On Sat, 15 Aug 2020 at 18:10:10 +0100, Simon McVittie wrote:
> I've reported it upstream and will see what happens. We should not change
> the 1.46.x ABI downstream until we have discussed the problem with upstream,
> because that will be storing up incompatibilities for us later, but it
> might make sense to go back to the new version in unstable for now, so
> that dependent packages like GTK don't get recompiled against the new ABI.
The set of libraries that reference PangoRenderer is manageably small, we
can binNMU them to comply with whichever ABI we end up with
(1.44.x-compatible or 1.46.0-compatible), and there don't seem to be
recursive dependencies on PangoRenderer (in particular it doesn't seem
to have any practical effect on the ABI that GTK provides to dependent
packages), so I'm going to see whether I get an answer from upstream on
which way they're going to go.
If I don't get an answer fairly soon, I'll do a +really version to revert
to 1.44.x, but I'd prefer to avoid that if I can.
Maintainers of gtk, cogl, mutter, muffin, ukwm, geda-gaf and lepton-eda:
to avoid spreading the ABI break around, please don't upload to unstable
until this is resolved, unless the upload is particularly important. If in
doubt, ask on this bug or in #debian-gnome on OFTC.
Thanks,
smcv
