gsl::Owner/gsl::Pointer: Add implicit annotations for some std types

Summary:
Hard code gsl::Owner/gsl::Pointer for std types. The paper mentions
some types explicitly. Generally, all containers and their iterators are
covered. For iterators, we cover both the case that they are defined
as an nested class or as an typedef/using. I have started to test this
implementation against some real standard library implementations, namely
libc++ 7.1.0, libc++ 8.0.1rc2, libstdc++ 4.6.4, libstdc++ 4.8.5,
libstdc++ 4.9.4, libstdc++ 5.4.0, libstdc++ 6.5.0, libstdc++ 7.3.0,
libstdc++ 8.3.0 and libstdc++ 9.1.0.

The tests are currently here
  https://github.com/mgehre/llvm-project/blob/lifetime-ci/lifetime-attr-test.sh
  https://github.com/mgehre/llvm-project/blob/lifetime-ci/lifetime-attr-test.cpp
I think due to their dependency on a standard library, they are not a good fit
for clang/test/. Where else could I put them?

Reviewers: gribozavr, xazax.hun

Subscribers: rnkovacs, cfe-commits

Tags: #clang

Differential Revision: https://reviews.llvm.org/D64448

llvm-svn: 368147

clang/include/clang/Basic/AttrDocs.td

@@ -4249,7 +4249,7 @@ object of type ``T``:
     int *getInt() { return &value; }
   };
 
-The argument ``T`` is optional and currently ignored.
+The argument ``T`` is optional and is ignored.
 This attribute may be used by analysis tools and has no effect on code
 generation.
 
@@ -4275,7 +4275,7 @@ like pointers to an object of type ``T``:
     int *getInt() { return &valuePointer; }
   };
 
-The argument ``T`` is optional and currently ignored.
+The argument ``T`` is optional and is ignored.
 This attribute may be used by analysis tools and has no effect on code
 generation.
 

clang/include/clang/Sema/Sema.h

@@ -6116,6 +6116,17 @@ public:
       ClassTemplateSpecializationDecl *BaseTemplateSpec,
       SourceLocation BaseLoc);
 
+  /// Add gsl::Pointer attribute to std::container::iterator
+  /// \param ND The declaration that introduces the name
+  /// std::container::iterator. \param UnderlyingRecord The record named by ND.
+  void inferGslPointerAttribute(NamedDecl *ND, CXXRecordDecl *UnderlyingRecord);
+
+  /// Add [[gsl::Owner]] and [[gsl::Pointer]] attributes for std:: types.
+  void inferGslOwnerPointerAttribute(CXXRecordDecl *Record);
+
+  /// Add [[gsl::Pointer]] attributes for std:: types.
+  void inferGslPointerAttribute(TypedefNameDecl *TD);
+
   void CheckCompletedCXXClass(CXXRecordDecl *Record);
 
   /// Check that the C++ class annoated with "trivial_abi" satisfies all the

clang/lib/Sema/SemaAttr.cpp

@@ -85,6 +85,126 @@ void Sema::AddMsStructLayoutForRecord(RecordDecl *RD) {
         MSVtorDispAttr::CreateImplicit(Context, VtorDispStack.CurrentValue));
 }
 
+template <typename Attribute>
+static void addGslOwnerPointerAttributeIfNotExisting(ASTContext &Context,
+                                                     CXXRecordDecl *Record) {
+  CXXRecordDecl *Canonical = Record->getCanonicalDecl();
+  if (Canonical->hasAttr<OwnerAttr>() || Canonical->hasAttr<PointerAttr>())
+    return;
+
+  Canonical->addAttr(::new (Context) Attribute(SourceRange{}, Context,
+                                               /*DerefType*/ nullptr,
+                                               /*Spelling=*/0));
+}
+
+void Sema::inferGslPointerAttribute(NamedDecl *ND,
+                                    CXXRecordDecl *UnderlyingRecord) {
+  if (!UnderlyingRecord)
+    return;
+
+  const auto *Parent = dyn_cast<CXXRecordDecl>(ND->getDeclContext());
+  if (!Parent)
+    return;
+
+  static llvm::StringSet<> Containers{
+      "array",
+      "basic_string",
+      "deque",
+      "forward_list",
+      "vector",
+      "list",
+      "map",
+      "multiset",
+      "multimap",
+      "priority_queue",
+      "queue",
+      "set",
+      "stack",
+      "unordered_set",
+      "unordered_map",
+      "unordered_multiset",
+      "unordered_multimap",
+  };
+
+  static llvm::StringSet<> Iterators{"iterator", "const_iterator",
+                                     "reverse_iterator",
+                                     "const_reverse_iterator"};
+
+  if (Parent->isInStdNamespace() && Iterators.count(ND->getName()) &&
+      Containers.count(Parent->getName()))
+    addGslOwnerPointerAttributeIfNotExisting<PointerAttr>(Context,
+                                                          UnderlyingRecord);
+}
+
+void Sema::inferGslPointerAttribute(TypedefNameDecl *TD) {
+
+  QualType Canonical = TD->getUnderlyingType().getCanonicalType();
+
+  CXXRecordDecl *RD = Canonical->getAsCXXRecordDecl();
+  if (!RD) {
+    if (auto *TST =
+            dyn_cast<TemplateSpecializationType>(Canonical.getTypePtr())) {
+
+      RD = dyn_cast_or_null<CXXRecordDecl>(
+          TST->getTemplateName().getAsTemplateDecl()->getTemplatedDecl());
+    }
+  }
+
+  inferGslPointerAttribute(TD, RD);
+}
+
+void Sema::inferGslOwnerPointerAttribute(CXXRecordDecl *Record) {
+  static llvm::StringSet<> StdOwners{
+      "any",
+      "array",
+      "basic_regex",
+      "basic_string",
+      "deque",
+      "forward_list",
+      "vector",
+      "list",
+      "map",
+      "multiset",
+      "multimap",
+      "optional",
+      "priority_queue",
+      "queue",
+      "set",
+      "stack",
+      "unique_ptr",
+      "unordered_set",
+      "unordered_map",
+      "unordered_multiset",
+      "unordered_multimap",
+      "variant",
+  };
+  static llvm::StringSet<> StdPointers{
+      "basic_string_view",
+      "reference_wrapper",
+      "regex_iterator",
+  };
+
+  if (!Record->getIdentifier())
+    return;
+
+  // Handle classes that directly appear in std namespace.
+  if (Record->isInStdNamespace()) {
+    CXXRecordDecl *Canonical = Record->getCanonicalDecl();
+    if (Canonical->hasAttr<OwnerAttr>() || Canonical->hasAttr<PointerAttr>())
+      return;
+
+    if (StdOwners.count(Record->getName()))
+      addGslOwnerPointerAttributeIfNotExisting<OwnerAttr>(Context, Record);
+    else if (StdPointers.count(Record->getName()))
+      addGslOwnerPointerAttributeIfNotExisting<PointerAttr>(Context, Record);
+
+    return;
+  }
+
+  // Handle nested classes that could be a gsl::Pointer.
+  inferGslPointerAttribute(Record, Record);
+}
+
 void Sema::ActOnPragmaOptionsAlign(PragmaOptionsAlignKind Kind,
                                    SourceLocation PragmaLoc) {
   PragmaMsStackAction Action = Sema::PSK_Reset;

clang/lib/Sema/SemaDecl.cpp

@@ -5858,6 +5858,8 @@ Sema::ActOnTypedefNameDecl(Scope *S, DeclContext *DC, TypedefNameDecl *NewTD,
   if (!Previous.empty()) {
     Redeclaration = true;
     MergeTypedefNameDecl(S, NewTD, Previous);
+  } else {
+    inferGslPointerAttribute(NewTD);
   }
 
   if (ShadowedDecl && !Redeclaration)
@@ -15131,6 +15133,9 @@ CreateNewDecl:
   if (PrevDecl)
     mergeDeclAttributes(New, PrevDecl);
 
+  if (auto *CXXRD = dyn_cast<CXXRecordDecl>(New))
+    inferGslOwnerPointerAttribute(CXXRD);
+
   // If there's a #pragma GCC visibility in scope, set the visibility of this
   // record.
   AddPushedVisibilityAttribute(New);

clang/lib/Sema/SemaTemplate.cpp

@@ -1690,6 +1690,7 @@ DeclResult Sema::CheckClassTemplate(
     mergeDeclAttributes(NewClass, PrevClassTemplate->getTemplatedDecl());
 
   AddPushedVisibilityAttribute(NewClass);
+  inferGslOwnerPointerAttribute(NewClass);
 
   if (TUK != TUK_Friend) {
     // Per C++ [basic.scope.temp]p2, skip the template parameter scopes.

clang/test/SemaCXX/attr-gsl-owner-pointer-std.cpp

@@ -0,0 +1,129 @@
+// RUN: %clang_cc1 -ast-dump %s | \
+// RUN: FileCheck --implicit-check-not OwnerAttr --implicit-check-not PointerAttr %s
+
+// Test attribute inference for types in the standard library.
+namespace std {
+// Attributes are inferred for a (complete) class.
+class any {
+  // CHECK: CXXRecordDecl {{.*}} any
+  // CHECK: OwnerAttr {{.*}}
+};
+
+// Attributes are inferred for instantiations of a complete template.
+template <typename T>
+class vector {
+public:
+  class iterator {};
+  // CHECK: ClassTemplateDecl {{.*}} vector
+  // CHECK: OwnerAttr {{.*}}
+  // CHECK: CXXRecordDecl {{.*}} iterator
+  // CHECK: PointerAttr {{.*}}
+  // CHECK: ClassTemplateSpecializationDecl {{.*}} vector
+  // CHECK: TemplateArgument type 'int'
+  // CHECK: OwnerAttr
+  // CHECK: CXXRecordDecl {{.*}} iterator
+  // CHECK: PointerAttr {{.*}}
+};
+static_assert(sizeof(vector<int>), "");           // Force instantiation.
+static_assert(sizeof(vector<int>::iterator), ""); // Force instantiation.
+
+// If std::container::iterator is a using declaration, attributes are inferred
+// for the underlying class.
+template <typename T>
+class __set_iterator {};
+// CHECK: ClassTemplateDecl {{.*}} __set_iterator
+// CHECK: PointerAttr
+// CHECK: ClassTemplateSpecializationDecl {{.*}} __set_iterator
+// CHECK: TemplateArgument type 'int'
+// CHECK: PointerAttr
+
+template <typename T>
+class set {
+  // CHECK: ClassTemplateDecl {{.*}} set
+  // CHECK: OwnerAttr {{.*}}
+  // CHECK: ClassTemplateSpecializationDecl {{.*}} set
+  // CHECK: OwnerAttr {{.*}}
+public:
+  using iterator = __set_iterator<T>;
+};
+static_assert(sizeof(set<int>::iterator), ""); // Force instantiation.
+
+// If std::container::iterator is a typedef, attributes are inferred for the
+// underlying class.
+template <typename T>
+class __map_iterator {};
+// CHECK: ClassTemplateDecl {{.*}} __map_iterator
+// CHECK: PointerAttr
+// CHECK: ClassTemplateSpecializationDecl {{.*}} __map_iterator
+// CHECK: TemplateArgument type 'int'
+// CHECK: PointerAttr
+
+template <typename T>
+class map {
+  // CHECK: ClassTemplateDecl {{.*}} map
+  // CHECK: OwnerAttr {{.*}}
+  // CHECK: ClassTemplateSpecializationDecl {{.*}} map
+  // CHECK: OwnerAttr {{.*}}
+public:
+  typedef __map_iterator<T> iterator;
+};
+static_assert(sizeof(map<int>::iterator), ""); // Force instantiation.
+
+// Inline namespaces are ignored when checking if
+// the class lives in the std namespace.
+inline namespace inlinens {
+template <typename T>
+class __unordered_map_iterator {};
+// CHECK: ClassTemplateDecl {{.*}} __unordered_map_iterator
+// CHECK: PointerAttr
+// CHECK: ClassTemplateSpecializationDecl {{.*}} __unordered_map_iterator
+// CHECK: TemplateArgument type 'int'
+// CHECK: PointerAttr
+
+template <typename T>
+class unordered_map {
+  // CHECK: ClassTemplateDecl {{.*}} unordered_map
+  // CHECK: OwnerAttr {{.*}}
+  // CHECK: ClassTemplateSpecializationDecl {{.*}} unordered_map
+  // CHECK: OwnerAttr {{.*}}
+public:
+  typedef __unordered_map_iterator<T> iterator;
+};
+static_assert(sizeof(unordered_map<int>::iterator), ""); // Force instantiation.
+} // namespace inlinens
+
+// std::list has an implicit gsl::Owner attribute,
+// but explicit attributes take precedence.
+template <typename T>
+class [[gsl::Pointer]] list{};
+// CHECK: ClassTemplateDecl {{.*}} list
+// CHECK: PointerAttr {{.*}}
+// CHECK: ClassTemplateSpecializationDecl {{.*}} list
+// CHECK: PointerAttr {{.*}}
+
+static_assert(sizeof(list<int>), ""); // Force instantiation.
+
+// Forward declared template (Owner).
+template <
+    class CharT,
+    class Traits>
+class basic_regex;
+// CHECK: ClassTemplateDecl {{.*}} basic_regex
+// CHECK: OwnerAttr {{.*}}
+
+// Forward declared template (Pointer).
+template <class T>
+class reference_wrapper;
+// CHECK: ClassTemplateDecl {{.*}} reference_wrapper
+// CHECK: PointerAttr {{.*}}
+
+class some_unknown_type;
+// CHECK: CXXRecordDecl {{.*}} some_unknown_type
+
+} // namespace std
+
+namespace user {
+// If a class is not in the std namespace, we don't infer the attributes.
+class any {
+};
+} // namespace user