Lex: Use the correct types for MS integer suffixes

We didn't properly implement support for the sized integer suffixes.
Suffixes like i16 were essentially ignored instead of mapping them to
the appropriately sized integer type.

This fixes PR20008.

Differential Revision: http://reviews.llvm.org/D4132

llvm-svn: 211426

6 files changed, 40 insertions, 29 deletions

diff --git a/clang/include/clang/Lex/LiteralSupport.h b/clang/include/clang/Lex/LiteralSupport.h
index 2eb7751f3c3..acfd174b63b 100644
--- a/clang/include/clang/Lex/LiteralSupport.h
+++ b/clang/include/clang/Lex/LiteralSupport.h
@@ -63,7 +63,7 @@ public:
   bool isLongLong;
   bool isFloat;       // 1.0f
   bool isImaginary;   // 1.0i
-  bool isMicrosoftInteger;  // Microsoft suffix extension i8, i16, i32, or i64.
+  uint8_t MicrosoftInteger;  // Microsoft suffix extension i8, i16, i32, or i64.
 
   bool isIntegerLiteral() const {
     return !saw_period && !saw_exponent;
diff --git a/clang/lib/AST/StmtPrinter.cpp b/clang/lib/AST/StmtPrinter.cpp
index 0f4fd552461..5c92521c7af 100644
--- a/clang/lib/AST/StmtPrinter.cpp
+++ b/clang/lib/AST/StmtPrinter.cpp
@@ -947,8 +947,10 @@ void StmtPrinter::VisitIntegerLiteral(IntegerLiteral *Node) {
   // FIXME: The Short and UShort cases are to handle cases where a short
   // integeral literal is formed during template instantiation.  They should
   // be removed when template instantiation no longer needs integer literals.
-  case BuiltinType::Short:
-  case BuiltinType::UShort:
+  case BuiltinType::SChar:     OS << "i8"; break;
+  case BuiltinType::UChar:     OS << "Ui8"; break;
+  case BuiltinType::Short:     OS << "i16"; break;
+  case BuiltinType::UShort:    OS << "Ui16"; break;
   case BuiltinType::Int:       break; // no suffix.
   case BuiltinType::UInt:      OS << 'U'; break;
   case BuiltinType::Long:      OS << 'L'; break;
diff --git a/clang/lib/Lex/LiteralSupport.cpp b/clang/lib/Lex/LiteralSupport.cpp
index 0103450ccab..c55054be306 100644
--- a/clang/lib/Lex/LiteralSupport.cpp
+++ b/clang/lib/Lex/LiteralSupport.cpp
@@ -522,7 +522,7 @@ NumericLiteralParser::NumericLiteralParser(StringRef TokSpelling,
   isLongLong = false;
   isFloat = false;
   isImaginary = false;
-  isMicrosoftInteger = false;
+  MicrosoftInteger = 0;
   hadError = false;
 
   if (*s == '0') { // parse radix
@@ -606,7 +606,8 @@ NumericLiteralParser::NumericLiteralParser(StringRef TokSpelling,
     case 'i':
     case 'I':
       if (PP.getLangOpts().MicrosoftExt) {
-        if (isLong || isLongLong) break;
+        if (isLong || isLongLong || MicrosoftInteger)
+          break;
 
         // Allow i8, i16, i32, i64, and i128.
         if (s + 1 != ThisTokEnd) {
@@ -614,20 +615,20 @@ NumericLiteralParser::NumericLiteralParser(StringRef TokSpelling,
             case '8':
               if (isFPConstant) break;
               s += 2; // i8 suffix
-              isMicrosoftInteger = true;
+              MicrosoftInteger = 8;
               break;
             case '1':
               if (isFPConstant) break;
               if (s + 2 == ThisTokEnd) break;
               if (s[2] == '6') {
                 s += 3; // i16 suffix
-                isMicrosoftInteger = true;
+                MicrosoftInteger = 16;
               }
               else if (s[2] == '2') {
                 if (s + 3 == ThisTokEnd) break;
                 if (s[3] == '8') {
                   s += 4; // i128 suffix
-                  isMicrosoftInteger = true;
+                  MicrosoftInteger = 128;
                 }
               }
               break;
@@ -636,8 +637,7 @@ NumericLiteralParser::NumericLiteralParser(StringRef TokSpelling,
               if (s + 2 == ThisTokEnd) break;
               if (s[2] == '2') {
                 s += 3; // i32 suffix
-                isLong = true;
-                isMicrosoftInteger = true;
+                MicrosoftInteger = 32;
               }
               break;
             case '6':
@@ -645,14 +645,13 @@ NumericLiteralParser::NumericLiteralParser(StringRef TokSpelling,
               if (s + 2 == ThisTokEnd) break;
               if (s[2] == '4') {
                 s += 3; // i64 suffix
-                isLongLong = true;
-                isMicrosoftInteger = true;
+                MicrosoftInteger = 64;
               }
               break;
             default:
               break;
           }
-          if (isMicrosoftInteger)
+          if (MicrosoftInteger)
             break;
         }
       }
@@ -682,7 +681,7 @@ NumericLiteralParser::NumericLiteralParser(StringRef TokSpelling,
       isLongLong = false;
       isFloat = false;
       isImaginary = false;
-      isMicrosoftInteger = false;
+      MicrosoftInteger = 0;
 
       saw_ud_suffix = true;
       return;
diff --git a/clang/lib/Sema/SemaExpr.cpp b/clang/lib/Sema/SemaExpr.cpp
index 2654c390ccf..69cc8dc2fd1 100644
--- a/clang/lib/Sema/SemaExpr.cpp
+++ b/clang/lib/Sema/SemaExpr.cpp
@@ -3190,7 +3190,7 @@ ExprResult Sema::ActOnNumericConstant(const Token &Tok, Scope *UDLScope) {
     // may be wider than [u]intmax_t.
     // FIXME: Actually, they don't. We seem to have accidentally invented the
     //        i128 suffix.
-    if (Literal.isMicrosoftInteger && MaxWidth < 128 &&
+    if (Literal.MicrosoftInteger && MaxWidth < 128 &&
         Context.getTargetInfo().hasInt128Type())
       MaxWidth = 128;
     llvm::APInt ResultVal(MaxWidth, 0);
@@ -3211,7 +3211,18 @@ ExprResult Sema::ActOnNumericConstant(const Token &Tok, Scope *UDLScope) {
 
       // Check from smallest to largest, picking the smallest type we can.
       unsigned Width = 0;
-      if (!Literal.isLong && !Literal.isLongLong) {
+
+      // Microsoft specific integer suffixes are explicitly sized.
+      if (Literal.MicrosoftInteger) {
+        Width = Literal.MicrosoftInteger;
+        if (Width < 128)
+          Ty = Context.getIntTypeForBitwidth(Width,
+                                             /*Signed=*/!Literal.isUnsigned);
+        else
+          Ty = Literal.isUnsigned ? Context.UnsignedInt128Ty : Context.Int128Ty;
+      }
+
+      if (Ty.isNull() && !Literal.isLong && !Literal.isLongLong) {
         // Are int/unsigned possibilities?
         unsigned IntSize = Context.getTargetInfo().getIntWidth();
 
@@ -3258,17 +3269,6 @@ ExprResult Sema::ActOnNumericConstant(const Token &Tok, Scope *UDLScope) {
           Width = LongLongSize;
         }
       }
-        
-      // If it doesn't fit in unsigned long long, and we're using Microsoft
-      // extensions, then its a 128-bit integer literal.
-      if (Ty.isNull() && Literal.isMicrosoftInteger &&
-          Context.getTargetInfo().hasInt128Type()) {
-        if (Literal.isUnsigned)
-          Ty = Context.UnsignedInt128Ty;
-        else
-          Ty = Context.Int128Ty;
-        Width = 128;
-      }
 
       // If we still couldn't decide a type, we probably have something that
       // does not fit in a signed long long, but has no U suffix.
diff --git a/clang/test/SemaCXX/ms_integer_suffix.cpp b/clang/test/SemaCXX/ms_integer_suffix.cpp
new file mode 100644
index 00000000000..20cd5a6f2cf
--- /dev/null
+++ b/clang/test/SemaCXX/ms_integer_suffix.cpp
@@ -0,0 +1,8 @@
+// RUN: %clang_cc1 -std=c++11 -fsyntax-only -fms-compatibility -verify %s
+// expected-no-diagnostics
+
+static_assert(sizeof(0i8  ) == sizeof(__INT8_TYPE__ ), "");
+static_assert(sizeof(0i16 ) == sizeof(__INT16_TYPE__), "");
+static_assert(sizeof(0i32 ) == sizeof(__INT32_TYPE__), "");
+static_assert(sizeof(0i64 ) == sizeof(__INT64_TYPE__), "");
+static_assert(sizeof(0i128) >  sizeof(__INT64_TYPE__), "");
diff --git a/clang/unittests/AST/StmtPrinterTest.cpp b/clang/unittests/AST/StmtPrinterTest.cpp
index c75cbdefbf2..541fb3df1d9 100644
--- a/clang/unittests/AST/StmtPrinterTest.cpp
+++ b/clang/unittests/AST/StmtPrinterTest.cpp
@@ -134,6 +134,8 @@ PrintedStmtCXX11Matches(StringRef Code, const StatementMatcher &NodeMatch,
                                               StringRef ContainingFunction,
                                               StringRef ExpectedPrinted) {
   std::vector<std::string> Args;
+  Args.push_back("-target");
+  Args.push_back("i686-pc-win32");
   Args.push_back("-std=c++98");
   Args.push_back("-fms-extensions");
   Args.push_back("-Wno-unused-value");
@@ -169,9 +171,9 @@ TEST(StmtPrinter, TestMSIntegerLiteral) {
     "  1i64, -1i64, 1ui64;"
     "}",
     "A",
+    "1i8 , -1i8 , 1Ui8 , "
+    "1i16 , -1i16 , 1Ui16 , "
     "1 , -1 , 1U , "
-    "1 , -1 , 1U , "
-    "1L , -1L , 1UL , "
     "1LL , -1LL , 1ULL"));
     // Should be: with semicolon
 }