[cfe-dev] Type::isIncompleteType()...

[Steve Naroff]

The spec seems to define completeness by the closing brace...

6.7.2.3p3 All declarations of structure, union, or enumerated types  
that have the same scope and
use the same tag declare the same type. The type is incomplete109)  
until the closing brace
of the list defining the content, and complete thereafter.

If so, changing the predicate doesn't sound correct (though it makes  
sense to me conceptually).

snaroff

On Feb 10, 2008, at 8:07 PM, Steve Naroff wrote:

>
> On Feb 10, 2008, at 7:13 PM, Ted Kremenek wrote:
>
>>
>> On Feb 10, 2008, at 6:35 PM, Steve Naroff wrote:
>>
>>> Folks,
>>>
>>> Do you think it makes sense for Type::isIncompleteType() to return
>>> true when the Tag definition is invalid?
>>
>> I guess so.  From the standard (6.2.5):
>>
>> "... incomplete types (types that describe objects but lack  
>> information needed to determine their sizes). "
>>
>> I guess it depends on what clients expect.  Another way to look at  
>> this is that a bad tag definition doesn't really type check at all;  
>> it's a bogus definition, and doesn't define any kind of type,  
>> including an incomplete one.  It's probably safer to have clients  
>> believe that a bad tag definition has an incomplete type, and then  
>> check *why* it is an incomplete type by interrogating the TagDecl.
>>
>
> I tend to agree with you. If we make this change, the code below  
> will result in 2 errors. Note that both EDG & GCC only produce 1  
> error (the first one). From my perspective, not flagging the 2nd  
> error is incorrect (though it could lead to some noisy diagnostics,  
> I guess).
>
> Neil, does your cfe produce 1 or 2 diagnostics for the following  
> test case?
>
> snaroff
>
> [steve-naroffs-imac:tools/clang/test] snaroff% cat xx.c
> struct S {
>   int a;
>   int foo();
> };
>
> struct S s;
>
> [steve-naroffs-imac:tools/clang/test] snaroff% ../../../Debug/bin/ 
> clang xx.c
> xx.c:3:7: error: field 'foo' declared as a function
>   int foo();
>       ^
> xx.c:6:10: error: variable has incomplete type 'struct S'
> struct S s;
>          ^
> 2 diagnostics generated.
>
>>>
>>>
>>> For example...
>>>
>>>   case Tagged:
>>>      // A tagged type (struct/union/enum/class) is incomplete if the
>>> decl is a
>>>      // forward declaration, but not a full definition (C99  
>>> 6.2.5p22).
>>> -    return !cast<TagType>(CanonicalType)->getDecl()- 
>>> >isDefinition();
>>> +    TagDecl *TD = cast<TagType>(CanonicalType)->getDecl();
>>> +    return !TD->isDefinition() || TD->isInvalidDecl();
>>>
>>> Since this predicate is fairly low-level, I wanted to get some
>>> feedback...
>>>
>>> snaroff
>>> _______________________________________________
>>> cfe-dev mailing list
>>> cfe-dev at cs.uiuc.edu
>>> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-dev
>>
>
> _______________________________________________
> cfe-dev mailing list
> cfe-dev at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-dev

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20080210/4b2ed0a5/attachment.html>