[clang] perform semantic checking in constant context

[Tyker]

Summary

Since the addition of __builtin_is_constant_evaluated the result of an expression can change based on whether it is evaluated in constant context. a lot of semantic checking performs evaluations with out specifying context. which can lead to wrong diagnostics.
for example:

constexpr int i0 = (long long)__builtin_is_constant_evaluated() * (1ll << 33); //#1
constexpr int i1 = (long long)!__builtin_is_constant_evaluated() * (1ll << 33); //#2

before the patch, #2 was diagnosed incorrectly and #1 wasn't diagnosed.
after the patch #1 is diagnosed as it should and #2 isn't.

Changes:

• add a flag to some Expr::Evaluate* operation to evaluate in constant context.
• in SemaChecking.cpp calls to Expr::Evaluate* are now done in constant context when they should.
• in SemaChecking.cpp diagnostics for UB are not checked for in constant context because an error will be emitted by the constant evaluator.
• in SemaChecking.cpp diagnostics for construct that cannot appear in constant context are not checked for in constant context.
• in SemaChecking.cpp diagnostics on constant expression are always emitted because constant expression are always evaluated.
• semantic checking for initialization of constexpr variables is now done in constant context.
• adapt test that were depending on warning changes.
• add test.

[rsmith]

The approach we've been taking for this up until now is to use a ConstantExpr node to mark the entry point of a constant context; it looks like that would continue to work here, but we're missing those nodes in some cases? (This is preferable to using a flag because it also -- eventually -- gives us a place to store the evaluated value, which we're going to need for various upcoming features, particularly consteval.)

[Tyker]

i added this bit because when we eventually store the result of evaluation in ConstantExpr we will probably want to trail-allocate the APValue's possible representations separately to limit memory consumption. but if we do this the ConstantExpr node will need to be created after evaluation of the value to allocate the right space and we will need an other way to mark that the expression should be evaluated in constant context.
if we won't store APValue's possible representations separately in trail-allocate space, wrapping the expression in a ConstantExpr is the right solution.

[rsmith]

In D62009#1505263, @Tyker wrote:

i added this bit because when we eventually store the result of evaluation in ConstantExpr we will probably want to trail-allocate the APValue's possible representations separately to limit memory consumption. but if we do this the ConstantExpr node will need to be created after evaluation of the value to allocate the right space and we will need an other way to mark that the expression should be evaluated in constant context.

Even if we want to tail-allocate (which does seem like a nice idea in the long term), we don't need to mark the expression for that: we can pass a flag into the constant evaluator to indicate that we're evaluating an expression with the intent of forming a ConstantExpr. (We already do that for EvaluateAsRValue; we'll need to do the same for other entry points that could be evaluating at the top level in a constant context.)

[Tyker]

I checked how we can propagate the information about constant context through semantic checking.
there are issues with using ConstantExpr to mark expressions as constant for semantic checking:

• #1 multpile Expr::Ignore* operation remove ConstantExpr from the expression.
• #2 Semantic checking Traverses the AST so all methods that only mark the top-level Node will not completely work.
• #3 at short term: adding ConstantExpr modifies the AST structure, so adding it everywhere it is needed for semantic checking will require changing code that depend closely on the AST structure.

Note: the limitation #2 also applies to the bit in ExprBitfield solution in its current form.

propagating constant context to the expression evaluator via a boolean value will be a lot of boilerplate and in my opinion this should be propagated more "automatically".
so I think the best solutions are:

• push a ExpressionEvaluationContext::ConstantEvaluated so Sema will propagate it and propagate from Sema to the expression evaluator via boolean values.
• put all semantic checking function's in a SemaCheckContext class and propagate via this class to the expression evaluator. this class will be usable to propagate Sema and probably other informations.
• keep the bit in ExprBitfield but make all nodes created in ExpressionEvaluationContext::ConstantEvaluated marked for constant evaluation to fixes limitation #2.

[rsmith]

In D62009#1506415, @Tyker wrote:

there are issues with using ConstantExpr to mark expressions as constant for semantic checking:

• #1 multpile Expr::Ignore* operation remove ConstantExpr from the expression.

Ignore* is generally used when walking the syntactic form of an expression, so this is usually appropriate.

• #2 Semantic checking Traverses the AST so all methods that only mark the top-level Node will not completely work.

Semantic checks that care about constant evaluation should not walk over a ConstantExpr node. Generally, we want rather different checking inside a ConstantExpr node than outside (for example, we shouldn't be performing any checks for potential overflow or narrowing or conversions, and instead should be finding out what *actually* happens by evaluating the expression and diagnosing problems in it). So this seems fine too, though there may still be places that need updates to properly handle ConstantExpr.

• #3 at short term: adding ConstantExpr modifies the AST structure, so adding it everywhere it is needed for semantic checking will require changing code that depend closely on the AST structure.

Yes. But that's going to be the case for any approach we take here.

• push a ExpressionEvaluationContext::ConstantEvaluated so Sema will propagate it and propagate from Sema to the expression evaluator via boolean values.
• put all semantic checking function's in a SemaCheckContext class and propagate via this class to the expression evaluator. this class will be usable to propagate Sema and probably other informations.
• keep the bit in ExprBitfield but make all nodes created in ExpressionEvaluationContext::ConstantEvaluated marked for constant evaluation to fixes limitation #2.

We don't always know whether an expression is constant evaluated until after we've parsed it (eg, for a call to a function that might be consteval we need to perform overload resolution before we find out). And this would require changing all the (many) places that create Expr nodes to pass in new information. That sounds unpleasant and expensive, and adds a permanent cost to all future AST changes.

So far we don't appear to need any per-Expr indicator of whether that expression is transitively within a ConstantExpr, so let's not add that.

[Tyker]

I changed the approach in the last revision. the information is now propagated using the expression evaluation context and then via booleans.

[rsmith]

Interesting. I think all of the new warnings in the test cases here are undesirable (they duplicate errors produced by the constant evaluator), but the removed warnings all look like improvements.

Generally, I think our goals should be:

For code patterns that might lead to undefined behavior when executed with certain values:

• If the code appears in a constant context (isConstantEvaluated() or a subexpression of a ConstantExpr), we should diagnose the problem from the constant evaluator if it actually happens and otherwise not diagnose it.
• Otherwise, we should warn on the problem if the code is reachable and otherwise not diagnose it (that is, use DiagRuntimeBehavior).

For code patterns that are suspicious but defined (eg, we think they might do something other than what the programmer intended), we should typically diagnose using Diag, regardless of whether they appear in a constant context.

DiagRuntimeBehavior already deals with much of this for us: if called in a ConstantEvaluated context, it suppresses diagnostics.

One way we could proceed would be to extend what you did for CheckCompletedExpr to the cases where SemaChecking.cpp (and friends) recurse into a ConstantExpr: temporarily create a ConstantEvaluated expression evaluation context for the purpose of the checks, so that we can track that we're within such a construct. But an ExpressionEvaluationContextRecord is not a trivial thing to create and pass around, and carries additional semantic implications on top of the context kind, so perhaps just adding a flag to it to indicate that we're performing semantic analysis inside a ConstantExpr within the context would be better.

clang/lib/Sema/SemaChecking.cpp

4066–4073

For __attribute__((nonnull)), I think the right thing to do is to change the constant expression evaluator to consider passing a null pointer argument for a nonnull-attributed parameter to be non-constant. (It's undefined behavior, so we should not permit it in constant evaluations.) For _Nonnull, I think the situation is less clear -- that case is explicitly not undefined behavior, so it's not clear that treating it as non-constant is the right choice. But as with the other cases here, isConstantEvaluated() does not imply the code is necessarily reachable, so bypassing DiagRuntimeBehavior's check for reachability doesn't seem like the appropriate behavior.

6089

isConstantEvaluated() doesn't imply that the expression is necessarily reachable, so always giving an error here doesn't seem right. Instead, we should (already) have the relevant range checks in the constant evaluator for such builtins that are usable in constant expressions, and should never need to check here (if the code is reached during a constant evaluation, then the evaluation will be non-constant and hence ill-formed).

I think we should just return early from this function if isConstantEvaluated() like you do in checkFortifiedBuiltinMemoryFunction.

6612–6613

If we're in a constant-evaluated context, I think we can and should simply skip all format string checking. There are three cases here:

1. The call with the format attribute is unreached: no diagnostic should be issued.
2. The call with the format attribute is reached but not constant-evaluatable: an error will be produced that we couldn't constant-evaluate.
3. The call with the format attribute is reached and is constant-evaluatable: the constant evaluator needs to check that the format string matches the arguments anyway (because in general we permit non-constant format strings), so it will need to properly handle this case.

Case 3 is currently theoretical (we don't constant-evaluate any format-string functions). But in any case, I think the conclusion remains: we don't need to do these checks in a constant-evaluated context.

[Tyker]

most comments were fixed. but there is a few point on which the direction isn't clear.

I think all of the new warnings in the test cases here are undesirable (they duplicate errors produced by the constant evaluator)

in the last revision warn_impcast_integer_precision_constant was always diagnosed in constant expression and diagnosed if reachable otherwise. this warning was responsible of all new warnings in test cases but isn't diagnosed by the constant evaluator. which is correct because integral casts have implementation defined or well defined behavior, not undefined behavior.

For code patterns that are suspicious but defined (eg, we think they might do something other than what the programmer intended), we should typically diagnose using Diag, regardless of whether they appear in a constant context.

this is not currently done even in non constant context. for example warn_impcast_integer_precision_constant is diagnosed via DiagRuntimeBehavior.

in this revision

constexpr int i0 = (long long)__builtin_is_constant_evaluated() * (1ll << 33); //#1

is diagnosed as it should but the reason it is diagnosed isn't correct. it is diagnosed only because CheckCompletedExpr doesn't push an ExpressionEvaluationContextRecord but just set isConstantEvaluatedOverride. so DiagRuntimeBehavior doesn't see the context as constant context. This means that other contexts that are correctly in constant context like:

case (long long)__builtin_is_constant_evaluated() * (1ll << 33):; //#1

will not be diagnosed even though they probably should.

[Tyker]

@rsmith friendly ping.

[Tyker]

@rsmith ping.

[rsmith]

Thanks, looks great.

clang/lib/Sema/SemaChecking.cpp

13941

permormed -> performed

clang/include/clang/AST/Expr.h

public:
   /// applied.
   bool EvaluateAsRValue(EvalResult &Result, const ASTContext &Ctx,
                         bool InConstantContext = false) const;
 
   /// EvaluateAsBooleanCondition - Return true if this is a constant
   /// which we can fold and convert to a boolean condition using
   /// any crazy technique that we want to, even if the expression has
   /// side-effects.
-  bool EvaluateAsBooleanCondition(bool &Result, const ASTContext &Ctx) const;
+  bool EvaluateAsBooleanCondition(bool &Result, const ASTContext &Ctx,
+                                  bool InConstantContext = false) const;
 
   enum SideEffectsKind {
     SE_NoSideEffects,          ///< Strictly evaluate the expression.
     SE_AllowUndefinedBehavior, ///< Allow UB that we can give a value, but not
                                ///< arbitrary unmodeled side effects.
     SE_AllowSideEffects        ///< Allow any unmodeled side effect.
   };
 
   /// EvaluateAsInt - Return true if this is a constant which we can fold and
   /// convert to an integer, using any crazy technique that we want to.
   bool EvaluateAsInt(EvalResult &Result, const ASTContext &Ctx,
-                     SideEffectsKind AllowSideEffects = SE_NoSideEffects) const;
+                     SideEffectsKind AllowSideEffects = SE_NoSideEffects,
+                     bool InConstantContext = false) const;
 
   /// EvaluateAsFloat - Return true if this is a constant which we can fold and
   /// convert to a floating point value, using any crazy technique that we
   /// want to.
-  bool
-  EvaluateAsFloat(llvm::APFloat &Result, const ASTContext &Ctx,
-                  SideEffectsKind AllowSideEffects = SE_NoSideEffects) const;
+  bool EvaluateAsFloat(llvm::APFloat &Result, const ASTContext &Ctx,
+                       SideEffectsKind AllowSideEffects = SE_NoSideEffects,
+                       bool InConstantContext = false) const;
 
   /// EvaluateAsFloat - Return true if this is a constant which we can fold and
   /// convert to a fixed point value.
-  bool EvaluateAsFixedPoint(
-      EvalResult &Result, const ASTContext &Ctx,
-      SideEffectsKind AllowSideEffects = SE_NoSideEffects) const;
+  bool EvaluateAsFixedPoint(EvalResult &Result, const ASTContext &Ctx,
+                            SideEffectsKind AllowSideEffects = SE_NoSideEffects,
+                            bool InConstantContext = false) const;
 
   /// isEvaluatable - Call EvaluateAsRValue to see if this expression can be
   /// constant folded without side-effects, but discard the result.
   bool isEvaluatable(const ASTContext &Ctx,
                      SideEffectsKind AllowSideEffects = SE_NoSideEffects) const;
 
   /// HasSideEffects - This routine returns true for all those expressions
   /// which have any effect other than producing a value. Example is a function
public:
   llvm::APSInt EvaluateKnownConstIntCheckOverflow(
       const ASTContext &Ctx,
       SmallVectorImpl<PartialDiagnosticAt> *Diag = nullptr) const;
 
   void EvaluateForOverflow(const ASTContext &Ctx) const;
 
   /// EvaluateAsLValue - Evaluate an expression to see if we can fold it to an
   /// lvalue with link time known address, with no side-effects.
-  bool EvaluateAsLValue(EvalResult &Result, const ASTContext &Ctx) const;
+  bool EvaluateAsLValue(EvalResult &Result, const ASTContext &Ctx,
+                        bool InConstantContext = false) const;
 
   /// EvaluateAsInitializer - Evaluate an expression as if it were the
   /// initializer of the given declaration. Returns true if the initializer
   /// can be folded to a constant, and produces any relevant notes. In C++11,
   /// notes will be produced if the expression is not a constant expression.
   bool EvaluateAsInitializer(APValue &Result, const ASTContext &Ctx,
                              const VarDecl *VD,
                              SmallVectorImpl<PartialDiagnosticAt> &Notes) const;

clang/include/clang/Sema/Sema.h

void pop() {
       SavedContext = nullptr;
     }
 
     ~ContextRAII() {
       pop();
     }
   };
 
+  bool isConstantEvaluated() {
+    return ExprEvalContexts.back().isConstantEvaluated();
+  }
+
   /// RAII object to handle the state changes required to synthesize
   /// a function body.
   class SynthesizedFunctionScope {
     Sema &S;
     Sema::ContextRAII SavedContext;
     bool PushedCodeSynthesisContext = false;
 
   public:

clang/lib/AST/ExprConstant.cpp

bool Expr::EvaluateAsRValue(EvalResult &Result, const ASTContext &Ctx,
                             bool InConstantContext) const {
   assert(!isValueDependent() &&
          "Expression evaluator can't be called on a dependent expression.");
   EvalInfo Info(Ctx, Result, EvalInfo::EM_IgnoreSideEffects);
   Info.InConstantContext = InConstantContext;
   return ::EvaluateAsRValue(this, Result, Ctx, Info);
 }
 
-bool Expr::EvaluateAsBooleanCondition(bool &Result,
-                                      const ASTContext &Ctx) const {
+bool Expr::EvaluateAsBooleanCondition(bool &Result, const ASTContext &Ctx,
+                                      bool InConstantContext) const {
   assert(!isValueDependent() &&
          "Expression evaluator can't be called on a dependent expression.");
   EvalResult Scratch;
   return EvaluateAsRValue(Scratch, Ctx) &&
          HandleConversionToBool(Scratch.Val, Result);
 }
 
 bool Expr::EvaluateAsInt(EvalResult &Result, const ASTContext &Ctx,
-                         SideEffectsKind AllowSideEffects) const {
+                         SideEffectsKind AllowSideEffects,
+                         bool InConstantContext) const {
   assert(!isValueDependent() &&
          "Expression evaluator can't be called on a dependent expression.");
   EvalInfo Info(Ctx, Result, EvalInfo::EM_IgnoreSideEffects);
+  Info.InConstantContext = InConstantContext;
   return ::EvaluateAsInt(this, Result, Ctx, AllowSideEffects, Info);
 }
 
 bool Expr::EvaluateAsFixedPoint(EvalResult &Result, const ASTContext &Ctx,
-                                SideEffectsKind AllowSideEffects) const {
+                                SideEffectsKind AllowSideEffects,
+                                bool InConstantContext) const {
   assert(!isValueDependent() &&
          "Expression evaluator can't be called on a dependent expression.");
   EvalInfo Info(Ctx, Result, EvalInfo::EM_IgnoreSideEffects);
+  Info.InConstantContext = InConstantContext;
   return ::EvaluateAsFixedPoint(this, Result, Ctx, AllowSideEffects, Info);
 }
 
 bool Expr::EvaluateAsFloat(APFloat &Result, const ASTContext &Ctx,
-                           SideEffectsKind AllowSideEffects) const {
+                           SideEffectsKind AllowSideEffects,
+                           bool InConstantContext) const {
   assert(!isValueDependent() &&
          "Expression evaluator can't be called on a dependent expression.");
 
   if (!getType()->isRealFloatingType())
     return false;
 
   EvalResult ExprResult;
   if (!EvaluateAsRValue(ExprResult, Ctx) || !ExprResult.Val.isFloat() ||
       hasUnacceptableSideEffect(ExprResult, AllowSideEffects))
     return false;
 
   Result = ExprResult.Val.getFloat();
   return true;
 }
 
-bool Expr::EvaluateAsLValue(EvalResult &Result, const ASTContext &Ctx) const {
+bool Expr::EvaluateAsLValue(EvalResult &Result, const ASTContext &Ctx,
+                            bool InConstantContext) const {
   assert(!isValueDependent() &&
          "Expression evaluator can't be called on a dependent expression.");
 
   EvalInfo Info(Ctx, Result, EvalInfo::EM_ConstantFold);
-
+  Info.InConstantContext = InConstantContext;
   LValue LV;
   if (!EvaluateLValue(this, LV, Info) || Result.HasSideEffects ||
       !CheckLValueConstantExpression(Info, getExprLoc(),
                                      Ctx.getLValueReferenceType(getType()), LV,
                                      Expr::EvaluateForCodeGen))
     return false;
 
   LV.moveInto(Result.Val);

clang/lib/Sema/SemaChecking.cpp

 /// __builtin_*_chk function, then use the object size argument specified in the
 /// source. Otherwise, infer the object size using __builtin_object_size.
 void Sema::checkFortifiedBuiltinMemoryFunction(FunctionDecl *FD,
                                                CallExpr *TheCall) {
   // FIXME: There are some more useful checks we could be doing here:
   //  - Analyze the format string of sprintf to see how much of buffer is used.
   //  - Evaluate strlen of strcpy arguments, use as object size.
 
-  if (TheCall->isValueDependent() || TheCall->isTypeDependent())
+  if (TheCall->isValueDependent() || TheCall->isTypeDependent() ||
+      isConstantEvaluated())
     return;
 
   unsigned BuiltinID = FD->getBuiltinID(/*ConsiderWrappers=*/true);
   if (!BuiltinID)
     return;
 
   unsigned DiagID = 0;
   bool IsChkVariant = false;
if (UT->getDecl()->hasAttr<TransparentUnionAttr>())
           dyn_cast<CompoundLiteralExpr>(Expr))
         if (const InitListExpr *ILE =
             dyn_cast<InitListExpr>(CLE->getInitializer()))
           Expr = ILE->getInit(0);
   }
 
   bool Result;
   return (!Expr->isValueDependent() &&
-          Expr->EvaluateAsBooleanCondition(Result, S.Context) &&
+          Expr->EvaluateAsBooleanCondition(Result, S.Context,
+                                           S.isConstantEvaluated()) &&
           !Result);
 }
 
+static void DiagRuntimeOrConstant(Sema &S, SourceLocation Loc, const Expr *Expr,
+                                  PartialDiagnostic PDiag) {
+  if (!S.isConstantEvaluated())
+    S.DiagRuntimeBehavior(Loc, Expr, PDiag);
+  else
+    S.Diag(Loc, PDiag);
+}
+
 static void CheckNonNullArgument(Sema &S,
                                  const Expr *ArgExpr,
                                  SourceLocation CallSiteLoc) {
   if (CheckNonNullExpr(S, ArgExpr))
-    S.DiagRuntimeBehavior(CallSiteLoc, ArgExpr,
-           S.PDiag(diag::warn_null_arg) << ArgExpr->getSourceRange());
+    DiagRuntimeOrConstant(S, CallSiteLoc, ArgExpr,
+                          S.PDiag(diag::warn_null_arg)
+                              << ArgExpr->getSourceRange());
 }
 
 bool Sema::GetFormatNSStringIdx(const FormatAttr *Format, unsigned &Idx) {
   FormatStringInfo FSI;
   if ((GetFormatStringType(Format) == FST_NSString) &&
       getFormatStringInfo(Format, false, &FSI)) {
     Idx = FSI.FormatIdx;
     return true;
   }
bool Sema::SemaBuiltinConstantArgRange(CallExpr *TheCall, int ArgNum,
   if (Arg->isTypeDependent() || Arg->isValueDependent())
     return false;
 
   // Check constant-ness first.
   if (SemaBuiltinConstantArg(TheCall, ArgNum, Result))
     return true;
 
   if (Result.getSExtValue() < Low || Result.getSExtValue() > High) {
-    if (RangeIsError)
+    if (RangeIsError || isConstantEvaluated())
       return Diag(TheCall->getBeginLoc(), diag::err_argument_invalid_range)
              << Result.toString(10) << Low << High << Arg->getSourceRange();
     else
       // Defer the warning until we know if the code will be emitted so that
       // dead code can ignore this.
       DiagRuntimeBehavior(TheCall->getBeginLoc(), TheCall,
                           PDiag(diag::warn_argument_invalid_range)
                               << Result.toString(10) << Low << High
const AbstractConditionalOperator *C =
         cast<AbstractConditionalOperator>(E);
 
     // Determine whether it is necessary to check both sub-expressions, for
     // example, because the condition expression is a constant that can be
     // evaluated at compile time.
     bool CheckLeft = true, CheckRight = true;
 
     bool Cond;
-    if (C->getCond()->EvaluateAsBooleanCondition(Cond, S.getASTContext())) {
+    if (C->getCond()->EvaluateAsBooleanCondition(Cond, S.getASTContext(),
+                                                 S.isConstantEvaluated())) {
       if (Cond)
         CheckRight = false;
       else
         CheckLeft = false;
     }
 
     // We need to maintain the offsets for the right and the left hand side
     // separately to check if every possible indexed expression is a valid
tryAgain:
   }
   case Stmt::BinaryOperatorClass: {
     const BinaryOperator *BinOp = cast<BinaryOperator>(E);
 
     // A string literal + an int offset is still a string literal.
     if (BinOp->isAdditiveOp()) {
       Expr::EvalResult LResult, RResult;
 
-      bool LIsInt = BinOp->getLHS()->EvaluateAsInt(LResult, S.Context);
-      bool RIsInt = BinOp->getRHS()->EvaluateAsInt(RResult, S.Context);
+      bool LIsInt = BinOp->getLHS()->EvaluateAsInt(
+          LResult, S.Context, Expr::SE_NoSideEffects, S.isConstantEvaluated());
+      bool RIsInt = BinOp->getRHS()->EvaluateAsInt(
+          RResult, S.Context, Expr::SE_NoSideEffects, S.isConstantEvaluated());
 
       if (LIsInt != RIsInt) {
         BinaryOperatorKind BinOpKind = BinOp->getOpcode();
 
         if (LIsInt) {
           if (BinOpKind == BO_Add) {
             sumOffsets(Offset, LResult.Val.getInt(), BinOpKind, RIsInt);
             E = BinOp->getRHS();
case Stmt::BinaryOperatorClass: {
 
     return SLCT_NotALiteral;
   }
   case Stmt::UnaryOperatorClass: {
     const UnaryOperator *UnaOp = cast<UnaryOperator>(E);
     auto ASE = dyn_cast<ArraySubscriptExpr>(UnaOp->getSubExpr());
     if (UnaOp->getOpcode() == UO_AddrOf && ASE) {
       Expr::EvalResult IndexResult;
-      if (ASE->getRHS()->EvaluateAsInt(IndexResult, S.Context)) {
+      if (ASE->getRHS()->EvaluateAsInt(IndexResult, S.Context,
+                                       Expr::SE_NoSideEffects,
+                                       S.isConstantEvaluated())) {
         sumOffsets(Offset, IndexResult.Val.getInt(), BO_Add,
                    /*RHS is int*/ true);
         E = ASE->getBase();
         goto tryAgain;
       }
     }
 
     return SLCT_NotALiteral;
if (const AtomicType *AtomicRHS = Ty->getAs<AtomicType>())
     Ty = AtomicRHS->getValueType();
   return Ty;
 }
 
 /// Pseudo-evaluate the given integer expression, estimating the
 /// range of values it might take.
 ///
 /// \param MaxWidth - the width to which the value will be truncated
-static IntRange GetExprRange(ASTContext &C, const Expr *E, unsigned MaxWidth) {
+static IntRange GetExprRange(ASTContext &C, const Expr *E, unsigned MaxWidth,
+                             bool InConstantContext) {
   E = E->IgnoreParens();
 
   // Try a full evaluation first.
   Expr::EvalResult result;
-  if (E->EvaluateAsRValue(result, C))
+  if (E->EvaluateAsRValue(result, C, InConstantContext))
     return GetValueRange(C, result.Val, GetExprType(E), MaxWidth);
 
   // I think we only want to look through implicit casts here; if the
   // user has an explicit widening cast, we should treat the value as
   // being of the new, wider type.
   if (const auto *CE = dyn_cast<ImplicitCastExpr>(E)) {
     if (CE->getCastKind() == CK_NoOp || CE->getCastKind() == CK_LValueToRValue)
-      return GetExprRange(C, CE->getSubExpr(), MaxWidth);
+      return GetExprRange(C, CE->getSubExpr(), MaxWidth, InConstantContext);
 
     IntRange OutputTypeRange = IntRange::forValueOfType(C, GetExprType(CE));
 
     bool isIntegerCast = CE->getCastKind() == CK_IntegralCast ||
                          CE->getCastKind() == CK_BooleanToSignedIntegral;
 
     // Assume that non-integer casts can span the full range of the type.
     if (!isIntegerCast)
       return OutputTypeRange;
 
-    IntRange SubRange
-      = GetExprRange(C, CE->getSubExpr(),
-                     std::min(MaxWidth, OutputTypeRange.Width));
+    IntRange SubRange = GetExprRange(C, CE->getSubExpr(),
+                                     std::min(MaxWidth, OutputTypeRange.Width),
+                                     InConstantContext);
 
     // Bail out if the subexpr's range is as wide as the cast type.
     if (SubRange.Width >= OutputTypeRange.Width)
       return OutputTypeRange;
 
     // Otherwise, we take the smaller width, and we're non-negative if
     // either the output type or the subexpr is.
     return IntRange(SubRange.Width,
                     SubRange.NonNegative || OutputTypeRange.NonNegative);
   }
 
   if (const auto *CO = dyn_cast<ConditionalOperator>(E)) {
     // If we can fold the condition, just take that operand.
     bool CondResult;
     if (CO->getCond()->EvaluateAsBooleanCondition(CondResult, C))
-      return GetExprRange(C, CondResult ? CO->getTrueExpr()
-                                        : CO->getFalseExpr(),
-                          MaxWidth);
+      return GetExprRange(C,
+                          CondResult ? CO->getTrueExpr() : CO->getFalseExpr(),
+                          MaxWidth, InConstantContext);
 
     // Otherwise, conservatively merge.
-    IntRange L = GetExprRange(C, CO->getTrueExpr(), MaxWidth);
-    IntRange R = GetExprRange(C, CO->getFalseExpr(), MaxWidth);
+    IntRange L =
+        GetExprRange(C, CO->getTrueExpr(), MaxWidth, InConstantContext);
+    IntRange R =
+        GetExprRange(C, CO->getFalseExpr(), MaxWidth, InConstantContext);
     return IntRange::join(L, R);
   }
 
   if (const auto *BO = dyn_cast<BinaryOperator>(E)) {
     switch (BO->getOpcode()) {
     case BO_Cmp:
       llvm_unreachable("builtin <=> should have class type");
 
if (const auto *BO = dyn_cast<BinaryOperator>(E)) {
     case BO_OrAssign:
       // TODO: bitfields?
       return IntRange::forValueOfType(C, GetExprType(E));
 
     // Simple assignments just pass through the RHS, which will have
     // been coerced to the LHS type.
     case BO_Assign:
       // TODO: bitfields?
-      return GetExprRange(C, BO->getRHS(), MaxWidth);
+      return GetExprRange(C, BO->getRHS(), MaxWidth, InConstantContext);
 
     // Operations with opaque sources are black-listed.
     case BO_PtrMemD:
     case BO_PtrMemI:
       return IntRange::forValueOfType(C, GetExprType(E));
 
     // Bitwise-and uses the *infinum* of the two source ranges.
     case BO_And:
     case BO_AndAssign:
-      return IntRange::meet(GetExprRange(C, BO->getLHS(), MaxWidth),
-                            GetExprRange(C, BO->getRHS(), MaxWidth));
+      return IntRange::meet(
+          GetExprRange(C, BO->getLHS(), MaxWidth, InConstantContext),
+          GetExprRange(C, BO->getRHS(), MaxWidth, InConstantContext));
 
     // Left shift gets black-listed based on a judgement call.
     case BO_Shl:
       // ...except that we want to treat '1 << (blah)' as logically
       // positive.  It's an important idiom.
       if (IntegerLiteral *I
             = dyn_cast<IntegerLiteral>(BO->getLHS()->IgnoreParenCasts())) {
         if (I->getValue() == 1) {
           IntRange R = IntRange::forValueOfType(C, GetExprType(E));
           return IntRange(R.Width, /*NonNegative*/ true);
         }
       }
       LLVM_FALLTHROUGH;
 
     case BO_ShlAssign:
       return IntRange::forValueOfType(C, GetExprType(E));
 
     // Right shift by a constant can narrow its left argument.
     case BO_Shr:
     case BO_ShrAssign: {
-      IntRange L = GetExprRange(C, BO->getLHS(), MaxWidth);
+      IntRange L = GetExprRange(C, BO->getLHS(), MaxWidth, InConstantContext);
 
       // If the shift amount is a positive constant, drop the width by
       // that much.
       llvm::APSInt shift;
       if (BO->getRHS()->isIntegerConstantExpr(shift, C) &&
           shift.isNonNegative()) {
         unsigned zext = shift.getZExtValue();
         if (zext >= L.Width)
           L.Width = (L.NonNegative ? 0 : 1);
         else
           L.Width -= zext;
       }
 
       return L;
     }
 
     // Comma acts as its right operand.
     case BO_Comma:
-      return GetExprRange(C, BO->getRHS(), MaxWidth);
+      return GetExprRange(C, BO->getRHS(), MaxWidth, InConstantContext);
 
     // Black-list pointer subtractions.
     case BO_Sub:
       if (BO->getLHS()->getType()->isPointerType())
         return IntRange::forValueOfType(C, GetExprType(E));
       break;
 
     // The width of a division result is mostly determined by the size
     // of the LHS.
     case BO_Div: {
       // Don't 'pre-truncate' the operands.
       unsigned opWidth = C.getIntWidth(GetExprType(E));
-      IntRange L = GetExprRange(C, BO->getLHS(), opWidth);
+      IntRange L = GetExprRange(C, BO->getLHS(), opWidth, InConstantContext);
 
       // If the divisor is constant, use that.
       llvm::APSInt divisor;
       if (BO->getRHS()->isIntegerConstantExpr(divisor, C)) {
         unsigned log2 = divisor.logBase2(); // floor(log_2(divisor))
         if (log2 >= L.Width)
           L.Width = (L.NonNegative ? 0 : 1);
         else
           L.Width = std::min(L.Width - log2, MaxWidth);
         return L;
       }
 
       // Otherwise, just use the LHS's width.
-      IntRange R = GetExprRange(C, BO->getRHS(), opWidth);
+      IntRange R = GetExprRange(C, BO->getRHS(), opWidth, InConstantContext);
       return IntRange(L.Width, L.NonNegative && R.NonNegative);
     }
 
     // The result of a remainder can't be larger than the result of
     // either side.
     case BO_Rem: {
       // Don't 'pre-truncate' the operands.
       unsigned opWidth = C.getIntWidth(GetExprType(E));
-      IntRange L = GetExprRange(C, BO->getLHS(), opWidth);
-      IntRange R = GetExprRange(C, BO->getRHS(), opWidth);
+      IntRange L = GetExprRange(C, BO->getLHS(), opWidth, InConstantContext);
+      IntRange R = GetExprRange(C, BO->getRHS(), opWidth, InConstantContext);
 
       IntRange meet = IntRange::meet(L, R);
       meet.Width = std::min(meet.Width, MaxWidth);
       return meet;
     }
 
     // The default behavior is okay for these.
     case BO_Mul:
     case BO_Add:
     case BO_Xor:
     case BO_Or:
       break;
     }
 
     // The default case is to treat the operation as if it were closed
     // on the narrowest type that encompasses both operands.
-    IntRange L = GetExprRange(C, BO->getLHS(), MaxWidth);
-    IntRange R = GetExprRange(C, BO->getRHS(), MaxWidth);
+    IntRange L = GetExprRange(C, BO->getLHS(), MaxWidth, InConstantContext);
+    IntRange R = GetExprRange(C, BO->getRHS(), MaxWidth, InConstantContext);
     return IntRange::join(L, R);
   }
 
   if (const auto *UO = dyn_cast<UnaryOperator>(E)) {
     switch (UO->getOpcode()) {
     // Boolean-valued operations are white-listed.
     case UO_LNot:
       return IntRange::forBoolType();
 
     // Operations with opaque sources are black-listed.
     case UO_Deref:
     case UO_AddrOf: // should be impossible
       return IntRange::forValueOfType(C, GetExprType(E));
 
     default:
-      return GetExprRange(C, UO->getSubExpr(), MaxWidth);
+      return GetExprRange(C, UO->getSubExpr(), MaxWidth, InConstantContext);
     }
   }
 
   if (const auto *OVE = dyn_cast<OpaqueValueExpr>(E))
-    return GetExprRange(C, OVE->getSourceExpr(), MaxWidth);
+    return GetExprRange(C, OVE->getSourceExpr(), MaxWidth, InConstantContext);
 
   if (const auto *BitField = E->getSourceBitField())
     return IntRange(BitField->getBitWidthValue(C),
                     BitField->getType()->isUnsignedIntegerOrEnumerationType());
 
   return IntRange::forValueOfType(C, GetExprType(E));
 }
 
-static IntRange GetExprRange(ASTContext &C, const Expr *E) {
-  return GetExprRange(C, E, C.getIntWidth(GetExprType(E)));
+static IntRange GetExprRange(ASTContext &C, const Expr *E,
+                             bool InConstantContext) {
+  return GetExprRange(C, E, C.getIntWidth(GetExprType(E)), InConstantContext);
 }
 
 /// Checks whether the given value, which currently has the given
 /// source semantics, has the same value when coerced through the
 /// target semantics.
 static bool IsSameFloatAfterCast(const llvm::APFloat &value,
                                  const llvm::fltSemantics &Src,
                                  const llvm::fltSemantics &Tgt) {
if (ED)
     OS << '\'' << *ED << "' (" << Value << ")";
   else
     OS << Value;
 
   // FIXME: We use a somewhat different formatting for the in-range cases and
   // cases involving boolean values for historical reasons. We should pick a
   // consistent way of presenting these diagnostics.
   if (!InRange || Other->isKnownToHaveBooleanValue()) {
-    S.DiagRuntimeBehavior(
-      E->getOperatorLoc(), E,
+    DiagRuntimeOrConstant(
+        S, E->getOperatorLoc(), E,
         S.PDiag(!InRange ? diag::warn_out_of_range_compare
                          : diag::warn_tautological_bool_compare)
-          << OS.str() << classifyConstantValue(Constant)
-          << OtherT << OtherIsBooleanDespiteType << *Result
+            << OS.str() << classifyConstantValue(Constant) << OtherT
+            << OtherIsBooleanDespiteType << *Result
             << E->getLHS()->getSourceRange() << E->getRHS()->getSourceRange());
   } else {
     unsigned Diag = (isKnownToHaveUnsignedValue(OriginalOther) && Value == 0)
                         ? (HasEnumType(OriginalOther)
                                ? diag::warn_unsigned_enum_always_true_comparison
                                : diag::warn_unsigned_always_true_comparison)
                         : diag::warn_tautological_constant_compare;
 
     S.Diag(E->getOperatorLoc(), Diag)
static void AnalyzeComparison(Sema &S, BinaryOperator *E) {
   } else if (RHS->getType()->hasSignedIntegerRepresentation()) {
     signedOperand = RHS;
     unsignedOperand = LHS;
   } else {
     return AnalyzeImpConvsInComparison(S, E);
   }
 
   // Otherwise, calculate the effective range of the signed operand.
-  IntRange signedRange = GetExprRange(S.Context, signedOperand);
+  IntRange signedRange =
+      GetExprRange(S.Context, signedOperand, S.isConstantEvaluated());
 
   // Go ahead and analyze implicit conversions in the operands.  Note
   // that we skip the implicit conversions on both sides.
   AnalyzeImplicitConversions(S, LHS, E->getOperatorLoc());
   AnalyzeImplicitConversions(S, RHS, E->getOperatorLoc());
 
   // If the signed range is non-negative, -Wsign-compare won't fire.
   if (signedRange.NonNegative)
     return;
 
   // For (in)equality comparisons, if the unsigned operand is a
   // constant which cannot collide with a overflowed signed operand,
   // then reinterpreting the signed operand as unsigned will not
   // change the result of the comparison.
   if (E->isEqualityOp()) {
     unsigned comparisonWidth = S.Context.getIntWidth(T);
-    IntRange unsignedRange = GetExprRange(S.Context, unsignedOperand);
+    IntRange unsignedRange =
+        GetExprRange(S.Context, unsignedOperand, S.isConstantEvaluated());
 
     // We should never be unable to prove that the unsigned operand is
     // non-negative.
     assert(unsignedRange.NonNegative && "unsigned range includes negative?");
 
     if (unsignedRange.Width < comparisonWidth)
       return;
   }
 
-  S.DiagRuntimeBehavior(E->getOperatorLoc(), E,
+  DiagRuntimeOrConstant(S, E->getOperatorLoc(), E,
                         S.PDiag(diag::warn_mixed_sign_comparison)
                             << LHS->getType() << RHS->getType()
                             << LHS->getSourceRange() << RHS->getSourceRange());
 }
 
 /// Analyzes an attempt to assign the given value to a bitfield.
 ///
 /// Returns true if there was something fishy about the attempt.
 static bool AnalyzeBitFieldAssignment(Sema &S, FieldDecl *Bitfield, Expr *Init,
                                       SourceLocation InitLoc) {
   assert(Bitfield->isBitField());
if (E->getLHS()->getType()->isAtomicType())
     S.Diag(E->getRHS()->getBeginLoc(), diag::warn_atomic_implicit_seq_cst);
 }
 
 /// Diagnose an implicit cast;  purely a helper for CheckImplicitConversion.
 static void DiagnoseImpCast(Sema &S, Expr *E, QualType SourceType, QualType T,
                             SourceLocation CContext, unsigned diag,
                             bool pruneControlFlow = false) {
   if (pruneControlFlow) {
-    S.DiagRuntimeBehavior(E->getExprLoc(), E,
+    DiagRuntimeOrConstant(S, E->getExprLoc(), E,
                           S.PDiag(diag)
                               << SourceType << T << E->getSourceRange()
                               << SourceRange(CContext));
     return;
   }
   S.Diag(E->getExprLoc(), diag)
     << SourceType << T << E->getSourceRange() << SourceRange(CContext);
 }
 
 /// Diagnose an implicit cast;  purely a helper for CheckImplicitConversion.
 static void DiagnoseImpCast(Sema &S, Expr *E, QualType T,
static void DiagnoseFloatingImpCast(Sema &S, Expr *E, QualType T,
 
   SmallString<16> PrettyTargetValue;
   if (IsBool)
     PrettyTargetValue = Value.isZero() ? "false" : "true";
   else
     IntegerValue.toString(PrettyTargetValue);
 
   if (PruneWarnings) {
-    S.DiagRuntimeBehavior(E->getExprLoc(), E,
+    DiagRuntimeOrConstant(S, E->getExprLoc(), E,
                           S.PDiag(DiagID)
                               << E->getType() << T.getUnqualifiedType()
                               << PrettySourceValue << PrettyTargetValue
                               << E->getSourceRange() << SourceRange(CContext));
   } else {
     S.Diag(E->getExprLoc(), DiagID)
         << E->getType() << T.getUnqualifiedType() << PrettySourceValue
         << PrettyTargetValue << E->getSourceRange() << SourceRange(CContext);
if (SourceBT && SourceBT->isFloatingPoint()) {
     }
     return;
   }
 
   // Valid casts involving fixed point types should be accounted for here.
   if (Source->isFixedPointType()) {
     if (Target->isUnsaturatedFixedPointType()) {
       Expr::EvalResult Result;
-      if (E->EvaluateAsFixedPoint(Result, S.Context,
-                                  Expr::SE_AllowSideEffects)) {
+      if (E->EvaluateAsFixedPoint(Result, S.Context, Expr::SE_AllowSideEffects,
+                                  S.isConstantEvaluated())) {
         APFixedPoint Value = Result.Val.getFixedPoint();
         APFixedPoint MaxVal = S.Context.getFixedPointMax(T);
         APFixedPoint MinVal = S.Context.getFixedPointMin(T);
         if (Value > MaxVal || Value < MinVal) {
-          S.DiagRuntimeBehavior(E->getExprLoc(), E,
+          DiagRuntimeOrConstant(S, E->getExprLoc(), E,
                                 S.PDiag(diag::warn_impcast_fixed_point_range)
                                     << Value.toString() << T
                                     << E->getSourceRange()
                                     << clang::SourceRange(CC));
           return;
         }
       }
     } else if (Target->isIntegerType()) {
       Expr::EvalResult Result;
-      if (E->EvaluateAsFixedPoint(Result, S.Context,
+      if (!S.isConstantEvaluated() &&
+          E->EvaluateAsFixedPoint(Result, S.Context,
                                   Expr::SE_AllowSideEffects)) {
         APFixedPoint FXResult = Result.Val.getFixedPoint();
 
         bool Overflowed;
         llvm::APSInt IntResult = FXResult.convertToInt(
             S.Context.getIntWidth(T),
             Target->isSignedIntegerOrEnumerationType(), &Overflowed);
 
         if (Overflowed) {
           S.DiagRuntimeBehavior(E->getExprLoc(), E,
                                 S.PDiag(diag::warn_impcast_fixed_point_range)
                                     << FXResult.toString() << T
                                     << E->getSourceRange()
                                     << clang::SourceRange(CC));
           return;
         }
       }
     }
   } else if (Target->isUnsaturatedFixedPointType()) {
     if (Source->isIntegerType()) {
       Expr::EvalResult Result;
-      if (E->EvaluateAsInt(Result, S.Context, Expr::SE_AllowSideEffects)) {
+      if (!S.isConstantEvaluated() &&
+          E->EvaluateAsInt(Result, S.Context, Expr::SE_AllowSideEffects)) {
         llvm::APSInt Value = Result.Val.getInt();
 
         bool Overflowed;
         APFixedPoint IntResult = APFixedPoint::getFromIntValue(
             Value, S.Context.getFixedPointSemantics(T), &Overflowed);
 
         if (Overflowed) {
           S.DiagRuntimeBehavior(E->getExprLoc(), E,
CheckImplicitConversion(Sema &S, Expr *E, QualType T, SourceLocation CC,
   if (!Source->isIntegerType() || !Target->isIntegerType())
     return;
 
   // TODO: remove this early return once the false positives for constant->bool
   // in templates, macros, etc, are reduced or removed.
   if (Target->isSpecificBuiltinType(BuiltinType::Bool))
     return;
 
-  IntRange SourceRange = GetExprRange(S.Context, E);
+  IntRange SourceRange = GetExprRange(S.Context, E, S.isConstantEvaluated());
   IntRange TargetRange = IntRange::forTargetOfCanonicalType(S.Context, Target);
 
   if (SourceRange.Width > TargetRange.Width) {
     // If the source is a constant, use a default-on diagnostic.
     // TODO: this should happen for bitfield stores, too.
     Expr::EvalResult Result;
-    if (E->EvaluateAsInt(Result, S.Context, Expr::SE_AllowSideEffects)) {
+    if (E->EvaluateAsInt(Result, S.Context, Expr::SE_AllowSideEffects,
+                         S.isConstantEvaluated())) {
       llvm::APSInt Value(32);
       Value = Result.Val.getInt();
 
       if (S.SourceMgr.isInSystemMacro(CC))
         return;
 
       std::string PrettySourceValue = Value.toString(10);
       std::string PrettyTargetValue = PrettyPrintInRange(Value, TargetRange);
 
-      S.DiagRuntimeBehavior(E->getExprLoc(), E,
+      DiagRuntimeOrConstant(
+          S, E->getExprLoc(), E,
           S.PDiag(diag::warn_impcast_integer_precision_constant)
-            << PrettySourceValue << PrettyTargetValue
-            << E->getType() << T << E->getSourceRange()
-            << clang::SourceRange(CC));
+              << PrettySourceValue << PrettyTargetValue << E->getType() << T
+              << E->getSourceRange() << clang::SourceRange(CC));
       return;
     }
 
     // People want to build with -Wshorten-64-to-32 and not -Wconversion.
     if (S.SourceMgr.isInSystemMacro(CC))
       return;
 
     if (TargetRange.Width == 32 && S.Context.getIntWidth(E->getType()) == 64)
if (TargetRange.Width == SourceRange.Width && !TargetRange.NonNegative &&
     Expr::EvalResult Result;
     if (E->EvaluateAsInt(Result, S.Context, Expr::SE_AllowSideEffects) &&
         !S.SourceMgr.isInSystemMacro(CC)) {
       llvm::APSInt Value = Result.Val.getInt();
       if (isSameWidthConstantConversion(S, E, T, CC)) {
         std::string PrettySourceValue = Value.toString(10);
         std::string PrettyTargetValue = PrettyPrintInRange(Value, TargetRange);
 
-        S.DiagRuntimeBehavior(
-            E->getExprLoc(), E,
+        DiagRuntimeOrConstant(
+            S, E->getExprLoc(), E,
             S.PDiag(diag::warn_impcast_integer_precision_constant)
                 << PrettySourceValue << PrettyTargetValue << E->getType() << T
                 << E->getSourceRange() << clang::SourceRange(CC));
         return;
       }
     }
 
     // Fall through for non-constants to give a sign conversion warning.
~EvaluationTracker() {
       Self.EvalTracker = Prev;
       if (Prev)
         Prev->EvalOK &= EvalOK;
     }
 
     bool evaluate(const Expr *E, bool &Result) {
       if (!EvalOK || E->isValueDependent())
         return false;
-      EvalOK = E->EvaluateAsBooleanCondition(Result, Self.SemaRef.Context);
+      EvalOK = E->EvaluateAsBooleanCondition(
+          Result, Self.SemaRef.Context, Self.SemaRef.isConstantEvaluated());
       return EvalOK;
     }
 
   private:
     SequenceChecker &Self;
     EvaluationTracker *Prev;
     bool EvalOK = true;
   } *EvalTracker = nullptr;
void Sema::CheckUnsequencedOperations(Expr *E) {
   while (!WorkList.empty()) {
     Expr *Item = WorkList.pop_back_val();
     SequenceChecker(*this, Item, WorkList);
   }
 }
 
 void Sema::CheckCompletedExpr(Expr *E, SourceLocation CheckLoc,
                               bool IsConstexpr) {
+  if (IsConstexpr)
+    PushExpressionEvaluationContext(
+        ExpressionEvaluationContext::ConstantEvaluated);
   CheckImplicitConversions(E, CheckLoc);
   if (!E->isInstantiationDependent())
     CheckUnsequencedOperations(E);
   if (!IsConstexpr && !E->isValueDependent())
     CheckForIntOverflow(E);
   DiagnoseMisalignedMembers();
+  if (IsConstexpr)
+    PopExpressionEvaluationContext();
 }
 
 void Sema::CheckBitFieldInitialization(SourceLocation InitLoc,
                                        FieldDecl *BitField,
                                        Expr *Init) {
   (void) AnalyzeBitFieldAssignment(*this, BitField, Init, InitLoc);
 }
 

 
 void Sema::CheckArrayAccess(const Expr *BaseExpr, const Expr *IndexExpr,
                             const ArraySubscriptExpr *ASE,
                             bool AllowOnePastEnd, bool IndexNegated) {
   IndexExpr = IndexExpr->IgnoreParenImpCasts();
   if (IndexExpr->isValueDependent())
     return;
 
+  // already diagnosed by the constant evaluator.
+  if (isConstantEvaluated())
+    return;
+
   const Type *EffectiveType =
       BaseExpr->getType()->getPointeeOrArrayElementType();
   BaseExpr = BaseExpr->IgnoreParenCasts();
   const ConstantArrayType *ArrayTy =
       Context.getAsConstantArrayType(BaseExpr->getType());
 
   if (!ArrayTy)
     return;

 
 /// Given a type tag expression find the type tag itself.
 ///
 /// \param TypeExpr Type tag expression, as it appears in user's code.
 ///
 /// \param VD Declaration of an identifier that appears in a type tag.
 ///
 /// \param MagicValue Type tag magic value.
+///
+/// \param isConstantEvaluated wether the evalaution should be permormed in
+/// constant context.
 static bool FindTypeTagExpr(const Expr *TypeExpr, const ASTContext &Ctx,
-                            const ValueDecl **VD, uint64_t *MagicValue) {
+                            const ValueDecl **VD, uint64_t *MagicValue,
+                            bool isConstantEvaluated) {
   while(true) {
     if (!TypeExpr)
       return false;
 
     TypeExpr = TypeExpr->IgnoreParenImpCasts()->IgnoreParenCasts();
 
     switch (TypeExpr->getStmtClass()) {
     case Stmt::UnaryOperatorClass: {
case Stmt::IntegerLiteralClass: {
         return false;
     }
 
     case Stmt::BinaryConditionalOperatorClass:
     case Stmt::ConditionalOperatorClass: {
       const AbstractConditionalOperator *ACO =
           cast<AbstractConditionalOperator>(TypeExpr);
       bool Result;
-      if (ACO->getCond()->EvaluateAsBooleanCondition(Result, Ctx)) {
+      if (ACO->getCond()->EvaluateAsBooleanCondition(Result, Ctx,
+                                                     isConstantEvaluated)) {
         if (Result)
           TypeExpr = ACO->getTrueExpr();
         else
           TypeExpr = ACO->getFalseExpr();
         continue;
       }
       return false;
     }

 /// Retrieve the C type corresponding to type tag TypeExpr.
 ///
 /// \param TypeExpr Expression that specifies a type tag.
 ///
 /// \param MagicValues Registered magic values.
 ///
 /// \param FoundWrongKind Set to true if a type tag was found, but of a wrong
 ///        kind.
 ///
 /// \param TypeInfo Information about the corresponding C type.
 ///
+/// \param isConstantEvaluated wether the evalaution should be permormed in
+/// constant context.
+///
 /// \returns true if the corresponding C type was found.
 static bool GetMatchingCType(
-        const IdentifierInfo *ArgumentKind,
-        const Expr *TypeExpr, const ASTContext &Ctx,
-        const llvm::DenseMap<Sema::TypeTagMagicValue,
-                             Sema::TypeTagData> *MagicValues,
-        bool &FoundWrongKind,
-        Sema::TypeTagData &TypeInfo) {
+    const IdentifierInfo *ArgumentKind, const Expr *TypeExpr,
+    const ASTContext &Ctx,
+    const llvm::DenseMap<Sema::TypeTagMagicValue, Sema::TypeTagData>
+        *MagicValues,
+    bool &FoundWrongKind, Sema::TypeTagData &TypeInfo,
+    bool isConstantEvaluated) {
   FoundWrongKind = false;
 
   // Variable declaration that has type_tag_for_datatype attribute.
   const ValueDecl *VD = nullptr;
 
   uint64_t MagicValue;
 
-  if (!FindTypeTagExpr(TypeExpr, Ctx, &VD, &MagicValue))
+  if (!FindTypeTagExpr(TypeExpr, Ctx, &VD, &MagicValue, isConstantEvaluated))
     return false;
 
   if (VD) {
     if (TypeTagForDatatypeAttr *I = VD->getAttr<TypeTagForDatatypeAttr>()) {
       if (I->getArgumentKind() != ArgumentKind) {
         FoundWrongKind = true;
         return false;
       }
if (TypeTagIdxAST >= ExprArgs.size()) {
     Diag(CallSiteLoc, diag::err_tag_index_out_of_range)
         << 0 << Attr->getTypeTagIdx().getSourceIndex();
     return;
   }
   const Expr *TypeTagExpr = ExprArgs[TypeTagIdxAST];
   bool FoundWrongKind;
   TypeTagData TypeInfo;
   if (!GetMatchingCType(ArgumentKind, TypeTagExpr, Context,
-                        TypeTagForDatatypeMagicValues.get(),
-                        FoundWrongKind, TypeInfo)) {
+                        TypeTagForDatatypeMagicValues.get(), FoundWrongKind,
+                        TypeInfo, isConstantEvaluated())) {
     if (FoundWrongKind)
       Diag(TypeTagExpr->getExprLoc(),
            diag::warn_type_tag_for_datatype_wrong_kind)
         << TypeTagExpr->getSourceRange();
     return;
   }
 
   // Retrieve the argument representing the 'arg_idx'.

clang/test/CXX/expr/expr.const/p3-0x.cpp

 //  integral conversions other than narrowing conversions,
 int b(unsigned n) {
   switch (n) {
     case E6:
     case EE::EE32: // expected-error {{not implicitly convertible}}
     case (int)EE::EE32:
     case 1000:
     case (long long)1e10: // expected-error {{case value evaluates to 10000000000, which cannot be narrowed to type 'unsigned int'}}
+    //expected-warning@-1 {{implicit conversion}}
     case -3: // expected-error {{case value evaluates to -3, which cannot be narrowed to type 'unsigned int'}}
       return n;
   }
   return 0;
 }
 enum class EEE : unsigned short {
   a = E6,
   b = EE::EE32, // expected-error {{not implicitly convertible}}

clang/test/Sema/integer-overflow.c

// expected-warning@+1 {{overflow in expression; result is 537919488 with type 'int'}}
   case (uint64_t)(4609 * 1024 * 1024):
     return 7;
 // expected-error@+1 {{expression is not an integer constant expression}}
   case ((uint64_t)(4608 * 1024 * 1024 * i)):
     return 8;
 // expected-warning@+1 {{overflow in expression; result is 536870912 with type 'int'}}
   case ((1ULL * ((4608) * ((1024) * (1024))) + 2ULL)):
     return 9;
+// expected-warning@+3 {{implicit conversion from}}
 // expected-warning@+2 2{{overflow in expression; result is 536870912 with type 'int'}}
 // expected-warning@+1 {{overflow converting case value to switch condition type (288230376151711744 to 0)}}
   case ((uint64_t)((uint64_t)(4608 * 1024 * 1024) * (uint64_t)(4608 * 1024 * 1024))):
     return 10;
   }
 
 // expected-warning@+1 {{overflow in expression; result is 536870912 with type 'int'}}
   while (4608 * 1024 * 1024);

clang/test/Sema/switch.c

 // RUN: %clang_cc1 -fsyntax-only -verify -Wswitch-enum -Wcovered-switch-default -triple x86_64-linux-gnu %s
 void f (int z) { 
   while (z) { 
     default: z--;            // expected-error {{statement not in switch}}
   } 
 }
 
 void foo(int X) {
   switch (X) {
   case 42: ;                 // expected-note {{previous case}}
-  case 5000000000LL:         // expected-warning {{overflow}}
+  case 5000000000LL:         // expected-warning {{overflow}} expected-warning {{implicit conversion}} 
   case 42:                   // expected-error {{duplicate case value '42'}}
    ;
 
   case 100 ... 99: ;         // expected-warning {{empty case range}}
 
   case 43: ;                 // expected-note {{previous case}}
   case 43 ... 45:  ;         // expected-error {{duplicate case value}}
 
void switch_on_ExtendedEnum1(enum ExtendedEnum1 e) {
   }
 }
 
 void PR11778(char c, int n, long long ll) {
   // Do not reject this; we don't have duplicate case values because we
   // check for duplicates in the promoted type.
   switch (c) case 1: case 257: ; // expected-warning {{overflow}}
 
-  switch (n) case 0x100000001LL: case 1: ; // expected-warning {{overflow}} expected-error {{duplicate}} expected-note {{previous}}
-  switch ((int)ll) case 0x100000001LL: case 1: ; // expected-warning {{overflow}} expected-error {{duplicate}} expected-note {{previous}}
+  switch (n) case 0x100000001LL: case 1: ; // expected-warning {{overflow}} expected-error {{duplicate}} expected-note {{previous}} expected-warning {{implicit conversion}}
+  switch ((int)ll) case 0x100000001LL: case 1: ; // expected-warning {{overflow}} expected-error {{duplicate}} expected-note {{previous}} expected-warning {{implicit conversion}}
   switch ((long long)n) case 0x100000001LL: case 1: ;
   switch (ll) case 0x100000001LL: case 1: ;
 }

clang/test/SemaCXX/builtin-is-constant-evaluated.cpp

 } // namespace test_ref_initialization
 
 #if defined(__cpp_conditional_explicit)
 struct TestConditionalExplicit {
   explicit(!__builtin_is_constant_evaluated()) TestConditionalExplicit(int) {}
 };
 TestConditionalExplicit e = 42;
 #endif
+
+constexpr int i1 = (long long)__builtin_is_constant_evaluated() * (1ll << 33);
+// expected-warning@-1 {{implicit conversion}}
+
+constexpr int i2 = (long long)!__builtin_is_constant_evaluated() * (1ll << 33);
+
+ void f(int i) {
+   switch (i) {
+   case (long long)__builtin_is_constant_evaluated() * (1ll << 33):;
+// expected-warning@-1 {{implicit conversion}}
+   }
+ }

clang/test/SemaCXX/constant-expression-cxx11.cpp

 static_assert(MangleChars("constexpr!") == 1768383, "");
 static_assert(MangleChars(u8"constexpr!") == 1768383, "");
 static_assert(MangleChars(L"constexpr!") == 1768383, "");
 static_assert(MangleChars(u"constexpr!") == 1768383, "");
 static_assert(MangleChars(U"constexpr!") == 1768383, "");
 
 constexpr char c0 = "nought index"[0];
 constexpr char c1 = "nice index"[10];
-constexpr char c2 = "nasty index"[12]; // expected-error {{must be initialized by a constant expression}} expected-warning {{is past the end}} expected-note {{read of dereferenced one-past-the-end pointer}}
-constexpr char c3 = "negative index"[-1]; // expected-error {{must be initialized by a constant expression}} expected-warning {{is before the beginning}} expected-note {{cannot refer to element -1 of array of 15 elements}}
+constexpr char c2 = "nasty index"[12]; // expected-error {{must be initialized by a constant expression}} expected-note {{read of dereferenced one-past-the-end pointer}}
+constexpr char c3 = "negative index"[-1]; // expected-error {{must be initialized by a constant expression}} expected-note {{cannot refer to element -1 of array of 15 elements}}
 constexpr char c4 = ((char*)(int*)"no reinterpret_casts allowed")[14]; // expected-error {{must be initialized by a constant expression}} expected-note {{cast that performs the conversions of a reinterpret_cast}}
 
 constexpr const char *p = "test" + 2;
 static_assert(*p == 's', "");
 
 constexpr const char *max_iter(const char *a, const char *b) {
   return *a < *b ? b : a;
 }

 
 constexpr const int *p = xs + 3;
 constexpr int xs4 = p[1]; // ok
 constexpr int xs5 = p[2]; // expected-error {{constant expression}} expected-note {{read of dereferenced one-past-the-end pointer}}
 constexpr int xs6 = p[3]; // expected-error {{constant expression}} expected-note {{cannot refer to element 6}}
 constexpr int xs0 = p[-3]; // ok
 constexpr int xs_1 = p[-4]; // expected-error {{constant expression}} expected-note {{cannot refer to element -1}}
 
-constexpr int zs[2][2][2][2] = { 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16 }; // expected-note {{array 'zs' declared here}}
+constexpr int zs[2][2][2][2] = { 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16 };
 static_assert(zs[0][0][0][0] == 1, "");
 static_assert(zs[1][1][1][1] == 16, "");
 static_assert(zs[0][0][0][2] == 3, ""); // expected-error {{constant expression}} expected-note {{read of dereferenced one-past-the-end pointer}}
 static_assert((&zs[0][0][0][2])[-1] == 2, "");
 static_assert(**(**(zs + 1) + 1) == 11, "");
 static_assert(*(&(&(*(*&(&zs[2] - 1)[0] + 2 - 2))[2])[-1][-1] + 1) == 11, ""); // expected-error {{constant expression}} expected-note {{cannot refer to element -1 of array of 2 elements in a constant expression}}
 static_assert(*(&(&(*(*&(&zs[2] - 1)[0] + 2 - 2))[2])[-1][2] - 2) == 11, "");
 constexpr int err_zs_1_2_0_0 = zs[1][2][0][0]; // \
 expected-error {{constant expression}} \
-expected-note {{cannot access array element of pointer past the end}} \
-expected-warning {{array index 2 is past the end of the array (which contains 2 elements)}}
+expected-note {{cannot access array element of pointer past the end}}
 
 constexpr int fail(const int &p) {
   return (&p)[64]; // expected-note {{cannot refer to element 64 of array of 2 elements}}
 }
 static_assert(fail(*(&(&(*(*&(&zs[2] - 1)[0] + 2 - 2))[2])[-1][2] - 2)) == 11, ""); // \
 expected-error {{static_assert expression is not an integral constant expression}} \
 expected-note {{in call to 'fail(zs[1][0][1][0])'}}
 

clang/test/SemaCXX/integer-overflow.cpp

case 1 + static_cast<uint64_t>(4609 * 1024 * 1024):
     return 7;
 // expected-error@+2 {{expression is not an integral constant expression}}
 // expected-note@+1 {{read of non-const variable 'i' is not allowed in a constant expression}}
   case ((uint64_t)(4608 * 1024 * 1024 * i)):
     return 8;
 // expected-warning@+1 {{overflow in expression; result is 536870912 with type 'int'}}
   case ((1ULL * ((4608) * ((1024) * (1024))) + 2ULL)):
     return 9;
+// expected-warning@+3 {{implicit conversion}}
 // expected-warning@+2 2{{overflow in expression; result is 536870912 with type 'int'}}
 // expected-warning@+1 {{overflow converting case value to switch condition type (288230376151711744 to 0)}}
   case ((uint64_t)((uint64_t)(4608 * 1024 * 1024) * (uint64_t)(4608 * 1024 * 1024))):
     return 10;
   }
 #endif
 
 // expected-warning@+1 {{overflow in expression; result is 536870912 with type 'int'}}