Re: Remove algo duplication

[Marc Glisse <marc dot glisse at inria dot fr>]

On Mon, 16 Sep 2013, Paolo Carlini wrote:

> Testcases? If we have solutions which would not get right some concrete 
> cases, we should really have testcases for those, in my opinion. Then we 
> should double check the standard about those, how much is QoI and how much 
> just plain conformance. Does C++11 eventually spell out more details about 
> proxy iterators? I think at some point it was a goal, but I'm not sure how 
> much went into the actual Standard. In principle we could even envisage two 
> different forwarding strategies, one for C++98, pretty good, not perfect, but 
> well within C++98 conformance. A better one for C++11, which would be able to 
> exploit the whole move semantics, "perfect" forwarding support we have got in 
> the language. The hope is that the whole predefined_ops.h would be anyway 
> decently compact and understandable.

I think having the simple lexicographical_compare forward to the predicate 
version with std::less<void>() is valid. All I am saying is that I'll be 
able to tell the difference if we do it, but that's just because some copy 
elision won't happen (in a contrived testcase), and compilers were never 
forced to do copy elision, so that's ok. It should even work in C++03 with 
an emulation of less<void> using 16 overloads (T cv1&, U cv2&).


By the way, I thought someone had contributed less<void> and the others 
already, but I can't see them. Did I imagine that patch?

--
Marc Glisse