Re: Patch to fix a hildon-1 bug

["Michael Dominic K." <mdk mdk am>]

On 7/25/07, Zhu, Peter J <peter j zhu intel com> wrote:
> Hi,
>
> Here I send a patch at end of this mail to fix a hildon-1 bug that will
> cause program crash as hildon window is destroied in some
> cases(Specially the it returns FALSE in the delete-event processing of
> that windo).
>
> Basically, the GtkBorder widget should be freed with gtk_border_free()
> rather than using g_free if it's NOT created by g_malloc.
>
> diff --git a/src/hildon-window.c b/src/hildon-window.c
> index 1f1fac6..6e67057 100644
> --- a/src/hildon-window.c
> +++ b/src/hildon-window.c
> @@ -380,9 +380,9 @@ hildon_window_finalize
> (GObject * o
>
>      self = HILDON_WINDOW (obj_self);
>
> -    g_free (priv->borders);
> -    g_free (priv->toolbar_borders);
> -
> +    gtk_border_free(priv->borders);
> +    gtk_border_free(priv->toolbar_borders);
> +
>      if (G_OBJECT_CLASS (parent_class)->finalize)
>          G_OBJECT_CLASS (parent_class)->finalize (obj_self);
>
> @@ -488,6 +488,8 @@ hildon_window_get_borders
> (HildonWindo
>
>      g_free (priv->borders);
>      g_free (priv->toolbar_borders);
> +    priv->borders = NULL;
> +    priv->toolbar_borders = NULL;
>
>      gtk_widget_style_get (GTK_WIDGET (window),
> "borders",&priv->borders,
>              "toolbar-borders", &priv->toolbar_borders,

Thanks for spotting the problem and the patch. This is not the first
time we run into this kind of issues (mismatched allocation/freeing
types). I personally think it's a bit API fault, which doesn't provide
gtk_border_new (similiar case -- GdkColor structure that has an _free
but not _new).


--
Michael.