[cfe-commits] [PATCH] Fix for reinterpret_cast (bug 11747)

[Aaron Ballman]

On Sun, Jan 22, 2012 at 8:58 AM, Eli Friedman <eli.friedman at gmail.com> wrote:
> On Sat, Jan 21, 2012 at 11:38 PM, Aaron Ballman <aaron at aaronballman.com> wrote:
>> On Thu, Jan 19, 2012 at 12:00 AM, Eli Friedman <eli.friedman at gmail.com> wrote:
>>> On Mon, Jan 16, 2012 at 8:38 PM, Aaron Ballman <aaron at aaronballman.com> wrote:
>>>> On Sun, Jan 15, 2012 at 9:05 PM, Eli Friedman <eli.friedman at gmail.com> wrote:
>>>>> On Sun, Jan 15, 2012 at 1:45 PM, Aaron Ballman <aaron at aaronballman.com> wrote:
>>>>>> This is a patch for fixing 11747, allowing reinterpret_cast to "cast"
>>>>>> where the source and destination types are the same as per
>>>>>> [expr.reinterpret.cast]p2.  The fix itself is fairly simple -- the
>>>>>> code was already there, it just needed to be moved up slightly since
>>>>>> the previous check was for source and dest being pointers.
>>>>>
>>>>> C++11 [expr.reinterpret.cast]p2 is as follows:
>>>>>
>>>>> The reinterpret_cast operator shall not cast away constness (5.2.11).
>>>>> An expression of integral, enumeration, pointer, or pointer-to-member
>>>>> type can be explicitly converted to its own type; such a cast yields
>>>>> the value of its operand.
>>>>
>>>> Attached is a new patch to hopefully better address this, as well as
>>>> some additional test cases.
>>>
>>> +    // C++ 5.2.10p2 has a note that mentions that, subject to all other
>>> +    // restrictions, a cast to the same type is allowed so long as it does not
>>> +    // cast away constness. The intent is not entirely clear here, since all
>>> +    // other paragraphs explicitly forbid casts to the same type.
>>> +    //
>>> +    // The only allowed types are: integral, enumeration, pointer, or
>>> +    // pointer-to-member types.
>>>
>>> This comment is really unclear... you should really explicitly state
>>> that the C++98 version is unclear, and C++11 clarifies this case.
>>
>> I've clarified now, thanks for pointing out the confusion!
>>
>>> +    } else if (SrcType->isPointerType() ||
>>> +      SrcType->isMemberPointerType()) {
>>>
>>> The original code allowed "DestType->isAnyPointerType() ||
>>> DestType->isBlockPointerType()".  I don't see any reason to restrict
>>> Objective-C or block pointers.
>>
>> Nor do I, I've added this in as well.
>>
>>> +      if (CastsAwayConstness(Self, SrcType, DestType, /*CheckCVR=*/!CStyle,
>>> +                           /*CheckObjCLifetime=*/CStyle)) {
>>>
>>> How could a cast where the source and destination types are identical
>>> cast away constness?
>>
>> From my testing, the equality test does not take cv qualifiers into
>> account.  Is there another way I should be testing for equality that
>> does?
>
> "==" should definitely be taking qualifiers into account.  Do you have
> a testcase?

It was in the patch -- but you can reproduce it yourself pretty easily:

void foo() {
  const int i = 0;
  (void)reinterpret_cast< int >( i );
}

Put a breakpoint in SemaCast.cpp on or around line 1629 after applying
the patch and you'll see that you break into the SrcType == DestType
block.

~Aaron