Re: [v3] 25_algo merge/flatten from libstdcxx_so_7

[Paolo Carlini <pcarlini at suse dot de>]

Benjamin Kosnik wrote:

> Another (IMHO better) approach would be to remove the
> dg-require-rvalrefs on the moveable.cc tests that actually work with
> your new functionality (looks like most just require std::move and or
> std::move_iterator...)
Unfortunately, that's not the case. Have a closer look... If I also add 
to the containers move constructor and assignment, which are almost 
trivial, things would slightly improve for those.

> and then adjust the test in
> testsuite/lib/libstdc++.exp:
>
>        puts $f "#include <iterator>"
>        puts $f "#include <utility>"
>        puts $f "using std::move;"
>        puts $f "using std::identity;"
>        puts $f "using std::forward;"
>        puts $f "using std::move_iterator;"
>        puts $f "using std::make_move_iterator;"
>
> to reflect something that is not implemented and will fail. Maybe
> pointer-to-rvalref member function in some container? Or some trait?
>  
Would mean checking each specific container and each specific algorithm. 
Frankly, I'm not sure anymore a libstdc++.exp test is appropriate in 
this case. Probably I will just add that std::move_iterator, and maybe 
the move constructors and assignement, temporarily xfail the remaining 
tests, and leave further improvements in this area to you...

Paolo.