Re: [boost] [Boost-commit] svn:boost r67792 - in branches/release: boost/iterator

Subject: Re: [boost] [Boost-commit] svn:boost r67792 - in branches/release: boost/iterator
From: Michel MORIN (mimomorin_at_[hidden])
Date: 2011-01-19 09:18:23

Marshall Clow wrote:
> On Jan 17, 2011, at 5:16 AM, Michel MORIN wrote:
>> * Ticket #1427
>> I also reopened #1427 because I found a bug in the updated code.
>> transform_iterator should use 'result_of<UnaryFunc(Iterator::reference)>'
>> but the code uses 'result_of<UnaryFunc(Iterator::reference)>'.
>>
>> A patch is attached in #1427 which is extremely simple.
>> Is there a chance that this will be fixed in Boost 1.46?
>
> Michel -
>
> I've committed your patches to the trunk - thanks!
>
> I don't think that they will be in 1.46, since I need to wait
> for the tests to cycle (which takes a couple days), and the deadline
> for merging stuff to release is _today_.

Yeah, that was too last minute...

In my experiments, this bug is a "showstopper" for using this feature
(i.e. using
result_of to determine the return type of functors in transform_iterator).
This feature has been requested by a not so small number of people,
and so it might be risky to ship this in an incomplete implementation, IMHO.

But, anyway, it is too late. Assuming we cannot change the code
(either merging the new commitment or reverting the result_of support),
what we can do is to mention this deficiency in the release note.

(Currently, the release note says:
 http://svn.boost.org/svn/boost/website/public_html/beta/feed/history/boost_1_46_0.qbk

  * `transform_iterator` now uses `boost::result_of` to determine
     functor result type
     ([@https://svn.boost.org/trac/boost/ticket/1427 #1427])

Another choice is to delete this release note?)

Regards,
Michel