[LLDB] Add array subscription and integer parsing to DIL (#141102)

Reapply #138551 with an xfailed test on Windows

lldb/docs/dil-expr-lang.ebnf

@@ -10,16 +10,17 @@ unary_expression = postfix_expression
 
 unary_operator = "*" | "&" ;
 
-postfix_expresson = primary_expression
-                  | postfix_expression "." id_expression
-                  | postfix_expression "->" id_expression ;
+postfix_expression = primary_expression
+                   | postfix_expression "[" integer_literal "]"
+                   | postfix_expression "." id_expression
+                   | postfix_expression "->" id_expression ;
 
 primary_expression = id_expression
-                   | "(" expression ")"  ;
+                   | "(" expression ")" ;
 
 id_expression = unqualified_id
               | qualified_id
-	      | register ;
+              | register ;
 
 unqualified_id = identifier ;
 
@@ -28,6 +29,8 @@ qualified_id = ["::"] [nested_name_specifier] unqualified_id
 
 identifier = ? C99 Identifier ? ;
 
+integer_literal = ? Integer constant: hexademical, decimal, octal, binary ? ;
+
 register = "$" ? Register name ? ;
 
 nested_name_specifier = type_name "::"

lldb/include/lldb/ValueObject/DILAST.h

@@ -18,6 +18,7 @@ namespace lldb_private::dil {
 
 /// The various types DIL AST nodes (used by the DIL parser).
 enum class NodeKind {
+  eArraySubscriptNode,
   eErrorNode,
   eIdentifierNode,
   eMemberOfNode,
@@ -120,8 +121,8 @@ public:
 
   llvm::Expected<lldb::ValueObjectSP> Accept(Visitor *v) const override;
 
-  UnaryOpKind kind() const { return m_kind; }
-  ASTNode *operand() const { return m_operand.get(); }
+  UnaryOpKind GetKind() const { return m_kind; }
+  ASTNode *GetOperand() const { return m_operand.get(); }
 
   static bool classof(const ASTNode *node) {
     return node->GetKind() == NodeKind::eUnaryOpNode;
@@ -132,6 +133,26 @@ private:
   ASTNodeUP m_operand;
 };
 
+class ArraySubscriptNode : public ASTNode {
+public:
+  ArraySubscriptNode(uint32_t location, ASTNodeUP base, int64_t index)
+      : ASTNode(location, NodeKind::eArraySubscriptNode),
+        m_base(std::move(base)), m_index(index) {}
+
+  llvm::Expected<lldb::ValueObjectSP> Accept(Visitor *v) const override;
+
+  ASTNode *GetBase() const { return m_base.get(); }
+  int64_t GetIndex() const { return m_index; }
+
+  static bool classof(const ASTNode *node) {
+    return node->GetKind() == NodeKind::eArraySubscriptNode;
+  }
+
+private:
+  ASTNodeUP m_base;
+  int64_t m_index;
+};
+
 /// This class contains one Visit method for each specialized type of
 /// DIL AST node. The Visit methods are used to dispatch a DIL AST node to
 /// the correct function in the DIL expression evaluator for evaluating that
@@ -145,6 +166,8 @@ public:
   Visit(const MemberOfNode *node) = 0;
   virtual llvm::Expected<lldb::ValueObjectSP>
   Visit(const UnaryOpNode *node) = 0;
+  virtual llvm::Expected<lldb::ValueObjectSP>
+  Visit(const ArraySubscriptNode *node) = 0;
 };
 
 } // namespace lldb_private::dil

lldb/include/lldb/ValueObject/DILEval.h

@@ -52,6 +52,8 @@ private:
   Visit(const IdentifierNode *node) override;
   llvm::Expected<lldb::ValueObjectSP> Visit(const MemberOfNode *node) override;
   llvm::Expected<lldb::ValueObjectSP> Visit(const UnaryOpNode *node) override;
+  llvm::Expected<lldb::ValueObjectSP>
+  Visit(const ArraySubscriptNode *node) override;
 
   // Used by the interpreter to create objects, perform casts, etc.
   lldb::TargetSP m_target;

lldb/include/lldb/ValueObject/DILLexer.h

@@ -30,8 +30,11 @@ public:
     eof,
     identifier,
     l_paren,
+    l_square,
+    numeric_constant,
     period,
     r_paren,
+    r_square,
     star,
   };
 

lldb/include/lldb/ValueObject/DILParser.h

@@ -95,6 +95,7 @@ private:
 
   std::string ParseIdExpression();
   std::string ParseUnqualifiedId();
+  std::optional<int64_t> ParseIntegerConstant();
 
   void BailOut(const std::string &error, uint32_t loc, uint16_t err_len);
 

lldb/source/ValueObject/DILAST.cpp

@@ -27,4 +27,9 @@ llvm::Expected<lldb::ValueObjectSP> UnaryOpNode::Accept(Visitor *v) const {
   return v->Visit(this);
 }
 
+llvm::Expected<lldb::ValueObjectSP>
+ArraySubscriptNode::Accept(Visitor *v) const {
+  return v->Visit(this);
+}
+
 } // namespace lldb_private::dil

lldb/source/ValueObject/DILEval.cpp

@@ -240,13 +240,13 @@ Interpreter::Visit(const IdentifierNode *node) {
 llvm::Expected<lldb::ValueObjectSP>
 Interpreter::Visit(const UnaryOpNode *node) {
   Status error;
-  auto rhs_or_err = Evaluate(node->operand());
+  auto rhs_or_err = Evaluate(node->GetOperand());
   if (!rhs_or_err)
     return rhs_or_err;
 
   lldb::ValueObjectSP rhs = *rhs_or_err;
 
-  switch (node->kind()) {
+  switch (node->GetKind()) {
   case UnaryOpKind::Deref: {
     lldb::ValueObjectSP dynamic_rhs = rhs->GetDynamicValue(m_use_dynamic);
     if (dynamic_rhs)
@@ -383,4 +383,51 @@ Interpreter::Visit(const MemberOfNode *node) {
       m_expr, errMsg, node->GetLocation(), node->GetFieldName().size());
 }
 
+llvm::Expected<lldb::ValueObjectSP>
+Interpreter::Visit(const ArraySubscriptNode *node) {
+  auto lhs_or_err = Evaluate(node->GetBase());
+  if (!lhs_or_err)
+    return lhs_or_err;
+  lldb::ValueObjectSP base = *lhs_or_err;
+
+  // Check to see if 'base' has a synthetic value; if so, try using that.
+  uint64_t child_idx = node->GetIndex();
+  if (lldb::ValueObjectSP synthetic = base->GetSyntheticValue()) {
+    llvm::Expected<uint32_t> num_children =
+        synthetic->GetNumChildren(child_idx + 1);
+    if (!num_children)
+      return llvm::make_error<DILDiagnosticError>(
+          m_expr, toString(num_children.takeError()), node->GetLocation());
+    if (child_idx >= *num_children) {
+      std::string message = llvm::formatv(
+          "array index {0} is not valid for \"({1}) {2}\"", child_idx,
+          base->GetTypeName().AsCString("<invalid type>"),
+          base->GetName().AsCString());
+      return llvm::make_error<DILDiagnosticError>(m_expr, message,
+                                                  node->GetLocation());
+    }
+    if (lldb::ValueObjectSP child_valobj_sp =
+            synthetic->GetChildAtIndex(child_idx))
+      return child_valobj_sp;
+  }
+
+  auto base_type = base->GetCompilerType().GetNonReferenceType();
+  if (!base_type.IsPointerType() && !base_type.IsArrayType())
+    return llvm::make_error<DILDiagnosticError>(
+        m_expr, "subscripted value is not an array or pointer",
+        node->GetLocation());
+  if (base_type.IsPointerToVoid())
+    return llvm::make_error<DILDiagnosticError>(
+        m_expr, "subscript of pointer to incomplete type 'void'",
+        node->GetLocation());
+
+  if (base_type.IsArrayType()) {
+    if (lldb::ValueObjectSP child_valobj_sp = base->GetChildAtIndex(child_idx))
+      return child_valobj_sp;
+  }
+
+  int64_t signed_child_idx = node->GetIndex();
+  return base->GetSyntheticArrayMember(signed_child_idx, true);
+}
+
 } // namespace lldb_private::dil

lldb/source/ValueObject/DILLexer.cpp

@@ -13,6 +13,7 @@
 
 #include "lldb/ValueObject/DILLexer.h"
 #include "lldb/Utility/Status.h"
+#include "lldb/ValueObject/DILParser.h"
 #include "llvm/ADT/StringSwitch.h"
 
 namespace lldb_private::dil {
@@ -31,10 +32,16 @@ llvm::StringRef Token::GetTokenName(Kind kind) {
     return "identifier";
   case Kind::l_paren:
     return "l_paren";
+  case Kind::l_square:
+    return "l_square";
+  case Kind::numeric_constant:
+    return "numeric_constant";
   case Kind::period:
     return "period";
   case Kind::r_paren:
     return "r_paren";
+  case Kind::r_square:
+    return "r_square";
   case Token::star:
     return "star";
   }
@@ -61,6 +68,18 @@ static std::optional<llvm::StringRef> IsWord(llvm::StringRef expr,
   return candidate;
 }
 
+static bool IsNumberBodyChar(char ch) { return IsDigit(ch) || IsLetter(ch); }
+
+static std::optional<llvm::StringRef> IsNumber(llvm::StringRef expr,
+                                               llvm::StringRef &remainder) {
+  if (IsDigit(remainder[0])) {
+    llvm::StringRef number = remainder.take_while(IsNumberBodyChar);
+    remainder = remainder.drop_front(number.size());
+    return number;
+  }
+  return std::nullopt;
+}
+
 llvm::Expected<DILLexer> DILLexer::Create(llvm::StringRef expr) {
   std::vector<Token> tokens;
   llvm::StringRef remainder = expr;
@@ -85,14 +104,17 @@ llvm::Expected<Token> DILLexer::Lex(llvm::StringRef expr,
     return Token(Token::eof, "", (uint32_t)expr.size());
 
   uint32_t position = cur_pos - expr.begin();
+  std::optional<llvm::StringRef> maybe_number = IsNumber(expr, remainder);
+  if (maybe_number)
+    return Token(Token::numeric_constant, maybe_number->str(), position);
   std::optional<llvm::StringRef> maybe_word = IsWord(expr, remainder);
   if (maybe_word)
     return Token(Token::identifier, maybe_word->str(), position);
 
   constexpr std::pair<Token::Kind, const char *> operators[] = {
-      {Token::amp, "&"},     {Token::arrow, "->"}, {Token::coloncolon, "::"},
-      {Token::l_paren, "("}, {Token::period, "."}, {Token::r_paren, ")"},
-      {Token::star, "*"},
+      {Token::amp, "&"},     {Token::arrow, "->"},   {Token::coloncolon, "::"},
+      {Token::l_paren, "("}, {Token::l_square, "["}, {Token::period, "."},
+      {Token::r_paren, ")"}, {Token::r_square, "]"}, {Token::star, "*"},
   };
   for (auto [kind, str] : operators) {
     if (remainder.consume_front(str))
@@ -100,7 +122,8 @@ llvm::Expected<Token> DILLexer::Lex(llvm::StringRef expr,
   }
 
   // Unrecognized character(s) in string; unable to lex it.
-  return llvm::createStringError("Unable to lex input string");
+  return llvm::make_error<DILDiagnosticError>(expr, "unrecognized token",
+                                              position);
 }
 
 } // namespace lldb_private::dil

lldb/source/ValueObject/DILParser.cpp

@@ -119,20 +119,46 @@ ASTNodeUP DILParser::ParseUnaryExpression() {
 //
 //  postfix_expression:
 //    primary_expression
+//    postfix_expression "[" integer_literal "]"
 //    postfix_expression "." id_expression
 //    postfix_expression "->" id_expression
 //
 ASTNodeUP DILParser::ParsePostfixExpression() {
   ASTNodeUP lhs = ParsePrimaryExpression();
-  while (CurToken().IsOneOf({Token::period, Token::arrow})) {
+  while (CurToken().IsOneOf({Token::l_square, Token::period, Token::arrow})) {
+    uint32_t loc = CurToken().GetLocation();
     Token token = CurToken();
-    m_dil_lexer.Advance();
-    Token member_token = CurToken();
-    std::string member_id = ParseIdExpression();
-    lhs = std::make_unique<MemberOfNode>(
-        member_token.GetLocation(), std::move(lhs),
-        token.GetKind() == Token::arrow, member_id);
+    switch (token.GetKind()) {
+    case Token::l_square: {
+      m_dil_lexer.Advance();
+      std::optional<int64_t> rhs = ParseIntegerConstant();
+      if (!rhs) {
+        BailOut(
+            llvm::formatv("failed to parse integer constant: {0}", CurToken()),
+            CurToken().GetLocation(), CurToken().GetSpelling().length());
+        return std::make_unique<ErrorNode>();
+      }
+      Expect(Token::r_square);
+      m_dil_lexer.Advance();
+      lhs = std::make_unique<ArraySubscriptNode>(loc, std::move(lhs),
+                                                 std::move(*rhs));
+      break;
+    }
+    case Token::period:
+    case Token::arrow: {
+      m_dil_lexer.Advance();
+      Token member_token = CurToken();
+      std::string member_id = ParseIdExpression();
+      lhs = std::make_unique<MemberOfNode>(
+          member_token.GetLocation(), std::move(lhs),
+          token.GetKind() == Token::arrow, member_id);
+      break;
+    }
+    default:
+      llvm_unreachable("invalid token");
+    }
   }
+
   return lhs;
 }
 
@@ -302,6 +328,23 @@ void DILParser::BailOut(const std::string &error, uint32_t loc,
   m_dil_lexer.ResetTokenIdx(m_dil_lexer.NumLexedTokens() - 1);
 }
 
+// Parse a integer_literal.
+//
+//  integer_literal:
+//    ? Integer constant ?
+//
+std::optional<int64_t> DILParser::ParseIntegerConstant() {
+  auto spelling = CurToken().GetSpelling();
+  llvm::StringRef spelling_ref = spelling;
+  int64_t raw_value;
+  if (!spelling_ref.getAsInteger<int64_t>(0, raw_value)) {
+    m_dil_lexer.Advance();
+    return raw_value;
+  }
+
+  return std::nullopt;
+}
+
 void DILParser::Expect(Token::Kind kind) {
   if (CurToken().IsNot(kind)) {
     BailOut(llvm::formatv("expected {0}, got: {1}", kind, CurToken()),

lldb/test/API/commands/frame/var-dil/basics/ArraySubscript/Makefile

@@ -0,0 +1,3 @@
+CXX_SOURCES := main.cpp
+
+include Makefile.rules

lldb/test/API/commands/frame/var-dil/basics/ArraySubscript/TestFrameVarDILArraySubscript.py

@@ -0,0 +1,119 @@
+"""
+Test DIL array subscript.
+"""
+
+import lldb
+from lldbsuite.test.lldbtest import *
+from lldbsuite.test.decorators import *
+from lldbsuite.test import lldbutil
+
+
+class TestFrameVarDILArraySubscript(TestBase):
+    NO_DEBUG_INFO_TESTCASE = True
+
+    def expect_var_path(self, expr, compare_to_framevar=False, value=None, type=None):
+        value_dil = super().expect_var_path(expr, value=value, type=type)
+        if compare_to_framevar:
+            self.runCmd("settings set target.experimental.use-DIL false")
+            value_frv = super().expect_var_path(expr, value=value, type=type)
+            self.runCmd("settings set target.experimental.use-DIL true")
+            self.assertEqual(value_dil.GetValue(), value_frv.GetValue())
+
+    def test_subscript(self):
+        self.build()
+        lldbutil.run_to_source_breakpoint(
+            self, "Set a breakpoint here", lldb.SBFileSpec("main.cpp")
+        )
+
+        self.runCmd("settings set target.experimental.use-DIL true")
+
+        # Test int[] and int*
+        self.expect_var_path("int_arr[0]", True, value="1")
+        self.expect_var_path("int_ptr[1]", True, value="2")
+        self.expect("frame var 'int_arr[enum_one]'", error=True)
+
+        # Test when base and index are references.
+        self.expect_var_path("int_arr[0]", True, value="1")
+        self.expect("frame var 'int_arr[idx_1_ref]'", error=True)
+        self.expect("frame var 'int_arr[enum_ref]'", error=True)
+        self.expect_var_path("int_arr_ref[0]", value="1")
+        self.expect("frame var 'int_arr_ref[idx_1_ref]'", error=True)
+        self.expect("frame var 'int_arr_ref[enum_ref]'", error=True)
+
+        # Test when base and index are typedefs.
+        self.expect_var_path("td_int_arr[0]", True, value="1")
+        self.expect("frame var 'td_int_arr[td_int_idx_1]'", error=True)
+        self.expect("frame var 'td_int_arr[td_td_int_idx_2]'", error=True)
+        self.expect_var_path("td_int_ptr[0]", True, value="1")
+        self.expect("frame var 'td_int_ptr[td_int_idx_1]'", error=True)
+        self.expect("frame var 'td_int_ptr[td_td_int_idx_2]'", error=True)
+
+        # Both typedefs and refs
+        self.expect("frame var 'td_int_arr_ref[td_int_idx_1_ref]'", error=True)
+
+        # Test for index out of bounds.
+        self.expect_var_path("int_arr[42]", True, type="int")
+        self.expect_var_path("int_arr[100]", True, type="int")
+
+        # Test address-of of the subscripted value.
+        self.expect_var_path("*(&int_arr[1])", value="2")
+
+        # Test for negative index.
+        self.expect(
+            "frame var 'int_arr[-1]'",
+            error=True,
+            substrs=["unrecognized token"],
+        )
+
+        # Test for floating point index
+        self.expect(
+            "frame var 'int_arr[1.0]'",
+            error=True,
+            substrs=["expected 'r_square', got: <'.'"],
+        )
+
+        # Base should be a "pointer to T" and index should be of an integral type.
+        self.expect(
+            "frame var 'idx_1[0]'",
+            error=True,
+            substrs=["subscripted value is not an array or pointer"],
+        )
+        self.expect(
+            "frame var 'idx_1_ref[0]'",
+            error=True,
+            substrs=["subscripted value is not an array or pointer"],
+        )
+        self.expect(
+            "frame var 'int_arr[int_ptr]'",
+            error=True,
+            substrs=["failed to parse integer constant"],
+        )
+        self.expect(
+            "frame var '1[2]'",
+            error=True,
+            substrs=["Unexpected token"],
+        )
+
+        # Base should not be a pointer to void
+        self.expect(
+            "frame var 'p_void[0]'",
+            error=True,
+            substrs=["subscript of pointer to incomplete type 'void'"],
+        )
+
+    @expectedFailureAll(oslist=["windows"])
+    def test_subscript_synthetic(self):
+        self.build()
+        lldbutil.run_to_source_breakpoint(
+            self, "Set a breakpoint here", lldb.SBFileSpec("main.cpp")
+        )
+
+        self.runCmd("settings set target.experimental.use-DIL true")
+
+        # Test synthetic value subscription
+        self.expect_var_path("vector[1]", value="2")
+        self.expect(
+            "frame var 'vector[100]'",
+            error=True,
+            substrs=["array index 100 is not valid"],
+        )

lldb/test/API/commands/frame/var-dil/basics/ArraySubscript/main.cpp

@@ -0,0 +1,32 @@
+#include <vector>
+
+int main(int argc, char **argv) {
+  int int_arr[] = {1, 2, 3};
+  int *int_ptr = int_arr;
+  int(&int_arr_ref)[3] = int_arr;
+  void *p_void = (void *)int_arr;
+
+  int idx_1 = 1;
+  const int &idx_1_ref = idx_1;
+
+  typedef int td_int_t;
+  typedef td_int_t td_td_int_t;
+  typedef int *td_int_ptr_t;
+  typedef int &td_int_ref_t;
+
+  td_int_t td_int_idx_1 = 1;
+  td_td_int_t td_td_int_idx_2 = 2;
+
+  td_int_t td_int_arr[3] = {1, 2, 3};
+  td_int_ptr_t td_int_ptr = td_int_arr;
+
+  td_int_ref_t td_int_idx_1_ref = td_int_idx_1;
+  td_int_t(&td_int_arr_ref)[3] = td_int_arr;
+
+  enum Enum { kZero, kOne } enum_one = kOne;
+  Enum &enum_ref = enum_one;
+
+  std::vector<int> vector = {1, 2, 3};
+
+  return 0; // Set a breakpoint here
+}

lldb/unittests/ValueObject/DILLexerTests.cpp

@@ -121,11 +121,11 @@ TEST(DILLexerTests, IdentifiersTest) {
       "a_b",       "this", "self", "a", "MyName", "namespace"};
 
   // The lexer can lex these strings, but they should not be identifiers.
-  std::vector<std::string> invalid_identifiers = {"", "::", "(", ")"};
+  std::vector<std::string> invalid_identifiers = {"", "::", "(", ")", "0abc"};
 
   // The lexer is expected to fail attempting to lex these strings (it cannot
   // create valid tokens out of them).
-  std::vector<std::string> invalid_tok_strings = {"234", "2a", "2", "1MyName"};
+  std::vector<std::string> invalid_tok_strings = {"#include", "a@a"};
 
   // Verify that all of the valid identifiers come out as identifier tokens.
   for (auto &str : valid_identifiers) {
@@ -150,7 +150,32 @@ TEST(DILLexerTests, IdentifiersTest) {
     DILLexer lexer(*maybe_lexer);
     Token token = lexer.GetCurrentToken();
     EXPECT_TRUE(token.IsNot(Token::identifier));
-    EXPECT_TRUE(token.IsOneOf(
-        {Token::eof, Token::coloncolon, Token::l_paren, Token::r_paren}));
+    EXPECT_TRUE(token.IsOneOf({Token::eof, Token::coloncolon, Token::l_paren,
+                               Token::r_paren, Token::numeric_constant}));
+  }
+}
+
+TEST(DILLexerTests, NumbersTest) {
+  // These strings should lex into number tokens.
+  std::vector<std::string> valid_numbers = {"123", "0x123", "0123", "0b101"};
+
+  // The lexer can lex these strings, but they should not be numbers.
+  std::vector<std::string> invalid_numbers = {"", "x123", "b123"};
+
+  for (auto &str : valid_numbers) {
+    SCOPED_TRACE(str);
+    EXPECT_THAT_EXPECTED(ExtractTokenData(str),
+                         llvm::HasValue(testing::ElementsAre(
+                             testing::Pair(Token::numeric_constant, str))));
+  }
+  // Verify that none of the invalid numbers come out as numeric tokens.
+  for (auto &str : invalid_numbers) {
+    SCOPED_TRACE(str);
+    llvm::Expected<DILLexer> maybe_lexer = DILLexer::Create(str);
+    EXPECT_THAT_EXPECTED(maybe_lexer, llvm::Succeeded());
+    DILLexer lexer(*maybe_lexer);
+    Token token = lexer.GetCurrentToken();
+    EXPECT_TRUE(token.IsNot(Token::numeric_constant));
+    EXPECT_TRUE(token.IsOneOf({Token::eof, Token::identifier}));
   }
 }