From: md2828 AT mclink DOT it (Marco Salvalaggio)
To: djgpp AT delorie DOT com
Cc: frenchc AT cadvision DOT com
Subject: Re: Large `new' request crashes: help!
Date: Wed, 26 Feb 1997 23:38:56 GMT
Message-ID: <331fc97c.12394538@newmail.mclink.it>
References: <330ED369 DOT 4742 AT cssa DOT org DOT za> <5eqksi$15gq AT elmo DOT cadvision DOT com>
In-Reply-To: <5eqksi$15gq@elmo.cadvision.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: quoted-printable

On Sun, 23 Feb 1997 12:23:42 GMT, you wrote:

>
>Don't have an exacty answer for you, but Borlands compiler let you set
>a callback function that would get called whenever new failed. The
>function was called set_new_handler(<whatever>) I think. If new fails
>it should still return zero, so you could set_new_handler to a dummy
>function (or there is probably a better solution) and just test for a
>zero...
>
>- Calvin -
>
set_new_handler() is not a Borland extension but a function that appears =
in
the standard C++ draft. Here's its definition (from the 25 April 1995
draft):

-------------------------------------------------------------------------=
------------------
18.4.2.3 set_new_handler

new_handler set_new_handler(new_handler new_p);

Effects: Establishes the function designated by new_p as the current
              new_handler.

Returns: the previous new_handler.
-------------------------------------------------------------------------=
------------------

Where new_handler is defined as follow:
                    =20
                       typedef void (*new_handler)();

Djgpp supports it, you'll have to include <std/new.h>
A better way to trap a new failure should be catch the standard defined
bad_alloc exception, but this one is not present in djgpp.

Marco.