Re: [patch] ConfigureNotify events

[Kazunobu Kuriyama]

Alexander Malmberg wrote:

> Kazunobu Kuriyama wrote:
>
> > Alexander Malmberg wrote:
> > <snip>
> >
> > I tried this patch and found that the windows of GNUstep apps appear
> > slightly higher
> > than before.  I have little time to track it down now, so I don't know
> > the reason
> > exactly.
>
> Is this a one-time thing, or a gradual shift? If it's just a one-time
> offset, I wouldn't worry about it. The old code was a bit off, and this
> would mess up the coordinates of events at times. Thus, I'm not
> surprised that there's a slight shift to bring everything in line,
> although I'm not entirely sure what the direct cause is.
What I saw was this: every window of GNUstep apps were shifed upward about
15 to 20 pixels (just about the width of title bars) after applying the 
patch.
The shift was persistent: even after I had dragged a window, expecting 
NSUserDefault
would remember this, the shift was not rectified when I gave an app a 
next run.
So I thought the patch gave some upward bias to the location of the windows.

> > Skimming the patch, I feel it should use XQeuryTree() and XGetGeometry()
> > before
> >
> >
> > > +                int root_x, root_y;
> > > +                Window root_child;
> > > +                XTranslateCoordinates(dpy, xEvent.xconfigure.window,
> > > +                                      RootWindow(dpy, cWin->screen),
> > > +                                      0, 0,
> > > +                                      &root_x, &root_y,
> > > +                                      &root_child);
> > > +                x.origin.x = root_x;
> > > +                x.origin.y = root_y;
> > to get the window of the parent and its coordinates, and pass them to
> > XTranslateCoordinates() if it does need them, not the ones of the root
> > window.
>
> I don't understand what you're trying to accomplish by doing that. The
> only thing I can think of is that you want to get the coordinates of the
> window including decorations added by the window manager. However, those
> aren't the coordinates we want, and we have no way of getting them since
> we can't assume that all window managers will use any particular
> structure of windows (parents, siblings, etc.) for decorations.
Ah..., what I wanted to suggest was this: use the global coordinates
system of GNUstep (though I'm not sure if there is it).  Because the
shift was persistent and constant, I thought it should be related to the
coordinates system defined internally by GNUstep.  So I thought use of such
coordinates, instead of ones of the root window, might fix the problem.

> Did you have something else in mind with that change?
Apart from window decorations, I'd like to encourage your investigation
on this issue because the current GNUstep's DnD mechanism somehow looks
interfering with other non-GNUstep X applications' DnD, too.  (I also
expect it might fix a strange cursor change problem.)  So if I find any
problem relevant to the patch, I will definitely report it to the list
(with a technical idea if I could).  In fact I still use the patched
-back to see what happens with it, though it may differ from the purpose
you intended.

Regards,
- Kazunobu Kuriyama

P.S. I got your reply about 12 hours later after you sent it. Other
mails are arriving at me in a quite chaotic order now.  So I'm afraid
my relply looks as if I ignored the comments of other people.  If so,
please don't get me wrong.