Hey,
Tomas Janousek <tomi@nomi.cz> (2020-04-13):
> Package: libpango-1.0-0
> Version: 1.44.7-3
> Severity: important
> Tags: upstream
>
> Since yesterday's apt full-upgrade of testing, font rendering in Gtk
> applications is very bad, kerning is broken, text is wider than
> necessary, it's very very unpleasant, I'd even say downright unuable
> without downgrading.
>
> What I'm seeing is exactly this:
> https://gitlab.gnome.org/GNOME/pango/issues/404#note_712911
> https://gitlab.gnome.org/GNOME/pango/issues/404#note_676284
> https://github.com/harfbuzz/harfbuzz/issues/1892#issuecomment-552877664
> so I won't include additional screenshots.
>
> I'm not sure what should Debian do provided upstream seems a bit
> dead... But breaking font rendering in Gtk apps across all Debian
> installations doesn't seem like a very good outcome either. :-(
>
> Any ideas?
>
> If anyone else is experiencing this, my workaround:
>
> /etc/apt/sources.list:
> deb http://snapshot.debian.org/archive/debian/20200401T000000Z/ testing main contrib non-free
>
> /etc/apt/preferences:
> Package: *pango*:*
> Pin: origin "snapshot.debian.org"
> Pin-Priority: 1100
I can confirm this in the Debian Installer as well (spotted early April
but only debugged right now).
Getting a little back in time and comparing what unstable used to be at
20200401T154401Z and at 20200405T084040Z, I'm seeing these changes in
the list of udebs used to build the netboot-gtk image:
-libcrypt1-udeb_1%3a4.4.15-1_amd64.udeb
+libcrypt1-udeb_1%3a4.4.16-1_amd64.udeb
-libdrm2-udeb_2.4.100-4_amd64.udeb
+libdrm2-udeb_2.4.101-1_amd64.udeb
-libpango1.0-udeb_1.42.4-8_amd64.udeb
+libpango1.0-udeb_1.44.7-3_amd64.udeb
-libudev1-udeb_245.2-1_amd64.udeb
+libudev1-udeb_245.4-2_amd64.udeb
-udev-udeb_245.2-1_amd64.udeb
+udev-udeb_245.4-2_amd64.udeb
Compare attached 1/2 screenshots to see the regression.
To make sure the pango update was responsible, I deb-reversion'ed its
udeb (taking the old udeb, faking a higher version so that it would be
preferred to the new udeb), and rebuilt the installer with unstable
from 20200405T084040Z, and I can confirm the regression disappears
(see 2+revert screenshot).
Cheers,
--
Cyril Brulebois (kibi@debian.org) <https://debamax.com/>
D-I release manager -- Release team member -- Freelance Consultant
Attachment:
1.png
Description: PNG image
Attachment:
2.png
Description: PNG image
Attachment:
2+revert.png
Description: PNG image
Attachment:
signature.asc
Description: PGP signature
