[lldb][TypeSystem] Better support for _BitInt types

lldb/source/Plugins/SymbolFile/DWARF/DWARFASTParserClang.cpp

@@ -814,13 +814,18 @@ DWARFASTParserClang::ParseTypeModifier(const SymbolContext &sc,
     // there...
     [[fallthrough]];
 
-  case DW_TAG_base_type:
+  case DW_TAG_base_type: {
     resolve_state = Type::ResolveState::Full;
+    // If a builtin type's size isn't a multiple of a byte, DWARF producers may
+    // add a precise bit-size to the type. Use the most precise bit-size
+    // possible.
+    const uint64_t bit_size = attrs.data_bit_size
+                                  ? *attrs.data_bit_size
+                                  : attrs.byte_size.value_or(0) * 8;
     clang_type = m_ast.GetBuiltinTypeForDWARFEncodingAndBitSize(
-        attrs.name.GetStringRef(), attrs.encoding,
-        attrs.byte_size.value_or(0) * 8);
+        attrs.name.GetStringRef(), attrs.encoding, bit_size);
     break;
-
+  }
   case DW_TAG_pointer_type:
     encoding_data_type = Type::eEncodingIsPointerUID;
     break;

lldb/source/Plugins/TypeSystem/Clang/TypeSystemClang.cpp

@@ -1000,6 +1000,8 @@ CompilerType TypeSystemClang::GetBuiltinTypeForDWARFEncodingAndBitSize(
 
   case DW_ATE_signed:
     if (!type_name.empty()) {
+      if (type_name.starts_with("_BitInt"))
+        return GetType(ast.getBitIntType(/*Unsigned=*/false, bit_size));
       if (type_name == "wchar_t" &&
           QualTypeMatchesBitSize(bit_size, ast, ast.WCharTy) &&
           (getTargetInfo() &&
@@ -1056,6 +1058,8 @@ CompilerType TypeSystemClang::GetBuiltinTypeForDWARFEncodingAndBitSize(
 
   case DW_ATE_unsigned:
     if (!type_name.empty()) {
+      if (type_name.starts_with("unsigned _BitInt"))
+        return GetType(ast.getBitIntType(/*Unsigned=*/true, bit_size));
       if (type_name == "wchar_t") {
         if (QualTypeMatchesBitSize(bit_size, ast, ast.WCharTy)) {
           if (!(getTargetInfo() &&
@@ -3893,6 +3897,13 @@ TypeSystemClang::GetTypeInfo(lldb::opaque_compiler_type_t type,
                            ->getModifiedType()
                            .getAsOpaquePtr(),
                        pointee_or_element_clang_type);
+  case clang::Type::BitInt: {
+    uint32_t type_flags = eTypeIsScalar | eTypeIsInteger | eTypeHasValue;
+    if (qual_type->isSignedIntegerType())
+      type_flags |= eTypeIsSigned;
+
+    return type_flags;
+  }
   case clang::Type::Builtin: {
     const clang::BuiltinType *builtin_type =
         llvm::cast<clang::BuiltinType>(qual_type->getCanonicalTypeInternal());

lldb/unittests/Symbol/TestTypeSystemClang.cpp

@@ -52,6 +52,12 @@ class TestTypeSystemClang : public testing::Test {
     return ClangUtil::GetQualType(
         m_ast->GetBuiltinTypeByName(ConstString(name)));
   }
+
+  CompilerType GetBuiltinTypeForDWARFEncodingAndBitSize(
+      llvm::StringRef type_name, uint32_t encoding, uint32_t bit_size) const {
+    return m_ast->GetBuiltinTypeForDWARFEncodingAndBitSize(type_name, encoding,
+                                                           bit_size);
+  }
 };
 
 TEST_F(TestTypeSystemClang, TestGetBasicTypeFromEnum) {
@@ -238,6 +244,91 @@ TEST_F(TestTypeSystemClang, TestBuiltinTypeForEncodingAndBitSize) {
   VerifyEncodingAndBitSize(*m_ast, eEncodingIEEE754, 64);
 }
 
+TEST_F(TestTypeSystemClang, TestGetBuiltinTypeForDWARFEncodingAndBitSize) {
+  EXPECT_FALSE(GetBuiltinTypeForDWARFEncodingAndBitSize(
+                   "_BitIn", llvm::dwarf::DW_ATE_signed, 2)
+                   .IsValid());
+  EXPECT_FALSE(GetBuiltinTypeForDWARFEncodingAndBitSize(
+                   "BitInt", llvm::dwarf::DW_ATE_signed, 2)
+                   .IsValid());
+  EXPECT_FALSE(GetBuiltinTypeForDWARFEncodingAndBitSize(
+                   "_BitInt(2)", llvm::dwarf::DW_ATE_signed_char, 2)
+                   .IsValid());
+  EXPECT_FALSE(GetBuiltinTypeForDWARFEncodingAndBitSize(
+                   "_BitInt", llvm::dwarf::DW_ATE_signed_char, 2)
+                   .IsValid());
+  EXPECT_FALSE(GetBuiltinTypeForDWARFEncodingAndBitSize(
+                   "_BitInt(2)", llvm::dwarf::DW_ATE_unsigned, 2)
+                   .IsValid());
+  EXPECT_FALSE(GetBuiltinTypeForDWARFEncodingAndBitSize(
+                   "_BitInt", llvm::dwarf::DW_ATE_unsigned, 2)
+                   .IsValid());
+
+  EXPECT_EQ(GetBuiltinTypeForDWARFEncodingAndBitSize(
+                "_BitInt(2)", llvm::dwarf::DW_ATE_signed, 2)
+                .GetTypeName(),
+            "_BitInt(2)");
+  EXPECT_EQ(GetBuiltinTypeForDWARFEncodingAndBitSize(
+                "_BitInt", llvm::dwarf::DW_ATE_signed, 2)
+                .GetTypeName(),
+            "_BitInt(2)");
+  EXPECT_EQ(GetBuiltinTypeForDWARFEncodingAndBitSize(
+                "_BitInt(129)", llvm::dwarf::DW_ATE_signed, 129)
+                .GetTypeName(),
+            "_BitInt(129)");
+  EXPECT_EQ(GetBuiltinTypeForDWARFEncodingAndBitSize(
+                "_BitInt", llvm::dwarf::DW_ATE_signed, 129)
+                .GetTypeName(),
+            "_BitInt(129)");
+
+  EXPECT_FALSE(GetBuiltinTypeForDWARFEncodingAndBitSize(
+                   "unsigned _BitIn", llvm::dwarf::DW_ATE_unsigned, 2)
+                   .IsValid());
+  EXPECT_FALSE(GetBuiltinTypeForDWARFEncodingAndBitSize(
+                   "unsigned BitInt", llvm::dwarf::DW_ATE_unsigned, 2)
+                   .IsValid());
+  EXPECT_FALSE(GetBuiltinTypeForDWARFEncodingAndBitSize(
+                   "unsigned _BitInt(2)", llvm::dwarf::DW_ATE_unsigned_char, 2)
+                   .IsValid());
+  EXPECT_FALSE(GetBuiltinTypeForDWARFEncodingAndBitSize(
+                   "unsigned _BitInt", llvm::dwarf::DW_ATE_unsigned_char, 2)
+                   .IsValid());
+  EXPECT_FALSE(GetBuiltinTypeForDWARFEncodingAndBitSize(
+                   "unsigned _BitInt(2)", llvm::dwarf::DW_ATE_signed, 2)
+                   .IsValid());
+  EXPECT_FALSE(GetBuiltinTypeForDWARFEncodingAndBitSize(
+                   "unsigned _BitInt", llvm::dwarf::DW_ATE_signed, 2)
+                   .IsValid());
+
+  EXPECT_EQ(GetBuiltinTypeForDWARFEncodingAndBitSize(
+                "unsigned _BitInt(2)", llvm::dwarf::DW_ATE_unsigned, 2)
+                .GetTypeName(),
+            "unsigned _BitInt(2)");
+  EXPECT_EQ(GetBuiltinTypeForDWARFEncodingAndBitSize(
+                "unsigned _BitInt", llvm::dwarf::DW_ATE_unsigned, 2)
+                .GetTypeName(),
+            "unsigned _BitInt(2)");
+  EXPECT_EQ(GetBuiltinTypeForDWARFEncodingAndBitSize(
+                "unsigned _BitInt(129)", llvm::dwarf::DW_ATE_unsigned, 129)
+                .GetTypeName(),
+            "unsigned _BitInt(129)");
+  EXPECT_EQ(GetBuiltinTypeForDWARFEncodingAndBitSize(
+                "unsigned _BitInt", llvm::dwarf::DW_ATE_unsigned, 129)
+                .GetTypeName(),
+            "unsigned _BitInt(129)");
+}
+
+TEST_F(TestTypeSystemClang, TestBitIntTypeInfo) {
+  EXPECT_EQ(GetBuiltinTypeForDWARFEncodingAndBitSize(
+                "_BitInt", llvm::dwarf::DW_ATE_signed, 2)
+                .GetTypeInfo(),
+            eTypeIsSigned | eTypeIsScalar | eTypeHasValue | eTypeIsInteger);
+  EXPECT_EQ(GetBuiltinTypeForDWARFEncodingAndBitSize(
+                "unsigned _BitInt", llvm::dwarf::DW_ATE_unsigned, 2)
+                .GetTypeInfo(),
+            eTypeIsScalar | eTypeHasValue | eTypeIsInteger);
+}
+
 TEST_F(TestTypeSystemClang, TestBuiltinTypeForEmptyTriple) {
   // Test that we can access type-info of builtin Clang AST
   // types without crashing even when the target triple is

lldb/unittests/SymbolFile/DWARF/DWARFASTParserClangTests.cpp

@@ -1741,3 +1741,215 @@ TEST_F(DWARFASTParserClangTests, TestTypeBitSize) {
   EXPECT_EQ(llvm::expectedToOptional(type_sp->GetByteSize(nullptr)).value_or(0),
             1U);
 }
+
+TEST_F(DWARFASTParserClangTests, TestBitIntParsing) {
+  // Tests that we correctly parse the DW_AT_base_type for a _BitInt.
+  // Older versions of Clang only emit the `_BitInt` string into the
+  // DW_AT_name (not including the bitsize). Make sure we understand
+  // those too.
+
+  const char *yamldata = R"(
+--- !ELF
+FileHeader:
+  Class:   ELFCLASS64
+  Data:    ELFDATA2LSB
+  Type:    ET_EXEC
+  Machine: EM_AARCH64
+DWARF:
+  debug_str:
+    - _BitInt(2)
+    - _BitInt
+    - unsigned _BitInt(2)
+    - unsigned _BitInt
+  debug_abbrev:
+    - ID:              0
+      Table:
+        - Code:            0x1
+          Tag:             DW_TAG_compile_unit
+          Children:        DW_CHILDREN_yes
+          Attributes:
+            - Attribute:       DW_AT_language
+              Form:            DW_FORM_data2
+        - Code:            0x2
+          Tag:             DW_TAG_base_type
+          Children:        DW_CHILDREN_no
+          Attributes:
+            - Attribute: DW_AT_name
+              Form:      DW_FORM_strp
+            - Attribute: DW_AT_encoding
+              Form:      DW_FORM_data1
+            - Attribute: DW_AT_byte_size
+              Form:      DW_FORM_data1
+            - Attribute: DW_AT_bit_size
+              Form:      DW_FORM_data1
+        - Code:            0x3
+          Tag:             DW_TAG_base_type
+          Children:        DW_CHILDREN_no
+          Attributes:
+            - Attribute: DW_AT_name
+              Form:      DW_FORM_strp
+            - Attribute: DW_AT_encoding
+              Form:      DW_FORM_data1
+            - Attribute: DW_AT_byte_size
+              Form:      DW_FORM_data1
+
+  debug_info:
+     - Version:  5
+       UnitType: DW_UT_compile
+       AddrSize: 8
+       Entries:
+
+# DW_TAG_compile_unit
+#   DW_AT_language [DW_FORM_data2]    (DW_LANG_C_plus_plus)
+
+        - AbbrCode: 0x1
+          Values:
+            - Value: 0x04
+
+#   DW_TAG_base_type
+#     DW_AT_name [DW_FORM_strp] ('_BitInt(2)')
+
+        - AbbrCode: 0x2
+          Values:
+            - Value: 0x0
+            - Value: 0x05
+            - Value: 0x01
+            - Value: 0x02
+
+#   DW_TAG_base_type
+#     DW_AT_name [DW_FORM_strp] ('_BitInt')
+
+        - AbbrCode: 0x2
+          Values:
+            - Value: 0x0b
+            - Value: 0x05
+            - Value: 0x08
+            - Value: 0x34
+
+#   DW_TAG_base_type
+#     DW_AT_name [DW_FORM_strp] ('unsigned _BitInt(2)')
+
+        - AbbrCode: 0x2
+          Values:
+            - Value: 0x13
+            - Value: 0x07
+            - Value: 0x01
+            - Value: 0x02
+
+#   DW_TAG_base_type
+#     DW_AT_name [DW_FORM_strp] ('unsigned _BitInt')
+
+        - AbbrCode: 0x2
+          Values:
+            - Value: 0x27
+            - Value: 0x07
+            - Value: 0x08
+            - Value: 0x34
+
+#   DW_TAG_base_type
+#     DW_AT_name [DW_FORM_strp] ('_BitInt')
+
+        - AbbrCode: 0x3
+          Values:
+            - Value: 0x0b
+            - Value: 0x05
+            - Value: 0x08
+...
+
+)";
+
+  YAMLModuleTester t(yamldata);
+
+  DWARFUnit *unit = t.GetDwarfUnit();
+  ASSERT_NE(unit, nullptr);
+  const DWARFDebugInfoEntry *cu_entry = unit->DIE().GetDIE();
+  ASSERT_EQ(cu_entry->Tag(), DW_TAG_compile_unit);
+  ASSERT_EQ(unit->GetDWARFLanguageType(), DW_LANG_C_plus_plus);
+  DWARFDIE cu_die(unit, cu_entry);
+
+  auto holder = std::make_unique<clang_utils::TypeSystemClangHolder>("ast");
+  auto &ast_ctx = *holder->GetAST();
+  DWARFASTParserClangStub ast_parser(ast_ctx);
+
+  auto type_die = cu_die.GetFirstChild();
+  ASSERT_TRUE(type_die.IsValid());
+
+  {
+    SymbolContext sc;
+    auto type_sp = ast_parser.ParseTypeFromDWARF(sc, type_die,
+                                                 /*type_is_new_ptr=*/nullptr);
+    ASSERT_NE(type_sp, nullptr);
+
+    EXPECT_EQ(
+        llvm::expectedToOptional(type_sp->GetByteSize(nullptr)).value_or(0),
+        1U);
+    EXPECT_EQ(type_sp->GetEncoding(), lldb::eEncodingSint);
+    EXPECT_EQ(type_sp->GetName(), "_BitInt(2)");
+    EXPECT_EQ(type_sp->GetForwardCompilerType().GetTypeName(), "_BitInt(2)");
+  }
+
+  {
+    type_die = type_die.GetSibling();
+    SymbolContext sc;
+    auto type_sp = ast_parser.ParseTypeFromDWARF(sc, type_die,
+                                                 /*type_is_new_ptr=*/nullptr);
+    ASSERT_NE(type_sp, nullptr);
+
+    EXPECT_EQ(
+        llvm::expectedToOptional(type_sp->GetByteSize(nullptr)).value_or(0),
+        8U);
+    EXPECT_EQ(type_sp->GetEncoding(), lldb::eEncodingSint);
+    EXPECT_EQ(type_sp->GetName(), "_BitInt");
+    EXPECT_EQ(type_sp->GetForwardCompilerType().GetTypeName(), "_BitInt(52)");
+  }
+
+  {
+    type_die = type_die.GetSibling();
+    SymbolContext sc;
+    auto type_sp = ast_parser.ParseTypeFromDWARF(sc, type_die,
+                                                 /*type_is_new_ptr=*/nullptr);
+    ASSERT_NE(type_sp, nullptr);
+
+    EXPECT_EQ(
+        llvm::expectedToOptional(type_sp->GetByteSize(nullptr)).value_or(0),
+        1U);
+    EXPECT_EQ(type_sp->GetEncoding(), lldb::eEncodingUint);
+    EXPECT_EQ(type_sp->GetName(), "unsigned _BitInt(2)");
+    EXPECT_EQ(type_sp->GetForwardCompilerType().GetTypeName(),
+              "unsigned _BitInt(2)");
+  }
+
+  {
+    type_die = type_die.GetSibling();
+    SymbolContext sc;
+    auto type_sp = ast_parser.ParseTypeFromDWARF(sc, type_die,
+                                                 /*type_is_new_ptr=*/nullptr);
+    ASSERT_NE(type_sp, nullptr);
+
+    EXPECT_EQ(
+        llvm::expectedToOptional(type_sp->GetByteSize(nullptr)).value_or(0),
+        8U);
+    EXPECT_EQ(type_sp->GetEncoding(), lldb::eEncodingUint);
+    EXPECT_EQ(type_sp->GetName(), "unsigned _BitInt");
+    EXPECT_EQ(type_sp->GetForwardCompilerType().GetTypeName(),
+              "unsigned _BitInt(52)");
+  }
+
+  {
+    type_die = type_die.GetSibling();
+    SymbolContext sc;
+    auto type_sp = ast_parser.ParseTypeFromDWARF(sc, type_die,
+                                                 /*type_is_new_ptr=*/nullptr);
+    ASSERT_NE(type_sp, nullptr);
+
+    EXPECT_EQ(
+        llvm::expectedToOptional(type_sp->GetByteSize(nullptr)).value_or(0),
+        8U);
+    EXPECT_EQ(type_sp->GetEncoding(), lldb::eEncodingSint);
+    EXPECT_EQ(type_sp->GetName(), "_BitInt");
+
+    // Older versions of Clang didn't emit a DW_AT_bit_size for _BitInt. In
+    // those cases we would format the CompilerType name using the byte-size.
+    EXPECT_EQ(type_sp->GetForwardCompilerType().GetTypeName(), "_BitInt(64)");
+  }
+}