Clarification on Gcc's strict aliasing rules

[Francis Moreau]

Andrew Haley <aph@redhat.com> writes:

> On 11/12/2010 07:45 PM, Francis Moreau wrote:
>> "Segher Boessenkool" <segher@kernel.crashing.org> writes:
>> 
>>>> But you finally said
>
>>>>
>>>>     - you can not access it as int:
>>>>
>>>>         that object (t.i) does not have a stored value therefore it
>>>>         doesn't exist.
>>>
>>> (Your words, not mine -- and such sloppy wording gets you into trouble,
>>> the standard does not talk about any of this.  It is one way of looking
>>> at it though).
>> 
>> So what did you mean by this ?
>> 
>>    > could you tell me what the effective type of 't.i' object ?
>>    
>>    int, if you can say that object exists at all: it does not have a stored
>>    value.  The stored value of t is a double with value 3.0 .  You can
>>    take its address and access it via that as "double" (or "char"), or you
>>    can access it as the union it is.  You can not access it as "int".
>> 
>>>> This is what I understood from what you said, please correct me if I'm
>>>> wrong.
>>>>
>>>> However doing:
>>>>
>>>>      int i = t.i;
>>>>
>>>> is defined in C (as long as there's no trap representation) even if 't.i'
>>>> object has no stored value.
>>>
>>> Actually, I think this is a GCC extension, and I was mistaken to say it
>>> is valid C99 before.  Standard C allows you to read from t.d or t, but not
>>> t.i, after storing into t.d .
>> 
>> No.
>> 
>>    t.d = 3.0;
>>    i = t.i;
>> 
>> is well defined in C.
>> 
>> Again, what's ambiguous is the example given by the GCC man:
>> 
>>    int *ip;
>>    t.d = 3.0;
>>    ip = &t.i;
>>    return *ip;
>> 
>> which produces code that might or not work.
>> 
>> 6.5p7 lists this as a possible alias case and I can't find any rule in
>> the standard that could invalidate it.
>> 
>> So either GCC is not conformant in this regard or I'm missing something.
>
> It's worth looking  at http://www.open-std.org/jtc1/sc22/wg14/www/docs/dr_236.htm

IMHO, I believe the standard broken in this regard and therefore GCC is
not conformant but for good reason.

-- 
Francis