[v3] get_temporary_buffer, redux

[Benjamin Kosnik <bkoz at redhat dot com>]

This patch has been dropped:
http://gcc.gnu.org/ml/libstdc++/2003-11/msg00169.html

The reasoning:

> I don't think this is conforming.  The standard doesn't require that it
> use malloc(), but I think that it does not allow it to use operator new.
> If these functions' semantics were affected by replacing operator new,
> it would have to say so.

I don't see this rationale present in the standard, or in any LWG
mailings. Several implemenations, in fact, do use new.

What gives? Here's an updated patch.

-benjamin

Attachment: p.20040129-1.patch
Description: Text document