Question about the "realize" signal

[Dan H <dunno stoptrick com>]

Hello,

I like to build my GUIs with Glade, but as Glade is quite limited in what it can do I usually write routines 
that "finalize" the GUI after the Glade stuff has been set up. For example, in a current project I have a 
table which is full of radiobuttons, but Glade always creates those with a label, and to remove that label I 
have to call gtk_container_remove() on each button in that table. Or in another dialog I like to use custom 
widgets, so I build the window in Glade using dummy widgets and then call a routine that finds and removes 
the dummies and replaces them with what I want.

Usually I place the "fixup" code in routines that connect to the "realize" signal for the containing widget 
(in the first example, for instance, the "realize" signal for the table that contains all the radiobuttons). 
This always works, but what got me thinking was: Is it safe? Can't the table be realized first without 
buttons, so my routine that iterates over all contained buttons finds nothing? Am I relying on some 
particular oder here in which Glade builds the GUI elements?

In short: What's the best way of doing what I need to do?

Thanks,
--D.