this looks like the system is putting a shadow around the invisible border we use for resizing.

we should not have invisible borders or shadows on Quartz: the system will do both for us in a consistent way.

The correct shadow around the window seems to be the NSView's native shadow.

I use this CSS as a workaround for 3.14 on 10.9:

GtkWindow {
    box-shadow: none;
}
.tooltip {
    border-radius: 0;
    padding: 0;
}
.tooltip.background {
    background-clip: border-box;
}

Created attachment 298995 [details] [review]
Quartz: Turn off csd

does this patch have the desired effect ?

Created attachment 299049 [details]
without-csd

>  if (GDK_IS_QUARTZ_DISPLAY (gtk_widget_get_display (widget))

Missing a ')'.


>does this patch have the desired effect ?

Well as you can see from the attachment it does avoid the issue, but I am not sure entirely disabling CSD is the ideal solution. Perhaps for now it is the best solution if nobody is going to improve the existing CSD support.

Created attachment 299054 [details] [review]
Quartz: Respect GDK_DECOR_BORDER

This will make it easier to make csd windows work well on OS X.

Does this patch make things better ?

(In reply to Matthias Clasen from comment #9)
> Does this patch make things better ?

Hmm, Upon testing master the shadow around the main CSD window is actually already fixed.
Though the other issues remain, shadow around popups, CSD window sizing, etc.

Your patch does not seem to change any visible behavior.

Created attachment 307695 [details] [review]
quartz: disable system shadows if a GdkWindow shadow is set.

this should fix the popup/menu/tooltip problems

Ting Ping, can you confirm that that this works for you as well ?

(In reply to Matthias Clasen from comment #12)
> Ting Ping, can you confirm that that this works for you as well ?

Yes patch seems to work fine, though very recently you broke the quartz backend in 6d04183c and 6ac07a2e.
Here is my patch to fix those, don't know if you want a full issue over it: http://fpaste.org/248385/97244914/

oh, can you push that fix, please ?

Review of attachment 307695 [details] [review]:

confirmed working, so please push

I guess this can be closed now