Re: [v3] Add missing forward_list<>::splice_after and merge overloads fix splice_after taking a range

[Paolo Carlini <paolo dot carlini at oracle dot com>]

Hi,
> Hi,
>
> Nico noticed in private email that we weren't implementing LWG 1310, 
> thus I proceede to add the required overloads, pretty trivial work 
> (suited for 4.7.1 too).
>
> But then, when I got to the testcases I noticed that really almost 
> nothing worked as expected for the splice_after overload taking a 
> range: memory leaks easily noticeable for existing testcases too; also 
> showing up as "disappearing" elements of the destination forward_list; 
> in some cases the end of the range improperly handled (thus, as , 
> last], instead of last), as must always be). Hey, we had a complexity 
> much lower than the expected O(distance(first, last)) per C++11 ;) 
> Thus I'm fixing this rather serious issue too. Have a look, tomorrow I 
> mean to commit to mainline.
>
> Tested x86_64-linux, normal and debug (and a lot of valgrind too).
I'm going with something a bit more invasive, cleaning up a bit the _M_* 
helpers.

Francois, can you please review the debug-mode checks for cases like:

    std::forward_list<int> fl1(1), fl2(1);
    fl1.splice_after(fl1.before_begin(), fl2, fl2.before_begin(), 
fl2.begin());

?

I don't think we should error out. We don't for things like:

    std::list<int> fl1(1), fl2(1);
    fl1.splice(fl1.begin(), fl2, fl2.begin(), fl2.begin());

ie, the source is in both cases just an empty range, not an invalid range.

Thanks,
Paolo.

/////////////////////

Attachment: CL_nico_3
Description: Text document

Attachment: patch_nico_3
Description: Text document