Re: Bogus (?) GtkDialog/GtkMessageDialog separator handling

[Christian Neumair <chris gnome-de org>]

Am Montag, den 11.10.2004, 10:42 -0400 schrieb Owen Taylor:
> On Sun, 2004-10-10 at 19:46 +0200, Christian Neumair wrote:
> > When browsing the GTK+ source code, I noted that GtkDialog contains a
> > private data struct with a ignore_separator bool, plus a pseudo-private
> > (protected in terms of java, through _gtk_dialog_set_ignore_separator)
> > setter API for that variable.
> > GtkMessageDialog sets ignore_separator to FALSE, calls the public
> > GtkDialog set_separator API and re-sets ignore_separator to TRUE. I
> > really don't see any point in doing so. Why should the public separator
> > property of the dialog be ignored and instead be overridden by the
> > message dialog property? Why does the message dialog class have the
> > "use_separator" property, while the dialog parent class already has a
> > "has_separator" property?
> 
> The ignore_separator flag is meant so that GtkMessageDialog can prevent
> people setting the has_separator flag on GtkMessageDialog, even though
> it's a subclass of GtkDialog.
> 
> It doesn't make sense to have different GtkMessageDialogs having
> separators or not depending only on application programmer whim. The
> reason that it was settable on GtkDialog was for adjusting to 
> dialog content; but all message dialogs have the same content.
> 
> Does that make the code make more sense?

Yes, it does. Thanks for exhausting this.
But it raises another question: Are dialog separators really useful? The
HIG [1] states:
"Do not use frames and separators to compensate for poor control layout
or alignnment."
Maybe that's 3.0 stuff, but if application developers *really* want that
separator thingie, couldn't they just create it and pack it into the
vbox?

regs,
 Chris

[1]
http://developer.gnome.org/projects/gup/hig/1.0/controls.html#controls-frames