[00:24] -queuebot:#ubuntu-release- New binary: fwupd [arm64] (focal-proposed/main) [1.3.5-1] (core)
[00:24] -queuebot:#ubuntu-release- New binary: meep [amd64] (focal-proposed/universe) [1.12.0-2] (no packageset)
[00:24] -queuebot:#ubuntu-release- New binary: gatb-core [amd64] (focal-proposed/universe) [1.4.1+git20191130.664696c+dfsg-1] (no packageset)
[00:27] -queuebot:#ubuntu-release- New binary: gatb-core [ppc64el] (focal-proposed/universe) [1.4.1+git20191130.664696c+dfsg-1] (no packageset)
[00:27] -queuebot:#ubuntu-release- New binary: meep [ppc64el] (focal-proposed/universe) [1.12.0-2] (no packageset)
[00:27] -queuebot:#ubuntu-release- New binary: gatb-core [s390x] (focal-proposed/universe) [1.4.1+git20191130.664696c+dfsg-1] (no packageset)
[00:27] -queuebot:#ubuntu-release- New binary: meep [s390x] (focal-proposed/universe) [1.12.0-2] (no packageset)
[00:30] -queuebot:#ubuntu-release- New: accepted fwupd [amd64] (focal-proposed) [1.3.5-1]
[00:30] -queuebot:#ubuntu-release- New: accepted fwupd [ppc64el] (focal-proposed) [1.3.5-1]
[00:30] -queuebot:#ubuntu-release- New: accepted opencolorio [amd64] (focal-proposed) [1.1.1~dfsg0-4]
[00:30] -queuebot:#ubuntu-release- New: accepted rust-gstreamer-base [arm64] (focal-proposed) [0.14.4-1]
[00:30] -queuebot:#ubuntu-release- New: accepted rust-gstreamer-base [ppc64el] (focal-proposed) [0.14.4-1]
[00:30] -queuebot:#ubuntu-release- New: accepted rust-sequoia-openpgp [amd64] (focal-proposed) [0.12.0-1]
[00:30] -queuebot:#ubuntu-release- New: accepted fwupd [arm64] (focal-proposed) [1.3.5-1]
[00:30] -queuebot:#ubuntu-release- New: accepted rust-gstreamer-base [amd64] (focal-proposed) [0.14.4-1]
[00:30] -queuebot:#ubuntu-release- New: accepted rust-gstreamer-base [s390x] (focal-proposed) [0.14.4-1]
[00:30] -queuebot:#ubuntu-release- New: accepted fwupd [s390x] (focal-proposed) [1.3.5-1]
[00:30] -queuebot:#ubuntu-release- New: accepted rust-gstreamer-base [armhf] (focal-proposed) [0.14.4-1]
[00:32] -queuebot:#ubuntu-release- New: accepted gatb-core [amd64] (focal-proposed) [1.4.1+git20191130.664696c+dfsg-1]
[00:32] -queuebot:#ubuntu-release- New: accepted gatb-core [s390x] (focal-proposed) [1.4.1+git20191130.664696c+dfsg-1]
[00:32] -queuebot:#ubuntu-release- New: accepted meep [ppc64el] (focal-proposed) [1.12.0-2]
[00:32] -queuebot:#ubuntu-release- New: accepted gatb-core [ppc64el] (focal-proposed) [1.4.1+git20191130.664696c+dfsg-1]
[00:32] -queuebot:#ubuntu-release- New: accepted meep [s390x] (focal-proposed) [1.12.0-2]
[00:32] -queuebot:#ubuntu-release- New: accepted meep [amd64] (focal-proposed) [1.12.0-2]
[03:21] -queuebot:#ubuntu-release- Unapproved: fwupd (focal-proposed/main) [1.3.5-1 => 1.3.5-1] (core)
[05:18] -queuebot:#ubuntu-release- New binary: liblouis [amd64] (focal-proposed/main) [3.12.0-2] (desktop-core, ubuntu-server)
[05:19] -queuebot:#ubuntu-release- New: accepted liblouis [amd64] (focal-proposed) [3.12.0-2]
[06:00] <vorlon> doko: do you happen to remember if there's a reason dh_python3 generates dependencies on python3:any, but python3.7 without the :any?
[16:33] -queuebot:#ubuntu-release- Unapproved: update-motd (eoan-proposed/main) [3.6-0ubuntu1 => 3.6-0ubuntu1.19.10.0] (core)
[19:46] <vorlon> infinity: we're left with a number of binary packages that are built from whitelisted source packages, and uninstallable on i386 because those binaries weren't seeded and they have deps on things not in the whitelist (e.g., audispd-plugins, bluez-cups, ...) do you have opinions on what to do with these?
[20:15] <infinity> vorlon: I find it odd that we're not seeding sources for the exercise, rather than binaries.
[20:15] <vorlon> does germinate have an option to seed by source?
[20:16] <infinity> vorlon: %source is a source entry in a seed (see supported-kernel-common, for instance)
[20:16] <vorlon> also I don't think it's to anyone's benefit to keep a bunch of extra source packages building on i386 just to make these packages "installable"
[20:19] <infinity> I think a bunch of uninstallable packages will lead to a distinct "lack of quality" feel to the whole thing.   I don't care so much about the solution (source deltas to not build those binaries, P-a-s to not publish them, or making them actually installable), but I don't think "a bunch of stuff in the archive we can't install" is great.
[20:19] <infinity> And yes, I know there are any number of arch:all packages that fit that description too (on many arches, but now a whole bunch on i386), and that's a stickier problem.
[20:20] <vorlon> infinity: right, I'm fine with us making the binaries go away, and yeah I was just going to make the same point re: arch:all
[20:20] <vorlon> although the difference with arch:all packages is that, if i386 is a foreign arch (the expected end-user use case), apt+dpkg alway treat that as a native-arch package
[20:20] <vorlon> so it will be installable
[20:21] <infinity> Making the binaries go away needs to be done "properly" (either by telling LP not to publish them in a P-a-s style feature that it doesn't have, or by altering the source), both of which are a bit icky.
[20:21] <vorlon> so, I'd definitely prefer introducing deltas to remove the binary packages rather than making them installable
[20:21] <vorlon> I'm not sure there's a way to do that in a way that's upstreamable to Debian... maybe some dh -p magic in debian/rules
[20:23] <infinity> Anyhow, having them uninstallable is something we can live with in the very short term, but it'll lead to stalled britney migrations without hinting, and then hinting will lead to unintended side-effects when britney decides that trading 3 uninst on s390x for 4 uninst on i386 is a good deal, or whatever.
[20:24] <infinity> (That said, my hatred of britney's uninst-trading feature is much lower these days now that we attempt to keep the uninst count close enough to zero that we can actually see what's happening)
[20:26] <infinity> vorlon: I am genuinely curious how much bigger the set would get if it was a full closure of all binaries generated from the sources we seed, though.  If it's only 5% or 10%, that seems worth the hassle to not have to think too hard, if it's 50%, maybe not.
[20:27] <vorlon> I didn't think britney traded uninstallables across architectures?  and also none of these should magically become installable
[20:28] <infinity> Fair point.
[20:28] <infinity> Still curious about how much the set would grow.
[20:28] <vorlon> infinity: tbh I'm still inclined to go the other direction and find more things to cut, for example gst-plugins-bad1.0 is explicitly seeded and that build depends on opencv which pulls in vtk and I want that to die
[20:28] <infinity> I suspect somehting like 'Extra-Include: *' would do the trick.
[20:30] <infinity> vorlon: I only have gst-x, gst-base, abd gst-good installed here on i386.  All of which I didn't do on purpose, so I assume those are "needed", and others maybe less so.
[20:30] <vorlon> infinity: wine in some cases goes looking for gst-plugins-bad
[20:30] <infinity> But, OTOH, if those are needed for something, then so is -bad as soon as that thing wants to play an MP3. :P
[20:30] <infinity> Or whateevr is in bad these days.
[20:34] <infinity> It occurs to me that the i386 Packages file could be much smaller if it wasn't for all the useless arch:all too.  I wonder if the whitelist feature could be extended to also not publish arch-all for non-whitelisted packages (and then we'd have to whitelist all the arch:all stuff we want, but that seems feasible)
[20:35]  * infinity wanders off again.
[20:35] -queuebot:#ubuntu-release- New binary: insighttoolkit4 [amd64] (focal-proposed/universe) [4.13.2-dfsg1-4ubuntu1] (no packageset)
[20:35] <vorlon> in practice the arch: all stuff we want is already included in the whitelist
[20:40] <vorlon>   samba (2:4.11.1+dfsg-3ubuntu1): libnss-winbind libpam-winbind python3-samba samba samba-common-bin samba-testsuite winbind
[20:40] <vorlon> wonder if we should've included nss modules in the whitelist explicitly
[20:51] <vorlon> why did ia32-libs depend on gvfs, that seems pointless
[20:51] <vorlon> hmm I guess that package provides the gio plugin that lets apps talk to gvfs.  So not actually pointless
[20:52] <vorlon> just another annoying dep tree that it pulls in as a result (leading to pacemaker ;)
[21:47] <vorlon> doko: britney thinks a large number of *-cross packages are NBS, but cron.NBS does not, and they're uninstallable.  Do you know what's needed there? https://people.canonical.com/~ubuntu-archive/proposed-migration/focal_uninst.txt
[23:08] <infinity> vorlon: cron.NBS fails miserably at finding stuff that's NBS on a subset of arches.
[23:09] <vorlon> infinity: so I should just manually nuke them?
[23:10] <infinity> Perhaps, yes.  Check outdate.
[23:10] <vorlon> "outdate"?
[23:11] <infinity> https://people.canonical.com/~ubuntu-archive/testing/focal_outdate.txt
[23:11] <infinity> Which doesn't seem to think anything's NBS...
[23:12] <infinity> I'm not seeing britney thinking things are NBS.  Am I blind?
[23:13] <vorlon> infinity: at the end of update_output it has a long list of foo-cross packages that it says it wants to remove...
[23:17] <infinity> Wow, now I'm confused.  uninst shows those are from 32ubuntu2, but the only version in the archive is 32ubuntu1...
[23:17] <infinity> Oh, 31ubuntu2.
[23:17] <infinity> Lysdexia.
[23:18] <infinity> So yeah, looks like all the dgb packages were dropped.  Why the outdate britney run isn't seeing that, I don't know.
[23:38] <vorlon> the trivialness of the fixes to make all of these packages cross-testable is really quite satisfying