Calling builtin new

[Shayne Wissler]

I've overloaded the global new operator, so that I can, detect when I've run
out of memory:

extern void* operator new(size_t s) {
void *r = malloc(s);
if (!r && s) {
fprintf(stderr, "Error: No more memory.");
exit(1);
}
return r;
}

The problem here is that I don't really want to call malloc (and--Valgrind
is quite noisy about me doing this); I want to call the implementation of
new that I overrode. Suggestions?


Thanks,
Shayne Wissler

 

[=?ISO-8859-1?Q?Le_G=E9ant_Vert?=]

I've overloaded the global new operator, so that I can, detect when I've run
out of memory:

extern void* operator new(size_t s) {
void *r = malloc(s);
if (!r && s) {
fprintf(stderr, "Error: No more memory.");
exit(1);
}
return r;
}

The problem here is that I don't really want to call malloc (and--Valgrind
is quite noisy about me doing this); I want to call the implementation of
new that I overrode. Suggestions?


Thanks,
Shayne Wissler

I've done quite the same thing to detect memory leaks due to mismatching
new/delete ; what I've done overload operator new with some more
parameters : the size, the filename of the source, and the line (to
display where memory leak occurs).
if you do as I've done (to display where "out of memory" occurs for
example), you'll have something like :

void *operator new(size_t, char *, unsigned long);

and you won't have any problem calling the original :perator new(size_t)

.... my solution... if anyone has a better idea...

 

[Thomas Matthews]

I've overloaded the global new operator, so that I can, detect when I've run
out of memory:

My understanding is that the "new" operator throws an exception
when there are memory problems. Perhaps you could catch the
exception to detect when you've run out of memory.

extern void* operator new(size_t s) {
void *r = malloc(s);
if (!r && s) {
fprintf(stderr, "Error: No more memory.");
exit(1);
}
return r;
}

The problem here is that I don't really want to call malloc (and--Valgrind
is quite noisy about me doing this); I want to call the implementation of
new that I overrode. Suggestions?

Here I don't understand. If you overload the global new operator,
then have your function call the global new operator, isn't this
called recursion (or perhaps an infinite loop)?

Thanks,
Shayne Wissler

--
Thomas Matthews

C++ newsgroup welcome message:
http://www.slack.net/~shiva/welcome.txt
C++ Faq: http://www.parashift.com/c++-faq-lite
C Faq: http://www.eskimo.com/~scs/c-faq/top.html
alt.comp.lang.learn.c-c++ faq:
http://www.raos.demon.uk/acllc-c++/faq.html
Other sites:
http://www.josuttis.com -- C++ STL Library book

 

[Shayne Wissler]

My understanding is that the "new" operator throws an exception
when there are memory problems. Perhaps you could catch the
exception to detect when you've run out of memory.

It does indeed, thanks. However, I caught with catch(...) when I tested
this, is there a standard exception I should be catching instead of this
catch-all?

Here I don't understand. If you overload the global new operator,
then have your function call the global new operator, isn't this
called recursion (or perhaps an infinite loop)?

What I meant was that I want to call the builtin-new, the one I had
overridden. Just like when I override a virtual method, I can still call the
base class's version of that method. How do I refer to the "base" version of
the new operator?


Shayne Wissler

 

[Rolf Magnus]

It does indeed, thanks. However, I caught with catch(...) when I
tested this, is there a standard exception I should be catching
instead of this catch-all?

yes, std::bad_alloc.

What I meant was that I want to call the builtin-new, the one I had
overridden. Just like when I override a virtual method, I can still
call the base class's version of that method.

That's a member of another class though.

How do I refer to the "base" version of the new operator?

I don't know if that's possible at all.

 

[Shayne Wissler]

yes, std::bad_alloc.
Thanks.


That's a member of another class though.

The principle is the same.

I don't know if that's possible at all.

One ought to be able to do such a thing.


Shayne Wissler

 

[Rolf Magnus]

The principle is the same.

Not really. You want to call a built-in function from a user-defined
function that replaces it, which is very different from the overloading
example, where you just call a user-defined fuction from another
user-defined function.
Other example: If you implement your own operator=(), how would you call
the builtin operator=() for the same class from that? I'd assume there
is none in this case (after all - you replaced it), so it can't be
called.

One ought to be able to do such a thing.

As I said, I don't know.

 

[lilburne]

My understanding is that the "new" operator throws an exception
when there are memory problems. Perhaps you could catch the
exception to detect when you've run out of memory.

Why not simply do a:

void myFunctionToCallWhenOutOfMemory();

set_new_handler(myFunctionToCallWhenOutOfMemory);

Section 18.4.2.2 and 18.4.2.3. Then you get to find out when
memory is exhausted, before exceptions are thrown, and have
the opportunity to free up some memory (page out to
secondary storage). If a handler is set two attempts are
made to obtain memory the exception is only thrown if the
second attempt fails.

 

[Shayne Wissler]

Not really. You want to call a built-in function from a user-defined
function that replaces it, which is very different from the overloading
example, where you just call a user-defined fuction from another
user-defined function.

I said the principle was the same, not the scenario.


Shayne Wissler

 

[Rolf Magnus]

I said the principle was the same, not the scenario.

And my point is that the principle is not at all the same. One is simple
calling a regular function, the other one is calling a built-in that
was replaced.

 

[Jerry Coffin]

I've overloaded the global new operator, so that I can, detect when I've run
out of memory:

There's no need to overload operator new to handle this -- the library
provides set_new_handler for precisely this purpose.

The problem here is that I don't really want to call malloc (and--Valgrind
is quite noisy about me doing this); I want to call the implementation of
new that I overrode. Suggestions?

Yes: don't even try this -- it's doomed from the beginning.

From the viewpoint of the compiler, you've overridden a function which
is fine and well. From the viewpoint of the linker (at least on the
typical implementations that use linkers) this operator satisfies all
external references in the program to the symbol with whatever name
global operator new mangles to with your compiler. Since the references
to the name have been satisfied, the linker will NOT include the one
that's in the standard library.

IOW, you can't call it because in your program it doesn't even exist.