compare
 // ADDITIONAL_COMPILE_FLAGS(gcc-style-warnings): -Wno-sign-compare
-// ADDITIONAL_COMPILE_FLAGS(clang-21): -Wno-implicit-unicode-conversion
+// ADDITIONAL_COMPILE_FLAGS(character-conversion-warnings): -Wno-character-conversion
 // MSVC warning C4245: conversion from 'int' to 'wchar_t', signed/unsigned mismatch
 // MSVC warning C4305: truncation from 'int' to 'bool'
 // MSVC warning C4310: cast truncates constant value

libcxx/utils/libcxx/test/features.py

@@ -144,6 +144,10 @@ def _mingwSupportsModules(cfg):
         when=lambda cfg: hasCompileFlag(cfg, "-Wuser-defined-warnings"),
         actions=[AddCompileFlag("-Wuser-defined-warnings")],
     ),
+    Feature(
+        name="character-conversion-warnings",
+        when=lambda cfg: hasCompileFlag(cfg, "-Wcharacter-conversion"),
+    ),
     # Tests to validate whether the compiler has a way to set the maximum number
     # of steps during constant evaluation. Since the flag differs per compiler
     # store the "valid" flag as a feature. This allows passing the proper compile