[cfe-commits] r103253 - in /cfe/trunk: lib/AST/ExprConstant.cpp test/Sema/init.c

[Eli Friedman]

On Fri, May 7, 2010 at 1:02 PM, John McCall <rjmccall at apple.com> wrote:
>
> On May 7, 2010, at 12:48 PM, Eli Friedman wrote:
>
>> On Fri, May 7, 2010 at 8:18 AM, Rafael Espindola
>> <rafael.espindola at gmail.com> wrote:
>>> Author: rafael
>>> Date: Fri May  7 10:18:43 2010
>>> New Revision: 103253
>>>
>>> URL: http://llvm.org/viewvc/llvm-project?rev=103253&view=rev
>>> Log:
>>> Fix PR4386 by implementing gcc's old behaviour (4.2) when initializing
>>> variables with a comparison of a function pointer with 0.
>>>
>>>
>>> Modified:
>>>    cfe/trunk/lib/AST/ExprConstant.cpp
>>>    cfe/trunk/test/Sema/init.c
>>>
>>> Modified: cfe/trunk/lib/AST/ExprConstant.cpp
>>> URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/AST/ExprConstant.cpp?rev=103253&r1=103252&r2=103253&view=diff
>>> ==============================================================================
>>> --- cfe/trunk/lib/AST/ExprConstant.cpp (original)
>>> +++ cfe/trunk/lib/AST/ExprConstant.cpp Fri May  7 10:18:43 2010
>>> @@ -70,9 +70,20 @@
>>>  //===----------------------------------------------------------------------===//
>>>
>>>  static bool EvalPointerValueAsBool(APValue& Value, bool& Result) {
>>> -  // FIXME: Is this accurate for all kinds of bases?  If not, what would
>>> -  // the check look like?
>>> -  Result = Value.getLValueBase() || !Value.getLValueOffset().isZero();
>>> +  const Expr* Base = Value.getLValueBase();
>>> +
>>> +  Result = Base || !Value.getLValueOffset().isZero();
>>> +
>>> +  const DeclRefExpr* DeclRef = dyn_cast<DeclRefExpr>(Base);
>>> +  if (!DeclRef)
>>> +    return true;
>>
>> Is it possible for Base to be null?
>
> Yes;  that's how we represent null pointer lvalues.

Okay, then there is in fact a potential crash here.

> (I hate this representation)

You can blame me, then... but I don't see how else you could do it.

-Eli