Re: inconsistency in naming functions...

[Paolo Molaro <lupus lettere unipd it>]

On 03/14/00 Antonio Campos wrote:
> Looking around gtklabel.h, I found that now gtk_label_set is deprecated
> in favour of gtk_label_set_text, but at the same time gtk_label_get is
> still being used when it should be gtk_label_get_text for a better
> naming consistency.
> So here it is the patch for having both gtk_label_set_text and
> gtk_label_get_text correctly. At the same time I have moved the
> compatibility defines:
> 
> #define gtk_label_set        gtk_label_set_text
> #define gtk_label_get        gtk_label_get_text

While getting rid sometime of the ugly gtk_label_set/get is fine,
the signature for gtk_label_get_text should be the saner:

const gchar* gtk_label_get_text (GtkLabel *label);

just as gtk_entry_get_text.

> But when after looking at the code,
> gtk_label_get_text(GTK_LABEL(label),&my_text) makes my_text point to the
> label that is in the internal structures of the widget, but doesn't
> first duplicate the label in the internal structures of the widget and
> then returns the duplication in my_text.
> I think this is annoying because is a little inconsistent. Besides that,
> suppose that the label widget is destroyed. Then you have a pointer in
> my_text to nowhere, and it's a possible "segmentation fault" problem in
> the future. (I know nothing about threads, but maybe this is even a
> better thing when using them).
> 
> My proprosal is making gtk_label_get_text first duplicate the label that
> it contains and then returning this duplication to the caller.
> I think it shouldn't break any code because I have tested the change and
> GNOME continued to work. Although we should advertise people about
> possible memory leaks if this change is someday done.

If you need a copy, you make a copy, with the above interface it's as simple
as:

g_strdup(gtk_label_get_text(label));

lupus