I have another suggestion about memory management of ISO C++. I think
it is very IMPORTANT. And I hope it can be discussed in C++0x.

----

A proposal to cancel overriding the DEFAULT operator new/delete

I don't think we need to override the DEFAULT operator new/delete.
This makes your code doesn't work well together 3rdparty code (if it
also override them).

I suggest ISO C++ cancel this feature, just like we can't override
operator+ of all C types. That is:

1. You CAN NOT override these operators:

void* operator new(size_t size);
void operator delete(void* p);

2. You CAN override NON-DEFAULT operator new. Such as:

void* operator new(size_t size, my_allocator& alloc);

3. There is no need to override operator delete. If you need delete
objects in special way, just use:

alloc.destroy(obj);

That's all.

Thanks.

On Apr 30, 7:13 pm, xushiwei gmail.com> wrote:

Practically, there's no way this will happen, as lots of existing code
makes use of this feature. If you think there's a problem, put your
effort into finding a way of structuring the resolution of memory
requests to avoid the issues you're experiencing, and if other people
see the merits, you might get it into a later C++ standard. Simply
banning everything is about as practical as asking the UN to make it
illegal for 3rd party libraries to include thread-unsafe code, or
requiring all libraries to be in namespaces registered with a global
body....

Tony

xushiwei wrote:

Forget it, it won't happen. Even if the Standards' Committees were not
in the final stages of developing C++0x it would still not happen
because your proposal gratuitously breaks existing code.

On 5

On May 1, 8:02 am, xushiwei gmail.com> wrote:

xushiwei wrote:

http://xushiwei.com/cancel-overriding-the-default-operator-new-delete

* Motivation

I don't think we need to override the DEFAULT operator new/delete.
This makes your code doesn't work well together 3rdparty code (if it
also overrides them).

Why overriding the DEFAULT operator new/delete is bad?

Allowing to override the DEFAULT operator new/delete means the DEFAULT
operator new/delete are special operators. When we have more than one
implementations of an operator, the compiler/linker will report...

{ clc++m uses English. It would be nicer for other English-only-speaking
readers if the attribution line used English and the whole article used
an encoding such as US-ASCII, ISO-8859 or UTF-8 that more English-based
systems have support for, though none of these is a rejection reason.
-mod }

On 4月30日, 下午9时16分, Tony Delroy yahoo.co.uk> wrote:

Why overriding the DEFAULT operator new/delete is bad?

Allowing to override the DEFAULT operator new/delete means the DEFAULT
operator new/delete are special operators. When we have more than one
implementations of an operator, the compiler/linker will report...

If your proposal ever got off the ground, how do you expect the authors of
the 3rd party library to use the compiler to compile their version of
operator new/delete?
Because your change means the compiler vendor should ignore the 3rd party
library's operator new/delete and prefer its own version.

In any case, it does not work like this.
It is usually the case that the programmers versions of
functions/data/objects etc are preferred by the linker before searching
libraries for missing symbols.

xushiwei wrote:

I can say with 100%% certainty that this proposal will NOT be considered
by WG21 and J16 for C++0x. Even if you could persuade us that it has
enough merit to justify breaking existing code and the coding techniques
that rely on the facility it is years too late for consideration.
That means that you have plenty of time to persuade people that your
proposal has sufficient merit to justify consideration for the standard
after C++0x. Frankly I think you would be wasting your time but you are
welcome to try.