On Tue, 2018-11-13 at 18:44 +0100, Daniel Dupont wrote: > Hi, > > On Tue, 2018-11-13 at 09:19 -0800, Diane Trout wrote: > > I thought they weren't supposed to be changing things in stable. I > > don't suppose you could try the version in unstable? > > > > Though honestly I need to fix a couple of bugs to release 0.17. If > > I > > can get that working maybe I can make a stretch backport for it. > > Thank you for your message. > > First of all, I would like: > - to confirm my analysis > - to identify the reason for this bug > - to identify the possible consequences > > I have two production servers affected by this bug and I would > prefer > not to try unstable packages.
I wouldn't really call any version of dnssec-trigger stable. It's just that 0.13 was bug free enough to make it into stretch. It's not terribly surprising to me that a change in libnm might break the python bindings. We had to make some changes in 0.15 because of changes in libnm. I had two ideas, either try downgrading to gir1.2-networkmanager- 1.0/libnm-glib4 1.6.2-3+debu9u2? (That's the stable version as opposed to the version from stretch proposed updates, and it might require downgrading some other network manager parts as well. Alternatively it might work if we apply the 0007-use-libnm.patch to the 0.13 package. https://salsa.debian.org/dns-team/dnssec-trigger/blob/master/debian/patches/0007-use-libnm.patch As a test could you try these two short python programs, and let me know if either if them works? Which would look like: Manager running? True ---- nmclient_test.py ---- #!/usr/bin/python import gi gi.require_version('NMClient', '1.0') from gi.repository import NMClient client = NMClient.Client().new() print('Manager running? %s' % (client.get_manager_running(),)) --------- end ------------ --- nmtest.py ---- #!/usr/bin/python import gi gi.require_version('NM', '1.0') from gi.repository import NM client = NM.Client().new() print('Manager running? %s' % (client.get_nm_running(),)) ------ end -------- Diane