Re: gtk_widget_ensure_style() and gtk_widget_style_get()

["BJörn Lindqvist" <bjourne gmail com>]

2008/4/2 Christian Neumair <cneumair gnome org>:
> Today I experienced an odd lockup in Nautilus (cf. attachment for the
>  full backtrace). The innermost frames are:
>
>  #1  0xb6e783ee in bsearch () from /lib/tls/i686/cmov/libc.so.6
>  #2  0xb7c50047 in _gtk_style_peek_property_value ()
>  from /opt/gnome/lib/libgtk-x11-2.0.so.0
>  #3  0xb7d02dae in gtk_widget_style_get_valist ()
>  from /opt/gnome/lib/libgtk-x11-2.0.so.0
>  #4  0xb7d02fa2 in gtk_widget_style_get ()
>  from /opt/gnome/lib/libgtk-x11-2.0.so.0
>  #5  0xb7c0d1c6 in gtk_range_get_props ()
>  from /opt/gnome/lib/libgtk-x11-2.0.so.0
>
>  Investigating #5 gtk_range_get_props() revealed that it calls #4
>  gtk_widget_style_get() correctly.
>
>  However, does a widget always have a valid style associated? Reading
>  gtk_widget_init() suggests that. It also looks like before realizing or
>  calling gtk_widget_ensure_style(), that style is just the default style,
>  meaning that no user changes (i.e. to ~/.gtkrc-2.0) are applied.
>
>  On the other hand, reading the gtk_widget_ensure_style() docs suggests
>  that a widget is not guaranteed to have a style until it is realized.

Then it should be possible to construct some test case in which a
widget does not have a style. But that seems impossible, unless you
cheat and set widget->style = NULL which is cheating so it doesn't
count. gtk_widget_ensure_style() is probably wrong and can be
deprecated.


-- 
mvh Björn