On Wed, Dec 2, 2009 at 12:55 PM, Anders Carlsson <[email protected]> wrote:
>
> On Dec 2, 2009, at 12:05 PM, Eli Friedman wrote:
>
>> On Wed, Dec 2, 2009 at 9:15 AM, Anders Carlsson <[email protected]> wrote:
>>> Author: andersca
>>> Date: Wed Dec  2 11:15:43 2009
>>> New Revision: 90327
>>>
>>> URL: http://llvm.org/viewvc/llvm-project?rev=90327&view=rev
>>> Log:
>>> In Sema, whenever we think that a function is going to cause a vtable to be 
>>> generated, we mark any virtual implicit member functions as referenced.
>>
>> Two questions:
>> 1. Does this mean r90283 should be reverted?
>
> It does, and I did that as part of the fix.
>
>> 2. Why just implicit virtual member functions?  Why not all virtual
>> member functions?  (This affects PR5557.)
>
> Because virtual member functions are already marked. (As for PR5557, I think 
> we should mimic what gcc does and just instantiate at the end of the 
> translation unit).

Well, we skip calling MarkDeclarationReferenced in some cases where we
end up with references, but I guess it's not too important in general
considering that the only effect of that for non-template non-implicit
member functions is that the Used bit gets set.

-Eli

_______________________________________________
cfe-commits mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits

Reply via email to