Re: List question

[Helmethead <hoshem mel comcen com au>]

On Thu, May 24, 2001 at 12:27:00PM -0600, Eric Lemings wrote:
> Hello all,
> 
> I was just looking at the source for glist.c and a thought occurred to
> me (which happens on occasion).  ;)  What if the client passes a pointer
> to the tail of the list, or any arbitrary node within the list, to some
> of these functions?  Would this break some functions, cause them to
> fail, or operate in unexpected ways?
> 
> For example,
> 
>     GList* node = NULL;
>     GList* head = NULL;
>     /* Add several nodes... */
>     GList* tail = g_list_last (head);
> 
>     /* Does this actually prepend to the head of the list?
>      */
>     node = g_list_prepend (tail, data);
> 
>     /* Is this the actual length of the list?
>      */
>     int len = g_list_length (node);
> 
> For efficiency reasons, I can see why the list functions may require
> that the 'list' parameter is the head of the list but, for sake of
> argument, why should some of these functions assume that?  Or maybe this
> policy should just be reiterated in the documentation somewhere.

Apparently, prepending to a list element inserts just before the element, correctly setting new_list->prev and list->prev->next. Only for doubly linked lists of course. What wacky behaviour!