Re: Alias analysis and C++. Was: Re: [BusyBox] Re: Long term code...

[Paolo Carlini <pcarlini at unitus dot it>]

Jan Hubicka wrote:

> > Seriously, Jan has already made some good progress (excellent!) and I'm
> > still hoping that this alias analysis issue with C++ will be satisfactorily
> > solved in time for 3.1: it seems to me that it is not just a matter of
>
> Don't rely too much on me in this case.  I've fixed the overhemly conservative
> predicate, but it don't help in your testcase, but it helps in others that
> do use arrays itself (not hid inside structure).

I see. And I would like to add that, as an interested user, I have always
appreciated your valuable and continuos contributions to the middle- and back-end
of the compiler (and your humility, of course)

> I understand briefly how the base classes are determined, but there are number
> of dificult details most probably behind my experinece in C++ frontend..
> Sadly :(

I think this could be a good occasion for me to begin studying in some detail the
C++ front-end. Could you possibly send me privately the patch you have already
prepared as a starting point for my investigations?

Thanks again,
Paolo.