Re: C++: STL 3.2 (fwd)

[Joe Buck <jbuck at Synopsys dot COM>]

> | > > I second the request.  I don't want to again be in the position of again
> | > > being one major release behind on STL, especially since this is only a
> | > > one-day miss.  By being in sync with STL we can point people to the
> | > > excellent SGI web site for documentation.
> 
> As a response to a previous message, Gerald suggested we just update
> what EGCS used to have and leave out <valarray> and <limits>. That is
> probably a good idea. As far as I can tell SGI's <valarray> is just a
> replicata of vector<> so I can't recommand it for  what valarray is
> supposed to be doing.  

I should have made clear that this is my intention as well.  In
particular, we don't want the SGI <string> implementation; what we
do want is their bug fixes and implementation improvements.

(SGI's a big believer in reuse; their <string> is a replica of vector<char>).

> A week ago, I announced on the libstdc++-v3 mailing list that
> <valarray> was nearly complete and just need testing. Right now, I got
> _one_ reply. I'd suggest to incorporate it into egcs snasphots and
> later move it into egcs-1.2.x. Thoughts?

In cases where it can't break anything, I'm in favor of efforts to
back-port stuff from libstdc++-v3 to v2.