|
Boost : |
From: David Abrahams (dave_at_[hidden])
Date: 2003-08-27 15:16:43
"E. Gladyshev" <egladysh_at_[hidden]> writes:
> BTW:
> I'd a bit suprised if the C++ committe
> accepts Boost memory management concept
> (or a complete lack of such) as
> an industry standard.
Too late (sort of)! Have you looked at TC1?
If you want the standard to tackle this problem for embedded users, I
suggest you come up with a comprehensive vision of:
a. how it would work throughout the library (e.g. iostreams and the
adaptive algorithms and basic_stringbuf have no allocator
parameters)
and
b. how it can be used in composite components (e.g. you build a new
component out of maps, strings, and shared_ptrs - how is that
customized?)
and present it to the committee. You could also present it here as a
first step and convince library authors such as Peter that supporting
the embedded market in this particular way is worth the costs incurred
by his library interface.
Another thing to keep in mind: most of the standard library functions
intentionally do not specify whether they are going to allocate
memory or not. It's perfectly legal and legitimate for any of them
to do so, using any means available to them. I believe that solving
the sort of problem you're having *in the standard text* is a much
bigger problem that you surmise.
-- Dave Abrahams Boost Consulting www.boost-consulting.com
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk