Re: Speeding up v3 basic_string

[Paolo Carlini <pcarlini at unitus dot it>]

Hi Vladimir,

the problem you are reporting seems strange to me. We (me!) are always
regtesting before committing and we are not experiencing problems since my last
changes to _M_replace. That being said, of course I'm willing to investigate
your problem and I have a few last resorts in mind (i.e., reintroducing a
_M_replace(forward_iterator_tag) which for the time being would be identical to
_M_replace(input_iterator_tag), I'm talking about last resort because at this
point I cannot understand why that would be necessary.
So I propose two things:
1- Please try to your best to prepare a short, non-linking test case. Of course
you have to look among the parts of you code that use the string class, some
member functions should be triggering the problem.
2- Give me a few hours and I will test that "last resort" I mentioned before,
it's a short patch, which you could use while I will investigate in greater
detail the problem.

Thanks,
Paolo.