[Patch] libstdc++/25288

[Paolo Carlini]

Howard Hinnant wrote:

> This part scares the hell out of me, especially without tests 
> confirming abi stability.  My ignorance is the source of my 
> nervousness, nothing more.  Otherwise this looks like a terrific 
> patch to me.  I'd appreciate more eyes on this, especially in the abi 
> department.

Yes, let's wait a bit more. Really, I think it's safe because nothing of
this is exported from the library, of course, we are touching only
internal implementation functions, and we are changing completely the
names. Frankly, if for some crazy reason we had to keep around
_M_fill_insert and _M_insert_dispatch, then I would rather prefer going
with my first try (out of line, of course).

Paolo.