operator new/delete forwarding

[Martin Sebor]

Thanks for the detailed history, Howard! I was not at the Batavia
meeting and I attended the joint EWG/LWG session the morning the
issue was being discussed at Oxford.

I wasn't aware of the use case given in N2158, or the requirement
for operator delete(void*) to be able to handle pointers obtained
from both the throwing as well as the nothrow overloads of operator
new (I just found it in [basic.stc.dynamic.deallocation], p3). It
seems to me that mixing and matching calls to a throwing new and
a nothrow delete (or vice versa) is at the root of the problem and
should not be allowed. Wouldn't that be an easier way to solve the
issue?

As for the pace we're at which we deal with issues, I have no
problem with it. I am, however, less than fully comfortable with
straw polls of a handful of people who just happen to be in the
room at the time determining the outcome of issues. I wish we
had a more accessible process in place, such as online polls.
Austin Group has a mechanism like that in place:
   http://www.opengroup.org/austin/interps/faq_polls.tpl
We could use one of the many online polling services or set up
our own.

Martin

Howard Hinnant wrote:
> On Apr 30, 2007, at 8:14 PM, Martin Sebor wrote:
> 
>> Howard Hinnant wrote:
>>> On Jul 26, 2006, Howard Hinnant wrote:
>>> (http://gcc.gnu.org/ml/libstdc++/2006-07/msg00157.html)
>>>> On Jul 25, 2006, at 8:35 PM, Gabriel Dos Reis wrote:
>>>>
>>>>> OK, you guys have changed my mind and convinced me we should proceed
>>>>> as Martin suggests.
>>>>> We'll meet at Portalnd and rubber-stamp Howard's suggestion if LWG
>>>>> feels OK :-)
>>>>> Howard, you know the Master of LWG issue list, right? :-)
>>>>
>>>> I'll interpret that as a request to reopen lwg 206. Consider it done:
>>>>
>>>> <note>
>>>> Reopened due to a gcc conversation between Howard, Martin and Gaby. 
>>>> Forwarding
>>>> or not is visible behavior to the client and it would be useful for 
>>>> the client
>>>> to know which behavior it could depend on.
>>>> </note>
>>> http://www.open-std.org/jtc1/sc22/wg21/docs/lwg-active.html#206
>>> was voted into the C++0X working draft on Friday April 20, 2007.
>>
>> I must have been asleep when we discussed this issue in Oxford.
>> The Rationale in the issue on your home page:
>> http://home.twcny.rr.com/hinnant/cpp_extensions/issues_preview/lwg-defects.html#206 
>>
>> still says that "[the operators] may become unlinked, and that
>> is by design. If a user replaces one, the user should also
>> replace the other." Doesn't the resolution go against that design?
>> If it does, what was the rationale for the change? And how did we
>> address Robert's compatibility concern? (Sorry to be asking these
>> questions now instead of when we talked about it.)
> 
> The issue was discussed in Batavia in January, and listed in Tentatively 
> Ready status at that point.  All of the issues from Batavia were given 
> visibility on the std-lib reflector on Jan 25 in message 
> c++std-lib-17919.  The Batavia issues were again highlighted in the 
> pre-Oxford mailing paper:
> 
>    http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2183.html
> 
> I agree we are moving at a frightening pace.  But it is my impression 
> that this pace is required to meet o 2009 deadline.
> 
> Yes, the resolution goes against the design that the no-throw and 
> throwing signatures become unlinked.  N2158 explains why the former 
> design is fragile and leads to client code crashing (providing rationale 
> for the change).
> 
> Robert expressed concern publicly in the Batavia meeting, and later to 
> me privately.  So I noted his concern in the issue (indeed I value his 
> opinion highly).  However there was no further concern or objection 
> noted between N2183 (pre-Oxford mailing Mar 9), and the full committee 
> vote on this issue Apr 20.
> 
> The Batavia issues were reviewed in Oxford on the morning of Tuesday, 
> Apr 17, however the LWG was split between doing this and meeting with 
> the EWG concerning threading issues (I attended the latter, and do not 
> know who attended which).  The part of the LWG that was reviewing the 
> Batavia issues Ok'd transitioning 206 from Tentatively Ready to Ready 
> (Ready for end of the week vote).
> 
> Robert's concerns were noted in Batavia but overruled by straw poll at 
> that time.  At no time since the Batavia meeting were concerns raised or 
> straw polls asked for to the best of my knowledge.  And repeated 
> visibility was given to this issue, as well as several other 
> controversial issues, in preparation for the Oxford meeting.  -- It's a 
> busy time right now with lots of critical/controversial decisions being 
> made.
> 
> -Howard
> 
>