Created attachment 113926 [details] [review]
gtk_widget_tooltip_empty_is_null.patch

Comment on attachment 113926 [details] [review]
gtk_widget_tooltip_empty_is_null.patch

Thanks, the patch looks good Murray, please apply after the following coding style issues have been fixed.

>Index: gtk/gtkwidget.c
>===================================================================
>--- gtk/gtkwidget.c	(revision 20744)
>+++ gtk/gtkwidget.c	(working copy)
>@@ -2478,6 +2478,13 @@
>       tooltip_window = g_object_get_qdata (object, quark_tooltip_window);
>       tooltip_markup = g_value_dup_string (value);
> 
>+      /* Treat an empty string as a NULL string, 
>+         because an empty string would be useless for a tooltip: */

All comment lines are prefixed with '*' in Gtk+ coding style. also '*/' should go on its own line.

>+      if (tooltip_markup && (strlen (tooltip_markup) == 0)) {

Braces also go on their own line.
>+	g_free (tooltip_markup);
>+        tooltip_markup = NULL;
>+      }
>+
>       g_object_set_qdata_full (object, quark_tooltip_markup,
> 			       tooltip_markup, g_free);
> 
>@@ -2486,7 +2493,14 @@
>       break;
>     case PROP_TOOLTIP_TEXT:
>       tooltip_window = g_object_get_qdata (object, quark_tooltip_window);
>+
>       tooltip_text = g_value_get_string (value);
>+
>+      /* Treat an empty string as a NULL string, 
>+         because an empty string would be useless for a tooltip: */

dito.

>+      if (tooltip_text && (strlen (tooltip_text) == 0))
>+        tooltip_text = NULL;
>+
>       tooltip_markup = tooltip_text ? g_markup_escape_text (tooltip_text, -1) : NULL;
> 
>       g_object_set_qdata_full (object, quark_tooltip_markup,

Admittedly I have no clue about GTK's coding styles/rules, so sorry if I ask something stupid, but wouldn't it be faster to replace the "strlen"-calls in the patch with a check if the first char is a '\0', i.e.

  if (tooltip_markup && *tooltip_markup)  /* or (*tooltip_markup != '\0') */

and

  if (tooltip_text && *tooltip_text)

(In reply to comment #3)
> Admittedly I have no clue about GTK's coding styles/rules, so sorry if I ask
> something stupid, but wouldn't it be faster to replace the "strlen"-calls in
> the patch with a check if the first char is a '\0', i.e.
> 
>   if (tooltip_markup && *tooltip_markup)  /* or (*tooltip_markup != '\0') */
> 
> and
> 
>   if (tooltip_text && *tooltip_text)

Yes, it would be faster, but only a tiny bit and the code in question isn't performance critical in anyway. In general, readability should be the prime concern, personally I don't have a strong preference for one over the other here though.

Committed to trunk and and gtk-2-12 with Tim's code formatting changes.