incubator-stdcxx-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Martin Sebor <se...@roguewave.com>
Subject Re: "Pluggable Memory"
Date Tue, 06 Sep 2005 00:33:00 GMT
Lance Diduck wrote:
> To all: Here is an issue that may impact stdcxx.
> http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2005/n1850.pdf 

Odd. Adobe Reader on Linux seems to have trouble with the document
-- it says it can't find Helvetica and the text is unreadable. Xpdf
opened it fine, though.

I skimmed the paper but I suspect I'll need to re-read it and spend
some time playing with the proposed classes to get a better
understanding of it. I'll try to do that before the end of the month
(i.e., before the October WG21 meeting) and post my feedback here.

> This was just put up there, and written by a guy I work with Palbo
> Halpern,
> and is a proposal to overhaul the STL containers to take polymorphic
> allocators as the default, and furthermore to pass these allocators
> around
> from container to container.
> 
> I don't know how the library committee reviews these proposals.

I see N1850 on the agenda for our Thursday 10/6 meeting (of the WG21
at Mont Tremblant).

> However,
> a
> way we could promote Apache stdcxx is to offer a realistic solution,
> that
> operates with the Apache "Pluggable Memory" concept.
> (http://xml.apache.org/xerces-c/program-others.html and
> http://xml.apache.org/xalan-c/programming.html ) There is an STL
> solution
> that shouldn't require changes to the standard,

You mean the one that defines allocate() and deallocate() virtual?
That would be a detectable change and for some implementations also
a binary incompatible, change.

> and works for all
> vendors
> (at least the ones that take optional allocator arguments). Using the
> STL
> with polymorphic allocators makes some of the interface ambiguous
> however. A
> more usable solution, could be an adapter that would not assume that
> allocators compare equal, and enforces post-conditions (based on a
> policy)
> that th standard does not address.
> 
> I have a paper here
> http://www.lancediduck.com/papers/Cpp/StatefulSTL.pdf
> that could be the start of a doc on how to use Pluggable Memory with STL
> containers.

I briefly looked at the paper a few weeks back (Ravi forwarded me
a copy) but I'm not done reviewing it yet. I'll review it before
the meeting and get back to you.

Martin

Mime
View raw message