Hi *, I'm concerned about the garbage collector moving an instance of an object to another location.

From D Specification (page 171):
“... compacting the remaining used memory by copying the allocated objects (called a
 copying collector)”

Suppose I do following:

alias extern(C) void function(void *context) callback;
extern(C) void xyz(callback handler, void *context);
class A
{
	void foo()
	{
		// some stuff goes here
	}
	static extern(C) void handler(void *context)
	{
		A a = cast(A)context;
		a.foo();
	}
	void test()
	{
		// call xyz - both handler address and this will be stored internally in the C-library
		xyz(&handler, cast(void*)this);
	}
}

Should I std.gc.addRoot to the address of the callback function in order to prevent the GC from moving the object? Furthermore, will the GC honor this root if it points to a member method of an instance of an object? And last question - is this one correct: std.gc.addRoot(cast(void*)this)?

If above is answered somewhere, please point me to it. I searched this forum, but to no avail.
Thanks,
-- teo

teo wrote:
> Hi *, I'm concerned about the garbage collector moving an instance of an object to another location.
> 
> From D Specification (page 171):
> “... compacting the remaining used memory by copying the allocated objects (called a
>  copying collector)”
> 

Page 171?

However, the current GC implementation is not a copying collector. If it ever becomes one, you should expect an API for "pinning" allocated memory (so that it won't be moved) to be added.

-- 
Kirk McDonald
http://kirkmcdonald.blogspot.com
Pyd: Connecting D and Python
http://pyd.dsource.org

Kirk McDonald Wrote:

> teo wrote:
> > Hi *, I'm concerned about the garbage collector moving an instance of an object to another location.
> > 
> > From D Specification (page 171):
> > “... compacting the remaining used memory by copying the allocated objects (called a
> >  copying collector)”
> > 
> 
> Page 171?
> 
> However, the current GC implementation is not a copying collector. If it ever becomes one, you should expect an API for "pinning" allocated memory (so that it won't be moved) to be added.
> 
That's why I thought about addRoot. Anyway thanks Kirk.