[PDB] Parse UDT symbols and pointers to members (combined patch)

Summary:
In this patch I've tried to combine the best ideas from D49368 and D49410,
so it implements following:

- Completion of UDTs from a PDB with a filling of a layout info;
- Pointers to members;
- Fixes the bug relating to a virtual base offset reading from `vbtable`.
  The offset was treated as an unsigned, but it can be a negative sometimes.
- Support of MSInheritance attribute

Reviewers: asmith, zturner, rnk, labath, clayborg, lldb-commits

Reviewed By: zturner

Subscribers: aleksandr.urakov, stella.stamenova, JDevlieghere, lldb-commits

Differential Revision: https://reviews.llvm.org/D49980

llvm-svn: 339649
diff --git a/lldb/source/Plugins/SymbolFile/PDB/PDBASTParser.cpp b/lldb/source/Plugins/SymbolFile/PDB/PDBASTParser.cpp
index 8bea994..713fa6e 100644
--- a/lldb/source/Plugins/SymbolFile/PDB/PDBASTParser.cpp
+++ b/lldb/source/Plugins/SymbolFile/PDB/PDBASTParser.cpp
@@ -9,11 +9,15 @@
 
 #include "PDBASTParser.h"
 
+#include "SymbolFilePDB.h"
+
 #include "clang/AST/CharUnits.h"
 #include "clang/AST/Decl.h"
 #include "clang/AST/DeclCXX.h"
 
+#include "lldb/Core/Module.h"
 #include "lldb/Symbol/ClangASTContext.h"
+#include "lldb/Symbol/ClangExternalASTSourceCommon.h"
 #include "lldb/Symbol/ClangUtil.h"
 #include "lldb/Symbol/Declaration.h"
 #include "lldb/Symbol/SymbolFile.h"
@@ -48,8 +52,9 @@
     return clang::TTK_Union;
   case PDB_UdtType::Interface:
     return clang::TTK_Interface;
+  default:
+    llvm_unreachable("unsuported PDB UDT type");
   }
-  return -1;
 }
 
 lldb::Encoding TranslateBuiltinEncoding(PDB_BuiltinType type) {
@@ -199,6 +204,68 @@
   decl.SetLine(first_line_up->getLineNumber());
   return true;
 }
+
+AccessType TranslateMemberAccess(PDB_MemberAccess access) {
+  switch (access) {
+  case PDB_MemberAccess::Private:
+    return eAccessPrivate;
+  case PDB_MemberAccess::Protected:
+    return eAccessProtected;
+  case PDB_MemberAccess::Public:
+    return eAccessPublic;
+  default:
+    return eAccessNone;
+  }
+}
+
+AccessType GetDefaultAccessibilityForUdtKind(PDB_UdtType udt_kind) {
+  switch (udt_kind) {
+  case PDB_UdtType::Struct:
+  case PDB_UdtType::Union:
+    return eAccessPublic;
+  case PDB_UdtType::Class:
+  case PDB_UdtType::Interface:
+    return eAccessPrivate;
+  default:
+    llvm_unreachable("unsupported PDB UDT type");
+  }
+}
+
+AccessType GetAccessibilityForUdt(const PDBSymbolTypeUDT &udt) {
+  AccessType access = TranslateMemberAccess(udt.getAccess());
+  if (access != lldb::eAccessNone || !udt.isNested())
+    return access;
+
+  auto parent = udt.getClassParent();
+  if (!parent)
+    return lldb::eAccessNone;
+
+  auto parent_udt = llvm::dyn_cast<PDBSymbolTypeUDT>(parent.get());
+  if (!parent_udt)
+    return lldb::eAccessNone;
+
+  return GetDefaultAccessibilityForUdtKind(parent_udt->getUdtKind());
+}
+
+clang::MSInheritanceAttr::Spelling GetMSInheritance(
+    const PDBSymbolTypeUDT &udt) {
+  int base_count = 0;
+  bool has_virtual = false;
+
+  auto bases_enum = udt.findAllChildren<PDBSymbolTypeBaseClass>();
+  if (bases_enum) {
+    while (auto base = bases_enum->getNext()) {
+      base_count++;
+      has_virtual |= base->isVirtualBaseClass();
+    }
+  }
+
+  if (has_virtual)
+    return clang::MSInheritanceAttr::Keyword_virtual_inheritance;
+  if (base_count > 1)
+    return clang::MSInheritanceAttr::Keyword_multiple_inheritance;
+  return clang::MSInheritanceAttr::Keyword_single_inheritance;
+}
 } // namespace
 
 PDBASTParser::PDBASTParser(lldb_private::ClangASTContext &ast) : m_ast(ast) {}
@@ -216,29 +283,90 @@
   Declaration decl;
 
   switch (type.getSymTag()) {
+  case PDB_SymType::BaseClass: {
+    auto symbol_file = m_ast.GetSymbolFile();
+    if (!symbol_file)
+      return nullptr;
+
+    auto ty = symbol_file->ResolveTypeUID(type.getRawSymbol().getTypeId());
+    return ty ? ty->shared_from_this() : nullptr;
+  } break;
   case PDB_SymType::UDT: {
     auto udt = llvm::dyn_cast<PDBSymbolTypeUDT>(&type);
     assert(udt);
-    AccessType access = lldb::eAccessPublic;
-    PDB_UdtType udt_kind = udt->getUdtKind();
-    auto tag_type_kind = TranslateUdtKind(udt_kind);
-    if (tag_type_kind == -1)
+
+    // Note that, unnamed UDT being typedef-ed is generated as a UDT symbol
+    // other than a Typedef symbol in PDB. For example,
+    //    typedef union { short Row; short Col; } Union;
+    // is generated as a named UDT in PDB:
+    //    union Union { short Row; short Col; }
+    // Such symbols will be handled here.
+
+    // Some UDT with trival ctor has zero length. Just ignore.
+    if (udt->getLength() == 0)
       return nullptr;
 
-    if (udt_kind == PDB_UdtType::Class)
-      access = lldb::eAccessPrivate;
+    // Ignore unnamed-tag UDTs.
+    if (udt->getName().empty())
+      return nullptr;
+
+    auto access = GetAccessibilityForUdt(*udt);
+
+    auto tag_type_kind = TranslateUdtKind(udt->getUdtKind());
+
+    ClangASTMetadata metadata;
+    metadata.SetUserID(type.getSymIndexId());
+    metadata.SetIsDynamicCXXType(false);
 
     CompilerType clang_type = m_ast.CreateRecordType(
         tu_decl_ctx, access, udt->getName().c_str(), tag_type_kind,
-        lldb::eLanguageTypeC_plus_plus, nullptr);
+        lldb::eLanguageTypeC_plus_plus, &metadata);
+    assert(clang_type.IsValid());
 
-    m_ast.SetHasExternalStorage(clang_type.GetOpaqueQualType(), true);
+    if (udt->isConstType())
+      clang_type = clang_type.AddConstModifier();
 
+    if (udt->isVolatileType())
+      clang_type = clang_type.AddVolatileModifier();
+
+    clang::CXXRecordDecl *record_decl =
+      m_ast.GetAsCXXRecordDecl(clang_type.GetOpaqueQualType());
+    assert(record_decl);
+    auto inheritance_attr = clang::MSInheritanceAttr::CreateImplicit(
+        *m_ast.getASTContext(), GetMSInheritance(*udt));
+    record_decl->addAttr(inheritance_attr);
+
+    ClangASTContext::StartTagDeclarationDefinition(clang_type);
+
+    Type::ResolveStateTag type_resolve_state_tag;
+    auto children = udt->findAllChildren();
+    if (!children || children->getChildCount() == 0) {
+      // PDB does not have symbol of forwarder. We assume we get an udt w/o any
+      // fields. Just complete it at this point.
+      ClangASTContext::CompleteTagDeclarationDefinition(clang_type);
+
+      m_ast.SetHasExternalStorage(clang_type.GetOpaqueQualType(), false);
+
+      type_resolve_state_tag = Type::eResolveStateFull;
+    } else {
+      // Add the type to the forward declarations. It will help us to avoid
+      // an endless recursion in CompleteTypeFromUdt function.
+      auto clang_type_removed_fast_quals =
+          ClangUtil::RemoveFastQualifiers(clang_type).GetOpaqueQualType();
+      m_forward_decl_clang_type_to_uid[clang_type_removed_fast_quals] =
+          type.getSymIndexId();
+
+      m_ast.SetHasExternalStorage(clang_type.GetOpaqueQualType(), true);
+
+      type_resolve_state_tag = Type::eResolveStateForward;
+    }
+
+    GetDeclarationForSymbol(type, decl);
     return std::make_shared<lldb_private::Type>(
         type.getSymIndexId(), m_ast.GetSymbolFile(),
         ConstString(udt->getName()), udt->getLength(), nullptr,
         LLDB_INVALID_UID, lldb_private::Type::eEncodingIsUID, decl, clang_type,
-        lldb_private::Type::eResolveStateForward);
+        type_resolve_state_tag);
   } break;
   case PDB_SymType::Enum: {
     auto enum_type = llvm::dyn_cast<PDBSymbolTypeEnum>(&type);
@@ -441,6 +569,26 @@
     if (!pointee_type)
       return nullptr;
 
+    if (pointer_type->isPointerToDataMember() ||
+        pointer_type->isPointerToMemberFunction()) {
+      auto class_parent_uid = pointer_type->getRawSymbol().getClassParentId();
+      auto class_parent_type =
+          m_ast.GetSymbolFile()->ResolveTypeUID(class_parent_uid);
+      assert(class_parent_type);
+
+      CompilerType pointer_ast_type;
+      pointer_ast_type = ClangASTContext::CreateMemberPointerType(
+          class_parent_type->GetLayoutCompilerType(),
+          pointee_type->GetForwardCompilerType());
+      assert(pointer_ast_type);
+
+      return std::make_shared<lldb_private::Type>(
+          pointer_type->getSymIndexId(), m_ast.GetSymbolFile(), ConstString(),
+          pointer_type->getLength(), nullptr, LLDB_INVALID_UID,
+          lldb_private::Type::eEncodingIsUID, decl, pointer_ast_type,
+          lldb_private::Type::eResolveStateForward);
+    }
+
     CompilerType pointer_ast_type;
     pointer_ast_type = pointee_type->GetFullCompilerType();
     if (pointer_type->isReference())
@@ -471,6 +619,47 @@
   return nullptr;
 }
 
+bool PDBASTParser::CompleteTypeFromPDB(
+    lldb_private::CompilerType &compiler_type) {
+  if (GetClangASTImporter().CanImport(compiler_type))
+    return GetClangASTImporter().CompleteType(compiler_type);
+
+  // Remove the type from the forward declarations to avoid
+  // an endless recursion for types like a linked list.
+  CompilerType compiler_type_no_qualifiers =
+      ClangUtil::RemoveFastQualifiers(compiler_type);
+  auto uid_it = m_forward_decl_clang_type_to_uid.find(
+      compiler_type_no_qualifiers.GetOpaqueQualType());
+  if (uid_it == m_forward_decl_clang_type_to_uid.end())
+    return true;
+
+  auto symbol_file = static_cast<SymbolFilePDB *>(m_ast.GetSymbolFile());
+  if (!symbol_file)
+    return false;
+
+  std::unique_ptr<PDBSymbol> symbol =
+      symbol_file->GetPDBSession().getSymbolById(uid_it->getSecond());
+  if (!symbol)
+    return false;
+
+  m_forward_decl_clang_type_to_uid.erase(uid_it);
+
+  ClangASTContext::SetHasExternalStorage(compiler_type.GetOpaqueQualType(),
+                                         false);
+
+  switch (symbol->getSymTag()) {
+  case PDB_SymType::UDT: {
+    auto udt = llvm::dyn_cast<PDBSymbolTypeUDT>(symbol.get());
+    if (!udt)
+      return false;
+
+    return CompleteTypeFromUDT(*symbol_file, compiler_type, *udt);
+  }
+  default:
+    llvm_unreachable("not a forward clang type decl!");
+  }
+}
+
 bool PDBASTParser::AddEnumValue(CompilerType enum_type,
                                 const PDBSymbolData &enum_value) const {
   Declaration decl;
@@ -513,3 +702,187 @@
       enum_type.GetOpaqueQualType(), underlying_type, decl, name.c_str(),
       raw_value, byte_size * 8);
 }
+
+bool PDBASTParser::CompleteTypeFromUDT(
+    lldb_private::SymbolFile &symbol_file,
+    lldb_private::CompilerType &compiler_type,
+    llvm::pdb::PDBSymbolTypeUDT &udt) {
+  ClangASTImporter::LayoutInfo layout_info;
+  layout_info.bit_size = udt.getLength() * 8;
+
+  auto nested_enums = udt.findAllChildren<PDBSymbolTypeUDT>();
+  if (nested_enums)
+    while (auto nested = nested_enums->getNext())
+      symbol_file.ResolveTypeUID(nested->getSymIndexId());
+
+  auto bases_enum = udt.findAllChildren<PDBSymbolTypeBaseClass>();
+  if (bases_enum)
+    AddRecordBases(symbol_file, compiler_type,
+                   TranslateUdtKind(udt.getUdtKind()), *bases_enum,
+                   layout_info);
+
+  auto members_enum = udt.findAllChildren<PDBSymbolData>();
+  if (members_enum)
+    AddRecordMembers(symbol_file, compiler_type, *members_enum, layout_info);
+
+  auto methods_enum = udt.findAllChildren<PDBSymbolFunc>();
+  if (methods_enum)
+    AddRecordMethods(symbol_file, compiler_type, *methods_enum);
+
+  m_ast.AddMethodOverridesForCXXRecordType(compiler_type.GetOpaqueQualType());
+  ClangASTContext::BuildIndirectFields(compiler_type);
+  ClangASTContext::CompleteTagDeclarationDefinition(compiler_type);
+
+  clang::CXXRecordDecl *record_decl =
+      m_ast.GetAsCXXRecordDecl(compiler_type.GetOpaqueQualType());
+  if (!record_decl)
+    return static_cast<bool>(compiler_type);
+
+  GetClangASTImporter().InsertRecordDecl(record_decl, layout_info);
+
+  return static_cast<bool>(compiler_type);
+}
+
+void PDBASTParser::AddRecordMembers(
+    lldb_private::SymbolFile &symbol_file,
+    lldb_private::CompilerType &record_type,
+    PDBDataSymbolEnumerator &members_enum,
+    lldb_private::ClangASTImporter::LayoutInfo &layout_info) const {
+  while (auto member = members_enum.getNext()) {
+    if (member->isCompilerGenerated())
+        continue;
+
+    auto member_name = member->getName();
+
+    auto member_type = symbol_file.ResolveTypeUID(member->getTypeId());
+    if (!member_type)
+      continue;
+
+    auto member_comp_type = member_type->GetLayoutCompilerType();
+    if (!member_comp_type.GetCompleteType()) {
+      symbol_file.GetObjectFile()->GetModule()->ReportError(
+          ":: Class '%s' has a member '%s' of type '%s' "
+          "which does not have a complete definition.",
+          record_type.GetTypeName().GetCString(), member_name.c_str(),
+          member_comp_type.GetTypeName().GetCString());
+      if (ClangASTContext::StartTagDeclarationDefinition(member_comp_type))
+        ClangASTContext::CompleteTagDeclarationDefinition(member_comp_type);
+    }
+
+    auto access = TranslateMemberAccess(member->getAccess());
+
+    switch (member->getDataKind()) {
+    case PDB_DataKind::Member: {
+      auto location_type = member->getLocationType();
+
+      auto bit_size = member->getLength();
+      if (location_type == PDB_LocType::ThisRel)
+        bit_size *= 8;
+
+      auto decl = ClangASTContext::AddFieldToRecordType(
+          record_type, member_name.c_str(), member_comp_type, access, bit_size);
+      if (!decl)
+        continue;
+
+      auto offset = member->getOffset() * 8;
+      if (location_type == PDB_LocType::BitField)
+        offset += member->getBitPosition();
+
+      layout_info.field_offsets.insert(std::make_pair(decl, offset));
+
+      break;
+    }
+    case PDB_DataKind::StaticMember:
+      ClangASTContext::AddVariableToRecordType(record_type, member_name.c_str(),
+                                               member_comp_type, access);
+      break;
+    default:
+      llvm_unreachable("unsupported PDB data kind");
+    }
+  }
+}
+
+void PDBASTParser::AddRecordBases(
+    lldb_private::SymbolFile &symbol_file,
+    lldb_private::CompilerType &record_type, int record_kind,
+    PDBBaseClassSymbolEnumerator &bases_enum,
+    lldb_private::ClangASTImporter::LayoutInfo &layout_info) const {
+  std::vector<clang::CXXBaseSpecifier *> base_classes;
+  while (auto base = bases_enum.getNext()) {
+    auto base_type = symbol_file.ResolveTypeUID(base->getTypeId());
+    if (!base_type)
+      continue;
+
+    auto base_comp_type = base_type->GetFullCompilerType();
+    if (!base_comp_type.GetCompleteType()) {
+      symbol_file.GetObjectFile()->GetModule()->ReportError(
+          ":: Class '%s' has a base class '%s' "
+          "which does not have a complete definition.",
+          record_type.GetTypeName().GetCString(),
+          base_comp_type.GetTypeName().GetCString());
+      if (ClangASTContext::StartTagDeclarationDefinition(base_comp_type))
+        ClangASTContext::CompleteTagDeclarationDefinition(base_comp_type);
+    }
+
+    auto access = TranslateMemberAccess(base->getAccess());
+
+    auto is_virtual = base->isVirtualBaseClass();
+
+    auto base_class_spec = m_ast.CreateBaseClassSpecifier(
+        base_comp_type.GetOpaqueQualType(), access, is_virtual,
+        record_kind == clang::TTK_Class);
+    if (!base_class_spec)
+        continue;
+
+    base_classes.push_back(base_class_spec);
+
+    if (is_virtual)
+        continue;
+
+    auto decl = m_ast.GetAsCXXRecordDecl(base_comp_type.GetOpaqueQualType());
+    if (!decl)
+      continue;
+
+    auto offset = clang::CharUnits::fromQuantity(base->getOffset());
+    layout_info.base_offsets.insert(std::make_pair(decl, offset));
+  }
+  if (!base_classes.empty()) {
+    m_ast.SetBaseClassesForClassType(record_type.GetOpaqueQualType(),
+                                     &base_classes.front(),
+                                     base_classes.size());
+    ClangASTContext::DeleteBaseClassSpecifiers(&base_classes.front(),
+                                               base_classes.size());
+  }
+}
+
+void PDBASTParser::AddRecordMethods(
+    lldb_private::SymbolFile &symbol_file,
+    lldb_private::CompilerType &record_type,
+    PDBFuncSymbolEnumerator &methods_enum) const {
+  while (auto method = methods_enum.getNext()) {
+    auto method_type = symbol_file.ResolveTypeUID(method->getSymIndexId());
+    // MSVC specific __vecDelDtor.
+    if (!method_type)
+      break;
+
+    auto method_comp_type = method_type->GetFullCompilerType();
+    if (!method_comp_type.GetCompleteType()) {
+      symbol_file.GetObjectFile()->GetModule()->ReportError(
+          ":: Class '%s' has a method '%s' whose type cannot be completed.",
+          record_type.GetTypeName().GetCString(),
+          method_comp_type.GetTypeName().GetCString());
+      if (ClangASTContext::StartTagDeclarationDefinition(method_comp_type))
+        ClangASTContext::CompleteTagDeclarationDefinition(method_comp_type);
+    }
+
+    // TODO: get mangled name for the method.
+    m_ast.AddMethodToCXXRecordType(
+        record_type.GetOpaqueQualType(), method->getName().c_str(),
+        /*mangled_name*/ nullptr, method_comp_type,
+        TranslateMemberAccess(method->getAccess()), method->isVirtual(),
+        method->isStatic(), method->hasInlineAttribute(),
+        /*is_explicit*/ false, // FIXME: Need this field in CodeView.
+        /*is_attr_used*/ false,
+        /*is_artificial*/ method->isCompilerGenerated());
+  }
+}