[cfe-dev] Failing test SemaCXX/underlying_type.cpp on Mingw/Msys

[Douglas Gregor]

On May 25, 2011, at 10:26 AM, Eli Friedman wrote:

> On Wed, May 25, 2011 at 10:00 AM, Matthieu Monrocq
> <matthieu.monrocq at gmail.com> wrote:
>> Hello,
>> 
>> SemaCXX/underlying_type.cpp fails on Mingw/Msys with the following message:
>> 
>>   Assertion failed: (PtrVal & ((1 << PtrTraits::NumLowBitsAvailable)-1)) ==
>> 0 && "Pointer is not sufficiently aligned", file
>> d:/clang/llvm/include/llvm/ADT/PointerIntPair.h, line 80
>> 
>> From this function:
>> 
>>   void setPointer(PointerTy Ptr) {
>>     intptr_t PtrVal
>>       = reinterpret_cast<intptr_t>(PtrTraits::getAsVoidPointer(Ptr));
>>     assert((PtrVal & ((1 << PtrTraits::NumLowBitsAvailable)-1)) == 0 &&
>>            "Pointer is not sufficiently aligned");
>>     // Preserve all low bits, just update the pointer.
>>     Value = PtrVal | (Value & ~PointerBitMask);
>>   }
>> 
>> I don't know whether the problem only occurs on mingw/msys or also occurs on
>> a "bare" windows system.
> 
> Also occurs on bare Windows, at least according to some Apple-internal
> buildbots. A backtrace would be helpful.
> 
> This is most likely caused by r132017; Sean, any ideas?


The problem is here:

+  UnaryTransformType *Ty =
+    new UnaryTransformType (BaseType, UnderlyingType, Kind,
+                            UnderlyingType->isDependentType() ?
+                              QualType() : UnderlyingType);

I believe I've fixed the issue in r132064.

	- Doug