From: Dave Abrahams (abrahams_at_[hidden])
Date: 2000-04-21 21:26:48
on 4/21/00 9:07 AM, Gary Powell at Gary.Powell_at_[hidden] wrote:
> I'd be glad to follow the std but its really broken in this case.
I agree that it is broken in some sense. I also think you can stay within
its bounds. I don't think inducing undefined behavior is appropriate for
libraries in boost (well, maybe only in libraries specifically designed to
test the limits of existing implementations).
> It's a commonly used container, I need the trait<hash_map>, SGI put it in
> the namespace std, Its not std,
> What would you like?
> a) A seperate include file that can be used or not by users of hash_map.
> b) use of defines to recognize that SGI STL is being used. And include the
> file <hash_map> ?
Either one of those sounds fine. <b> is probably the more seamless choice.
> This isn't important to me. I'm just trying to make this library as standard
> complient as possible. It's just frustrating.
Tell me about it. This should be easier, especially the swap stuff.
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk