Merge pull request #10257 from swiftlang/lldb/compressed-pair-to-20240723

🍒[lldb] Add support for new libc++ __compressed_pair layout

Author: Michael137

clang/include/clang/AST/ASTImporter.h

@@ -258,7 +258,6 @@ class TypeSourceInfo;
     FoundDeclsTy findDeclsInToCtx(DeclContext *DC, DeclarationName Name);
 
     void AddToLookupTable(Decl *ToD);
-    llvm::Error ImportAttrs(Decl *ToD, Decl *FromD);
 
   protected:
     /// Can be overwritten by subclasses to implement their own import logic.

clang/include/clang/AST/DeclCXX.h

@@ -1188,10 +1188,6 @@ class CXXRecordDecl : public RecordDecl {
   ///
   /// \note This does NOT include a check for union-ness.
   bool isEmpty() const { return data().Empty; }
-  /// Marks this record as empty. This is used by DWARFASTParserClang
-  /// when parsing records with empty fields having [[no_unique_address]]
-  /// attribute
-  void markEmpty() { data().Empty = true; }
 
   void setInitMethod(bool Val) { data().HasInitMethod = Val; }
   bool hasInitMethod() const { return data().HasInitMethod; }

clang/lib/AST/ASTImporter.cpp

@@ -4201,12 +4201,6 @@ ExpectedDecl ASTNodeImporter::VisitFieldDecl(FieldDecl *D) {
                               D->getInClassInitStyle()))
     return ToField;
 
-  // We need [[no_unqiue_address]] attributes to be added to FieldDecl, before
-  // we add fields in CXXRecordDecl::addedMember, otherwise record will be
-  // marked as having non-zero size.
-  Err = Importer.ImportAttrs(ToField, D);
-  if (Err)
-    return std::move(Err);
   ToField->setAccess(D->getAccess());
   ToField->setLexicalDeclContext(LexicalDC);
   ToField->setImplicit(D->isImplicit());
@@ -9423,19 +9417,6 @@ TranslationUnitDecl *ASTImporter::GetFromTU(Decl *ToD) {
   return FromDPos->second->getTranslationUnitDecl();
 }
 
-Error ASTImporter::ImportAttrs(Decl *ToD, Decl *FromD) {
-  if (!FromD->hasAttrs() || ToD->hasAttrs())
-    return Error::success();
-  for (const Attr *FromAttr : FromD->getAttrs()) {
-    auto ToAttrOrErr = Import(FromAttr);
-    if (ToAttrOrErr)
-      ToD->addAttr(*ToAttrOrErr);
-    else
-      return ToAttrOrErr.takeError();
-  }
-  return Error::success();
-}
-
 Expected<Decl *> ASTImporter::Import(Decl *FromD) {
   if (!FromD)
     return nullptr;
@@ -9569,8 +9550,15 @@ Expected<Decl *> ASTImporter::Import(Decl *FromD) {
   }
   // Make sure that ImportImpl registered the imported decl.
   assert(ImportedDecls.count(FromD) != 0 && "Missing call to MapImported?");
-  if (auto Error = ImportAttrs(ToD, FromD))
-    return std::move(Error);
+
+  if (FromD->hasAttrs())
+    for (const Attr *FromAttr : FromD->getAttrs()) {
+      auto ToAttrOrErr = Import(FromAttr);
+      if (ToAttrOrErr)
+        ToD->addAttr(*ToAttrOrErr);
+      else
+        return ToAttrOrErr.takeError();
+    }
 
   // Notify subclasses.
   Imported(FromD, ToD);

clang/unittests/AST/ASTImporterTest.cpp

@@ -9376,29 +9376,6 @@ TEST_P(ASTImporterOptionSpecificTestBase, VaListCpp) {
       ToVaList->getUnderlyingType(), ToBuiltinVaList->getUnderlyingType()));
 }
 
-TEST_P(ASTImporterOptionSpecificTestBase,
-       ImportDefinitionOfEmptyClassWithNoUniqueAddressField) {
-  Decl *FromTU = getTuDecl(
-      R"(
-      struct B {};
-      struct A { B b; };
-      )",
-      Lang_CXX20);
-
-  CXXRecordDecl *FromD = FirstDeclMatcher<CXXRecordDecl>().match(
-      FromTU, cxxRecordDecl(hasName("A")));
-
-  for (auto *FD : FromD->fields())
-    FD->addAttr(clang::NoUniqueAddressAttr::Create(FromD->getASTContext(),
-                                                   clang::SourceRange()));
-  FromD->markEmpty();
-
-  CXXRecordDecl *ToD = Import(FromD, Lang_CXX20);
-  EXPECT_TRUE(ToD->isEmpty());
-  for (auto *FD : ToD->fields())
-    EXPECT_EQ(true, FD->hasAttr<NoUniqueAddressAttr>());
-}
-
 TEST_P(ASTImporterOptionSpecificTestBase, ImportExistingTypedefToRecord) {
   const char *Code =
       R"(

lldb/examples/synthetic/libcxx.py

@@ -721,6 +721,12 @@ def _get_value_of_compressed_pair(self, pair):
     def update(self):
         logger = lldb.formatters.Logger.Logger()
         try:
+            has_compressed_pair_layout = True
+            alloc_valobj = self.valobj.GetChildMemberWithName("__alloc_")
+            size_valobj = self.valobj.GetChildMemberWithName("__size_")
+            if alloc_valobj.IsValid() and size_valobj.IsValid():
+                has_compressed_pair_layout = False
+
             # A deque is effectively a two-dim array, with fixed width.
             # 'map' contains pointers to the rows of this array. The
             # full memory area allocated by the deque is delimited
@@ -734,9 +740,13 @@ def update(self):
             # variable tells which element in this NxM array is the 0th
             # one, and the 'size' element gives the number of elements
             # in the deque.
-            count = self._get_value_of_compressed_pair(
-                self.valobj.GetChildMemberWithName("__size_")
-            )
+            if has_compressed_pair_layout:
+                count = self._get_value_of_compressed_pair(
+                    self.valobj.GetChildMemberWithName("__size_")
+                )
+            else:
+                count = size_valobj.GetValueAsUnsigned(0)
+
             # give up now if we cant access memory reliably
             if self.block_size < 0:
                 logger.write("block_size < 0")
@@ -748,9 +758,15 @@ def update(self):
             self.map_begin = map_.GetChildMemberWithName("__begin_")
             map_begin = self.map_begin.GetValueAsUnsigned(0)
             map_end = map_.GetChildMemberWithName("__end_").GetValueAsUnsigned(0)
-            map_endcap = self._get_value_of_compressed_pair(
-                map_.GetChildMemberWithName("__end_cap_")
-            )
+
+            if has_compressed_pair_layout:
+                map_endcap = self._get_value_of_compressed_pair(
+                    map_.GetChildMemberWithName("__end_cap_")
+                )
+            else:
+                map_endcap = map_.GetChildMemberWithName(
+                    "__end_cap_"
+                ).GetValueAsUnsigned(0)
 
             # check consistency
             if not map_first <= map_begin <= map_end <= map_endcap:

lldb/include/lldb/Symbol/SymbolFile.h

@@ -217,7 +217,15 @@ class SymbolFile : public PluginInterface {
   /// The characteristics of an array type.
   struct ArrayInfo {
     int64_t first_index = 0;
-    llvm::SmallVector<uint64_t, 1> element_orders;
+
+    ///< Each entry belongs to a distinct DW_TAG_subrange_type.
+    ///< For multi-dimensional DW_TAG_array_types we would have
+    ///< an entry for each dimension. An entry represents the
+    ///< optional element count of the subrange.
+    ///
+    ///< The order of entries follows the order of the DW_TAG_subrange_type
+    ///< children of this DW_TAG_array_type.
+    llvm::SmallVector<std::optional<uint64_t>, 1> element_orders;
     uint32_t byte_stride = 0;
     uint32_t bit_stride = 0;
   };

lldb/source/Plugins/Language/CPlusPlus/LibCxx.cpp

@@ -27,13 +27,40 @@
 #include "Plugins/LanguageRuntime/CPlusPlus/CPPLanguageRuntime.h"
 #include "Plugins/TypeSystem/Clang/TypeSystemClang.h"
 #include "lldb/lldb-enumerations.h"
+#include "lldb/lldb-forward.h"
 #include <optional>
 #include <tuple>
 
 using namespace lldb;
 using namespace lldb_private;
 using namespace lldb_private::formatters;
 
+static void consumeInlineNamespace(llvm::StringRef &name) {
+  // Delete past an inline namespace, if any: __[a-zA-Z0-9_]+::
+  auto scratch = name;
+  if (scratch.consume_front("__") && std::isalnum(scratch[0])) {
+    scratch = scratch.drop_while([](char c) { return std::isalnum(c); });
+    if (scratch.consume_front("::")) {
+      // Successfully consumed a namespace.
+      name = scratch;
+    }
+  }
+}
+
+bool lldb_private::formatters::isOldCompressedPairLayout(
+    ValueObject &pair_obj) {
+  return isStdTemplate(pair_obj.GetTypeName(), "__compressed_pair");
+}
+
+bool lldb_private::formatters::isStdTemplate(ConstString type_name,
+                                             llvm::StringRef type) {
+  llvm::StringRef name = type_name.GetStringRef();
+  // The type name may be prefixed with `std::__<inline-namespace>::`.
+  if (name.consume_front("std::"))
+    consumeInlineNamespace(name);
+  return name.consume_front(type) && name.starts_with("<");
+}
+
 lldb::ValueObjectSP lldb_private::formatters::GetChildMemberWithName(
     ValueObject &obj, llvm::ArrayRef<ConstString> alternative_names) {
   for (ConstString name : alternative_names) {
@@ -53,7 +80,7 @@ lldb_private::formatters::GetFirstValueOfLibCXXCompressedPair(
   if (first_child)
     value = first_child->GetChildMemberWithName("__value_");
   if (!value) {
-    // pre-r300140 member name
+    // pre-c88580c member name
     value = pair.GetChildMemberWithName("__first_");
   }
   return value;
@@ -70,7 +97,7 @@ lldb_private::formatters::GetSecondValueOfLibCXXCompressedPair(
     }
   }
   if (!value) {
-    // pre-r300140 member name
+    // pre-c88580c member name
     value = pair.GetChildMemberWithName("__second_");
   }
   return value;
@@ -176,7 +203,9 @@ bool lldb_private::formatters::LibcxxUniquePointerSummaryProvider(
   if (!ptr_sp)
     return false;
 
-  ptr_sp = GetFirstValueOfLibCXXCompressedPair(*ptr_sp);
+  if (isOldCompressedPairLayout(*ptr_sp))
+    ptr_sp = GetFirstValueOfLibCXXCompressedPair(*ptr_sp);
+
   if (!ptr_sp)
     return false;
 
@@ -363,13 +392,22 @@ lldb_private::formatters::LibcxxUniquePtrSyntheticFrontEnd::Update() {
 
   // Retrieve the actual pointer and the deleter, and clone them to give them
   // user-friendly names.
-  ValueObjectSP value_pointer_sp = GetFirstValueOfLibCXXCompressedPair(*ptr_sp);
-  if (value_pointer_sp)
-    m_value_ptr_sp = value_pointer_sp->Clone(ConstString("pointer"));
+  if (isOldCompressedPairLayout(*ptr_sp)) {
+    if (ValueObjectSP value_pointer_sp =
+            GetFirstValueOfLibCXXCompressedPair(*ptr_sp))
+      m_value_ptr_sp = value_pointer_sp->Clone(ConstString("pointer"));
+
+    if (ValueObjectSP deleter_sp =
+            GetSecondValueOfLibCXXCompressedPair(*ptr_sp))
+      m_deleter_sp = deleter_sp->Clone(ConstString("deleter"));
+  } else {
+    m_value_ptr_sp = ptr_sp->Clone(ConstString("pointer"));
 
-  ValueObjectSP deleter_sp = GetSecondValueOfLibCXXCompressedPair(*ptr_sp);
-  if (deleter_sp)
-    m_deleter_sp = deleter_sp->Clone(ConstString("deleter"));
+    if (ValueObjectSP deleter_sp =
+            valobj_sp->GetChildMemberWithName("__deleter_"))
+      if (deleter_sp->GetNumChildrenIgnoringErrors() > 0)
+        m_deleter_sp = deleter_sp->Clone(ConstString("deleter"));
+  }
 
   return lldb::ChildCacheState::eRefetch;
 }
@@ -407,24 +445,27 @@ namespace {
 enum class StringLayout { CSD, DSC };
 }
 
+static ValueObjectSP ExtractLibCxxStringData(ValueObject &valobj) {
+  if (auto rep_sp = valobj.GetChildMemberWithName("__rep_"))
+    return rep_sp;
+
+  ValueObjectSP valobj_r_sp = valobj.GetChildMemberWithName("__r_");
+  if (!valobj_r_sp || !valobj_r_sp->GetError().Success())
+    return nullptr;
+
+  if (!isOldCompressedPairLayout(*valobj_r_sp))
+    return nullptr;
+
+  return GetFirstValueOfLibCXXCompressedPair(*valobj_r_sp);
+}
+
 /// Determine the size in bytes of \p valobj (a libc++ std::string object) and
 /// extract its data payload. Return the size + payload pair.
 // TODO: Support big-endian architectures.
 static std::optional<std::pair<uint64_t, ValueObjectSP>>
 ExtractLibcxxStringInfo(ValueObject &valobj) {
-  ValueObjectSP valobj_r_sp = valobj.GetChildMemberWithName("__r_");
-  if (!valobj_r_sp || !valobj_r_sp->GetError().Success())
-    return {};
-
-  // __r_ is a compressed_pair of the actual data and the allocator. The data we
-  // want is in the first base class.
-  ValueObjectSP valobj_r_base_sp = valobj_r_sp->GetChildAtIndex(0);
-  if (!valobj_r_base_sp)
-    return {};
-
-  ValueObjectSP valobj_rep_sp =
-      valobj_r_base_sp->GetChildMemberWithName("__value_");
-  if (!valobj_rep_sp)
+  ValueObjectSP valobj_rep_sp = ExtractLibCxxStringData(valobj);
+  if (!valobj_rep_sp || !valobj_rep_sp->GetError().Success())
     return {};
 
   ValueObjectSP l = valobj_rep_sp->GetChildMemberWithName("__l");

lldb/source/Plugins/Language/CPlusPlus/LibCxx.h

@@ -25,7 +25,8 @@ GetChildMemberWithName(ValueObject &obj,
 
 lldb::ValueObjectSP GetFirstValueOfLibCXXCompressedPair(ValueObject &pair);
 lldb::ValueObjectSP GetSecondValueOfLibCXXCompressedPair(ValueObject &pair);
-
+bool isOldCompressedPairLayout(ValueObject &pair_obj);
+bool isStdTemplate(ConstString type_name, llvm::StringRef type);
 
 bool LibcxxStringSummaryProviderASCII(
     ValueObject &valobj, Stream &stream,