Re: [HIG] GtkMessageDialog

[Gregory Merchan <merchan phys lsu edu>]

On Fri, Dec 06, 2002 at 10:59:47AM +0100, Murray Cumming wrote:
> On Thu, 2002-12-05 at 20:11, Gregory Merchan wrote:
> > On Thu, Dec 05, 2002 at 04:01:42PM +0100, Murray Cumming wrote:
> > > Does GtkMessageDialog comply with the HIG by default:
<snip>
> > It does not have a type for authentication alerts. See bug 65765 :
> >   http://bugzilla.gnome.org/show_bug.cgi?id=65765
> 
> It looks like that patch will be applied.

Cool.

> > The API does not allow for fields as an authentication alert requires.
> 
> I don't think GtkMessageDialog means to solve every problem. There is
> still GtkDialog for more complex dialogs.

Yeah, that seems reasonable.


> > GtkButtonsType has GTK_BUTTONS_CLOSE, GTK_BUTTONS_CANCEL, and
> > GTK_BUTTONS_YES_NO, none of which should be used.
> 
> Cancel should never be used? But it's used in the HIG's "reactor
> meltdown" example.

The button is fine, insofar as it's the second button. But GTK_BUTTONS_CANCEL
provides only the Cancel button; an alert (or any dialog) with only Cancel
is what should not be used.


> > I say the alert image should be in a column all to itself, but the HIG
> > doesn't. Doing so would also make leave GtkDialog an unsuitable parent
> > class.
> 
> > I would prefer to just see GtkMessageDialog removed and have a GtkAlert
> > class.
> 
> Are you aware of any efforts to reconcile the HIG and GtkMessageDialog,
> or to add something to libgnomeui for this? How can anyone reasonably be
> expected to comply with the HIG if it requires such effort.

I'm not aware of any, but my awareness here is limited to the gnome mailing
lists. As I understand it, the code should go into libegg for eventual
inclusion in Gtk+ proper and libgnomeui should be just desktop "glue" like
session management.

Both Seth and I had written our own GtkAlert classes and (iirc) offered them
for inclusion in Gtk+ proper. But this happened during a freeze and wasn't
revisted later, afaik.

HIG 1.0 has some errors which I hope no one complies with. An override
redirect alert would really suck; that's fixed in CVS.  Surprising to me
there have been people who've gone to (IMO) great lengths to comply.
Those great lengths shouldn't be necessary for compliance, and there's
work being done already to eliminate some of them.

One thing I'd really like to see I mentioned here:
  http://mail.gnome.org/archives/hig/2002-October/msg00018.html


Cheers,
Greg Merchan