r290906 - [Win64] Don't widen integer literal zero arguments to unprototyped function calls

[Reid Kleckner via cfe-commits]

Author: rnk
Date: Tue Jan  3 15:23:35 2017
New Revision: 290906

URL: http://llvm.org/viewvc/llvm-project?rev=290906&view=rev
Log:
[Win64] Don't widen integer literal zero arguments to unprototyped function calls

The special case to widen the integer literal zero when passed to
variadic function calls should only apply to variadic functions, not
unprototyped functions. This is consistent with what MSVC does. In this
test case, MSVC uses a 4-byte store to pass the 5th argument to 'kr' and
an 8-byte store to pass the zero to 'v':

  void v(int, ...);
  void kr();
  void f(void) {
    v(1, 2, 3, 4, 0);
    kr(1, 2, 3, 4, 0);
  }

Aaron Ballman discovered this issue in https://reviews.llvm.org/D28166

Modified:
    cfe/trunk/lib/CodeGen/CodeGenFunction.h
    cfe/trunk/test/CodeGen/variadic-null-win64.c

Modified: cfe/trunk/lib/CodeGen/CodeGenFunction.h
URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/CodeGen/CodeGenFunction.h?rev=290906&r1=290905&r2=290906&view=diff
==============================================================================
--- cfe/trunk/lib/CodeGen/CodeGenFunction.h (original)
+++ cfe/trunk/lib/CodeGen/CodeGenFunction.h Tue Jan  3 15:23:35 2017
@@ -3571,7 +3571,7 @@ public:
 
     // If we still have any arguments, emit them using the type of the argument.
     for (auto *A : llvm::make_range(Arg, ArgRange.end()))
-      ArgTypes.push_back(getVarArgType(A));
+      ArgTypes.push_back(CallArgTypeInfo ? getVarArgType(A) : A->getType());
 
     EmitCallArgs(Args, ArgTypes, ArgRange, CalleeDecl, ParamsToSkip, Order);
   }

Modified: cfe/trunk/test/CodeGen/variadic-null-win64.c
URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/test/CodeGen/variadic-null-win64.c?rev=290906&r1=290905&r2=290906&view=diff
==============================================================================
--- cfe/trunk/test/CodeGen/variadic-null-win64.c (original)
+++ cfe/trunk/test/CodeGen/variadic-null-win64.c Tue Jan  3 15:23:35 2017
@@ -3,15 +3,21 @@
 
 // Make it possible to pass NULL through variadic functions on platforms where
 // NULL has an integer type that is more narrow than a pointer. On such
-// platforms we widen null pointer constants to a pointer-sized integer.
+// platforms we widen null pointer constants passed to variadic functions to a
+// pointer-sized integer. We don't apply this special case to K&R-style
+// unprototyped functions, because MSVC doesn't either.
 
 #define NULL 0
 
 void v(const char *f, ...);
+void kr();
 void f(const char *f) {
   v(f, 1, 2, 3, NULL);
+  kr(f, 1, 2, 3, 0);
 }
 // WINDOWS: define void @f(i8* %f)
 // WINDOWS: call void (i8*, ...) @v(i8* {{.*}}, i32 1, i32 2, i32 3, i64 0)
+// WINDOWS: call void bitcast (void (...)* @kr to void (i8*, i32, i32, i32, i32)*)(i8* {{.*}}, i32 1, i32 2, i32 3, i32 0)
 // LINUX: define void @f(i8* %f)
 // LINUX: call void (i8*, ...) @v(i8* {{.*}}, i32 1, i32 2, i32 3, i32 0)
+// LINUX: call void (i8*, i32, i32, i32, i32, ...) bitcast (void (...)* @kr to void (i8*, i32, i32, i32, i32, ...)*)(i8* %1, i32 1, i32 2, i32 3, i32 0)