[PATCH] Fix libstdc++/7186

[Paolo Carlini]

Phil Edwards wrote:

>On Sat, Jul 06, 2002 at 12:40:00AM +0200, Paolo Carlini wrote:
>  
>
>>it turns out that the PR can be fixed exactly in the same way of 
>>libstdc++/6642. Of course, I have also verified that the behavior for 
>>reverse_iterator is not affected (perhaps we can consider adding a *neg 
>>test for reverse iterators, here and for 6642, reminding the open state 
>>of DR280). Tested i686-pc-linux-gnu.
>>    
>>
>Since bkoz is offline for two weeks and some-odd days, I'm going to say
>that this looks good to me.
>
Thanks Phil, indeed Benjamin asked to put in ASAP this kind of quite 
straightforward but potentially ABI-breaking changes.

>>+template <typename _Tp, typename _RefL, typename _PtrL,
>>+                        typename _RefR, typename _PtrR>
>>+inline typename _Deque_iterator<_Tp, _RefL, _PtrL>::difference_type
>>+operator-(const _Deque_iterator<_Tp, _RefL, _PtrL>& __x,
>>+         const _Deque_iterator<_Tp, _RefR, _PtrR>& __y)
>>+{
>>+  return _Deque_iterator<_Tp, _RefL, _PtrL>::difference_type
>>    
>>
>I'm surprised that there doesn't need to be another 'typename' here.
>
I will check this bit. Thanks!

>I haven't followed the state of these changes on the branch.  Is this
>something that should be queued for 3.1 after a while, or is the branch
>unprepared for this change?
>
I think this should wait for the completion of the ABI docs Benjamin is 
preparing. It seems to me that the changes affecting std::deque (6503, 
7186) may be safely moved to the branch without breaking the ABI in any 
sensible meaning, perhaps not so those affecting __normal_iterator (6642).

Ciao, Paolo.