Re: [OE-core] [PATCH] gtk4: upgrade 4.22.5 -> 4.24.0

[Richard Purdie <richard.purdie@linuxfoundation.org>]

On Thu, 2026-10-01 at 20:50 +0200, Mathieu Dubois-Briand via
lists.openembedded.org wrote:
> On Thu Oct 1, 2026 at 1:54 PM CEST, Ross Burton via
> lists.openembedded.org wrote:
> > Hightlights from the upstream release notes:
> > 
> > * GtkSvgWidget is a widget to display SVG, with some support for
> >   interactive aspects of SVG
> > 
> > * Wayland:
> >  - Version 3.2 of the text protocol is supported
> >  - xdg-session-management-v1 is used
> > 
> > * GSK:
> >  - A number of 10bit image formats are supported
> >  - We are now using clip masks to reduce the number of offscreens
> > 
> > * The gtk4-encode-symbolic tool for producing symbolic png icons
> > has
> >   been dropped.
> > 
> > * Snapping: GTK can now snap contents to device pixel boundaries,
> >   for better control of rendering with fractional scales
> > 
> > * GTK now respects the reduced motion setting in various places
> >   and avoids animations (GtkSpinner, GtkStack, GtkRevealer)
> > 
> > * New GtkShortcutTrigger functions it easier to handle synonymous
> >   keys (in particular keypad keys).
> > 
> > * Type-to-search (in particular interaction with input methods)
> >   has been redone, switching from a "pull" to a "push" approach.
> > 
> > * The first part of ongoing frameclock work has landed, improving
> >   our ability to deliver frames on time
> > 
> > * GTK supports the ext-background-effect Wayland protocol to allow
> >   blurred background behind windows
> > 
> > * The visible focus timeout can be changed with the new
> >   gtk-keyboard-focus-visible-timeout setting
> > 
> > * The GDK frameclock has been reworked to improve precision and
> > reduce
> >   frame drops
> > 
> > Add a hook to rewrite the content of the meson-generated unity
> > sources,
> > as these contain absolute build paths.  Manipulate the file during
> > the
> > packaging process so that the actual build tree isn't altered,
> > meaning
> > rebuilds are not affected.
> > 
> > Based on work by Markus Volk <f_l_k@t-online.de>.
> > 
> > Signed-off-by: Ross Burton <ross.burton@arm.com>
> > ---
> 
> Hi Ross,
> 
> We might have some reproducibility issue here:
> 
> https://autobuilder.yoctoproject.org/valkyrie/#/builders/37/builds/5098
> https://valkyrie.yocto.io/pub/repro-fail/oe-reproducible-20261001-l7m_wa7_/packages/diff-html/

I think I know the answer to the gtk4 mystery. The find/sed runs on
packages-split and it breaks the hardlink between package/ and
packages-split. rpm builds off package/, ipk/deb build off packages-
split. That means only rpm is seeing the "wrong" file and this is why
the reproducibility test only reports rpm, not ipk and deb.

When we debug, we tend to go to the deb/ipk files since they're easier
to see inside than rpm...

Cheers,

Richard