Re: why no gtk_tree_iter_new()?

[James Henstridge <james daa com au>]

Murray Cumming wrote:

> I notice that functions such as gtk_tree_store_new() take pointers to
> allocated, but unset, GtkTreeIters, so in gtk-demo/main.c you have to do
> this:
>
> GtkTreeIter iter;
> ...
> gtk_tree_store_append (GTK_TREE_STORE (model), &iter, NULL);
>
>
> Isn't there a risk that someone could end up calling
> gtk_tree_iter_free() on this. And if there is a gtk_tree_iter(),
> shouldn't there be a matching gtk_tree_iter_new(), which could also zero
> the struct memory to avoid undefined behaviour.
The tree iter filled in by gtk_tree_store_append is owned by the caller. 
It is responsible for freeing it (if necessary).  Unlike iters 
allocated on the heap, you can't really transfer ownership of an iter 
allocated on the stack.  This is no more dangerous than the possibility 
of a program calling free() on an object it allocated on the stack.

If you are wondering about how to wrap the function, in python all 
functions that logically create GtkTextIters or GtkTreeIters, I changed 
them to return the iter, rather than taking a reference to it as an 
argument.  This may be appropriate for the C++ binding as well.

James.

--
Email: james daa com au
WWW:   http://www.daa.com.au/~james/