Re: Regarding the GTK+ bug pointed out by issue #3091

[Benoît Minisini <benoit.minisini@xxxxxxxxxxxxxxxx>]

Le 18/10/2025 à 19:44, Benoît Minisini a écrit :
> Hi,
>
> There is a big design bug in GTK+3.
>
> When you put a widget A inside a parent container widget B, and A is 
> clipped by B (for example its Y coordinate is negative), then only the 
> drawing of A is clipped on the screen.
>
> But the hidden part A is still active, and may prevent another widget C 
> at the same place to work correctly (depending on whether A is above or 
> below).
>
> In other words, there is draw clipping, but not event clipping!
>
> It's utterly stupid in my opinon, but hey it's GTK+.
>
> I found a workaround for the IDE form editor by ordering controls 
> differently. This workaround is available for the 3.21.0 version that 
> will be released as soon as possible.
>
> But it's not a long-term solution.
>
> As soon as 3.21.0 is out, I will commit a possible solution on the 
> master, but as this needs some tests, I can't put it in the stable 
> release just now.
>
> What is the solution? It is the 'GtkEventBox' widget, which is a 
> container that just uses its own X11 window for itself and all its 
> children.
>
> By using it on every container controls, the event clipping is ensured 
> by the X11 server, and so GTK+ can't mess things up.
>
> And now there is Wayland. No idea if GtkEventBox will work the same way 
> on it, we'll see. But Wayland is not my priority anymore, so...
>
> Sorry for the inconvenience.
>
> Regards,

I pushed the workaround I talked about on master to see what happens.

It should fix the clipping probelm for all containers inheriting 
UserContainer at the moment, i.e. everything except uncached DrawingArea 
and TabStrip.

I will see if something can done for DrawingArea and TabStrip later.

Regards,

--
Benoît Minisini.