GLib BoF

GTK+ developers are happy with it, as long as its behaviour is the same as what GTK+ does; Philip to verify, then merge

Update on GProperty

Some discussion about JS trampolining costs for animations in the shell
People seem OK with this idea
Who’s going to do the work? Can we guarantee it for 2.60?
Reevaluate what to do about property documentation (getter + setter + property documentation + blurb)
Some discussion of how to present this better in gtk-doc (list getter/setter prototypes below documentation from property doc block); bear in mind the need for _dup_ rather than _get_, etc.; does the getter return const or (transfer full)?

GError improvements:

https://gitlab.gnome.org/GNOME/glib/issues/14
Complex, lots of use cases
> Have requested people to add their use cases to https://gitlab.gnome.org/GNOME/glib/issues/14
Some discussion about ways of ABI-compatibly extending GError: probably need to have a new error domain which is 'extended' and overallocates the GError
Could we also have a way of getting a backtrace from where a GError was *set*?

Thumbnailing:

Long, rambly, off-topic discussion
If Bastien wants it, he can implement it
Thumbnail sizing, normal vs large, do we want symlinks for actual hard pixel sizes?

libicu has complex symbol versioning and is huge, and is a pain to build on Windows
Can’t find any compelling alternatives
Make sure we have a list of places to update UCD when it updates

Anything which people want to get upstream into GLib

g_clear_pointer() typeof() improvements; more typeof() improvements everywhere
Can we do something like g_clear_autoptr() which uses the autoptr-registered free function?
Some discussion about getting g_autoptr() support working on Windows; Nirbheek to take a look at it so that we can potentially use it for GTK+ 4.0? Evaluating clang vs clang-cl next month

Finalisation of objects from the thread which created them

Philip to take a look
Need to think about whether we marshal dispose() as well as finalize() — could this cause deadlocks?
Once it’s done in GLib, GTK+ needs to implement the vfunc in GtkWidget

Misc discussion about atomics:

Do we make higher atomic refcount safety guarantees than we actually need? Do the new acquire()/release() APIs match up with C11?
Fallback atomic refcounting code is not actually safe, according to Behdad

2.60 planning

Elementary have some utility functions which might be upstreamable into GLib (GDateTime comparisons, ISO8601 formatting)
Topological sort?