[01:06] -queuebot:#ubuntu-release- New binary: haskell-http-link-header [amd64] (artful-proposed/universe) [1.0.3-2] (no packageset)
[01:06] -queuebot:#ubuntu-release- New binary: haskell-http-link-header [ppc64el] (artful-proposed/universe) [1.0.3-2] (no packageset)
[01:06] -queuebot:#ubuntu-release- New binary: haskell-http-link-header [i386] (artful-proposed/universe) [1.0.3-2] (no packageset)
[01:09] -queuebot:#ubuntu-release- New binary: haskell-http-link-header [arm64] (artful-proposed/universe) [1.0.3-2] (no packageset)
[03:29] -queuebot:#ubuntu-release- New: accepted haskell-http-link-header [arm64] (artful-proposed) [1.0.3-2]
[03:29] -queuebot:#ubuntu-release- New: accepted haskell-http-link-header [ppc64el] (artful-proposed) [1.0.3-2]
[03:29] -queuebot:#ubuntu-release- New: accepted haskell-http-link-header [i386] (artful-proposed) [1.0.3-2]
[03:29] -queuebot:#ubuntu-release- New: accepted haskell-cabal-install [s390x] (artful-proposed) [1.24.0.2-2]
[03:29] -queuebot:#ubuntu-release- New: accepted haskell-http-link-header [amd64] (artful-proposed) [1.0.3-2]
[03:45] <slangasek> jbicha: are you tracking the mapnik that you synced from experimental?  It has missing symbols relative to the previous version and regresses node-mapnik
[03:46] <slangasek> (and probably more; but node-mapnik has the autopkgtest that notices)
[04:07] -queuebot:#ubuntu-release- New binary: haskell-github [amd64] (artful-proposed/universe) [0.15.0-1build1] (no packageset)
[04:07] -queuebot:#ubuntu-release- New binary: haskell-github [ppc64el] (artful-proposed/universe) [0.15.0-1build1] (no packageset)
[04:07] -queuebot:#ubuntu-release- New binary: haskell-github [i386] (artful-proposed/universe) [0.15.0-1build1] (no packageset)
[04:32] -queuebot:#ubuntu-release- New: accepted haskell-github [amd64] (artful-proposed) [0.15.0-1build1]
[04:32] -queuebot:#ubuntu-release- New: accepted haskell-github [ppc64el] (artful-proposed) [0.15.0-1build1]
[04:32] -queuebot:#ubuntu-release- New: accepted haskell-github [i386] (artful-proposed) [0.15.0-1build1]
[08:03] <Laney> slangasek: I do it frequently
[08:03] <Laney> But not at the weekend ......
[08:03] <Laney> slangasek: Are you interested in helping out, or something? You seem to be keeping an eye on this.
[09:05] <LocutusOfBorg> hello, did anybody stop haskell autosync? btw slangasek great success :)
[09:08] -queuebot:#ubuntu-release- New binary: paste [amd64] (artful-proposed/main) [2.0.3+dfsg-4ubuntu1] (ubuntu-desktop, ubuntu-server)
[09:21] <LocutusOfBorg> btw why is ninja failing in that strange way? I would blame debhelper or similar...
[09:22] <LocutusOfBorg> src:ninja-build, seems that debhelper is missing some target
[09:23] <cjwatson> I see no evidence that haskell autosyncing is stopped
[09:26] <LocutusOfBorg> I had to manually sinc ghc some minutes ago
[09:33] <Laney> LocutusOfBorg: what happens if you rebuild in unstable now?
[09:33] <Laney> 10.5 is there
[09:35] <LocutusOfBorg> Laney, I finished 10 secoonds ago
[09:35] <LocutusOfBorg> damn, same issue
[09:35] <LocutusOfBorg> adding an empty override_dh_auto_install works, meh
[09:35] <LocutusOfBorg> I'm opening an RC bug in debian
[09:36] <LocutusOfBorg> or maybe I'll open it if you ask me, because it might be a debhelper bug
[09:38] <Laney> seems likely that it is
[09:38] -queuebot:#ubuntu-release- New binary: linux-signed-lts-xenial [amd64] (trusty-proposed/main) [4.4.0-82.105~14.04.1] (kernel)
[09:38] <LocutusOfBorg> I'll ask in debian devel irc
[09:39] <cjwatson> LocutusOfBorg: ghc> you're just impatient
[09:40] <cjwatson> LocutusOfBorg: by about an hour or so
[09:40] <LocutusOfBorg> ok, it was uploaded yesterday... anyway, even better
[09:44] <cjwatson> LocutusOfBorg: didn't get imported into LP last night because of losing a race by a couple of minutes, which delayed it by six hours
[09:44] <LocutusOfBorg> oh ok
[10:07] -queuebot:#ubuntu-release- Unapproved: at-spi2-atk (xenial-proposed/main) [2.18.1-2ubuntu1 => 2.24.1-0ubuntu1] (kubuntu, ubuntu-desktop)
[10:09] -queuebot:#ubuntu-release- Unapproved: rejected at-spi2-atk [source] (xenial-proposed) [2.24.1-0ubuntu1]
[10:15]  * cjwatson finally gets round to proposing a crontab merge to fix said race.  Only been meaning to do that for about six year
[10:15] <cjwatson> s
[10:32] <LocutusOfBorg> cjwatson, <3
[10:32] <LocutusOfBorg> lovely cron
[10:52] -queuebot:#ubuntu-release- New: accepted linux-signed-lts-xenial [amd64] (trusty-proposed) [4.4.0-82.105~14.04.1]
[11:35] <jbicha> slangasek: yes I believe we just need a newer node-mapnik version, we can do that ourselves or wait a bit longer for Debian
[11:51] -queuebot:#ubuntu-release- Unapproved: libinput (xenial-proposed/main) [1.2.3-1ubuntu1 => 1.6.3-1ubuntu1~16.04.1] (kubuntu, ubuntu-desktop)
[12:12] -queuebot:#ubuntu-release- Unapproved: xorg-server (xenial-proposed/main) [2:1.18.4-0ubuntu0.2 => 2:1.18.4-0ubuntu0.3] (desktop-core, xorg)
[12:18] -queuebot:#ubuntu-release- Unapproved: xfonts-utils (xenial-proposed/main) [1:7.7+3ubuntu0.16.04.1 => 1:7.7+3ubuntu0.16.04.2] (core)
[12:57] -queuebot:#ubuntu-release- Unapproved: mesa (xenial-proposed/main) [12.0.6-0ubuntu0.16.04.1 => 17.0.7-0ubuntu0.16.04.1] (core, xorg)
[13:40] -queuebot:#ubuntu-release- New: accepted paste [amd64] (artful-proposed) [2.0.3+dfsg-4ubuntu1]
[14:21] -queuebot:#ubuntu-release- New binary: rustc [amd64] (artful-proposed/universe) [1.17.0+dfsg2-7] (no packageset)
[14:21] -queuebot:#ubuntu-release- New binary: rustc [s390x] (artful-proposed/universe) [1.17.0+dfsg2-7] (no packageset)
[14:27] -queuebot:#ubuntu-release- New binary: rustc [i386] (artful-proposed/universe) [1.17.0+dfsg2-7] (no packageset)
[14:36] -queuebot:#ubuntu-release- Unapproved: accepted curtin [source] (xenial-proposed) [0.1.0~bzr505-0ubuntu1~16.04.1]
[15:33] -queuebot:#ubuntu-release- Unapproved: intel-microcode (xenial-proposed/restricted) [3.20151106.1 => 3.20170511.1~ubuntu16.04.0] (ubuntu-desktop, ubuntu-server)
[15:34] -queuebot:#ubuntu-release- Unapproved: intel-microcode (yakkety-proposed/restricted) [3.20160714.1 => 3.20170511.1~ubuntu16.10.0] (ubuntu-desktop, ubuntu-server)
[15:34] -queuebot:#ubuntu-release- Unapproved: intel-microcode (zesty-proposed/restricted) [3.20161104.1 => 3.20170511.1~ubuntu17.04.0] (ubuntu-desktop, ubuntu-server)
[15:46] <rbasak> infinity: xnox: ^ who's reviewing these? Want me to do it? I don't see sil2100 here.
[15:50] -queuebot:#ubuntu-release- New binary: rustc [arm64] (artful-proposed/universe) [1.17.0+dfsg2-7] (no packageset)
[15:50] <xnox> rbasak, i review no, i applaud only =)
[15:53] <rbasak> xnox: huh?
[15:53] <xnox> rbasak, i'm just the uploader, and i have not pinged anybody to review it yet.
[15:54]  * xnox translates ranglish to english
[15:54] <rbasak> Ah
[15:54] <rbasak> OK. Well then I'll take it.
[15:58] <rbasak> xnox: not doing Trusty?
[16:00] <xnox> rbasak, nope.
[16:01] <xnox> rbasak, these new microcodes require early initramfs loading code; which is not in trusty. And I don't think intel-microcode is widely installed in trusty.
[16:02] <xnox> rbasak, in xenial we have started automated intel-microcode installation with ubuntu-drivers
[16:02] <rbasak> I see. Thanks.
[16:04] <rbasak> xnox: in the past the intel-microcode SRU updated the dat file only. With your backport it has the effect of changing more, such as debhelper compat level and the initramfs-tools hook. Do you have an opinion on one approach over the other?
[16:11] -queuebot:#ubuntu-release- New: accepted rustc [arm64] (artful-proposed) [1.17.0+dfsg2-7]
[16:11] -queuebot:#ubuntu-release- New: accepted rustc [amd64] (artful-proposed) [1.17.0+dfsg2-7]
[16:11] -queuebot:#ubuntu-release- New: accepted rustc [s390x] (artful-proposed) [1.17.0+dfsg2-7]
[16:11] -queuebot:#ubuntu-release- New: accepted rustc [i386] (artful-proposed) [1.17.0+dfsg2-7]
[16:11] <xnox> rbasak, i'd rather do wholesome backports. the changes are improve text output during initramfs generation. Pre-v4.4 handling of kernels is not as relevant on xenial, unless people boot trusty kernel (still)
[16:12] <rbasak> xnox: that's not general SRU policy though :-/
[16:14] <slangasek> Laney: well, I was working on transitions over the weekend, and the number of stuck tests that I was running across was significant enough that I batch retried, and 800 seemed a bit high.  Do you have some idea of the failure rate here, that would inform whether it's worth putting effort into the cause vs. just batch retrying on the regular?
[16:15] <xnox> rbasak, this is hwe backport of binary blobs for restricted.
[16:15] <rbasak> xnox: it is; however the packaging is not a binary blob.
[16:15] <slangasek> Laney: but also, if you need someone else to hit the mass-retry button, I can certainly do that ;)
[16:15] <Laney> slangasek: do you know what the problem was?
[16:15] <xnox> rbasak, e.g. are nvidia packages backported piecewise, or just straight backports of the new upstream releases (and the new packaging)?
[16:15] <slangasek> no clue
[16:15] <Laney> it's not regular, at least not on that scale
[16:15] <Laney> but there could be a systemic problem that hits at one time
[16:15] <rbasak> xnox: I don't know. Is it?
[16:15] <Laney> something like: the initial dist-upgrade fails
[16:16] <xnox> rbasak, there is more room for error by modifying this; since one has to be careful and exclude packaging of the broken microcodes.
[16:16] <rbasak> I can see that a wholesale packaging update might be necessary depending on the nature of the upstream blob update (eg. if old packaging cannot cope with it).
[16:16] <ypwong> infinity, can you help us to review the xenial sru for fwupdate?
[16:17] <slangasek> Laney: well, there had also been several big packages going through the system (nodejs, perl, ...), so it's possible 800 was in line with the overall error rate?  But I don't know how to even measure that error rate
[16:17] <rbasak> In this case, isn't the packaging more closely tied to the Free bits, rather than the non-Free bits? The blob itself remains a blob all the way through to the hardware, no?
[16:18] <xnox> zesty diff is smaller, since zesty base is newer thus that one looks a lot more like just the blob update
[16:21] <rbasak> That statement sounds likely to be true, but I don't follow what point you're making there. I was reviewing the Xenial diff to start with.
[16:21] <xnox> that's backwards no? =)
[16:21] <xnox> yakkety all diffs are needed.
[16:21] <Laney> slangasek: We have the information needed (exit code of autopkgtest) in a database, so it could be exposed on http://autopkgtest.ubuntu.com/statistics
[16:21] <Laney> 800 at once is definitely more than I'd expect
[16:21] <Laney> I suspect a package was failing to install or similar
[16:21] <rbasak> I find it easier to start with the most invasive set of decisions. Then reviewing the smaller diffs is easy :)
[16:22] <Laney> I pushed a change earlier that should result in more runs being recorded as failures (with version 'unknown') rather than being left in running state
[16:22] <slangasek> Laney: you have an error code for these test requests that end up stuck in 'running'?  They don't seem to ever show up on the per-package pages
[16:23] <Laney> It's because they fail too early for autopkgtest-web to know the current version, so it skips them
[16:23] <slangasek> ah
[16:23] <Laney> that's what my fix remedies
[16:23] <xnox> rbasak, the initramfs hook updates are good / wanted in xenial; update from debhelper compat 7->9 is imho also good as we do want microcode package to build exactly the same way on all releases. And ideally we do want to update it on continious basis.
[16:23] <Laney> well, it fakes up a version of 'unknown'
[16:23] <xnox> not taking compat changes, creates more work and divergence
[16:23] <slangasek> yeah, that seems like a good improvement :)
[16:24] <Laney> so you should at least easily be able to browse them rather than having to surf swift via its API
[16:24]  * xnox ponders if intel-microcode really should be a snap
[16:24] <slangasek> makes the state more discoverable
[16:24] <rbasak> xnox: perhaps so, but in that case should we not have those changes go through the usual SRU process with their own bugs, justifications, regression tests, etc?
[16:25] <rbasak> xnox: and for the future "exactly the same way" / "update it on continious basis", I agree this might be what we want in general, but I think that should probably form an SRU exception that we want to formulate, review, approve, etc.
[16:26] <xnox> rbasak, perhaps, but all of them are "backport the latest micocode package to stable releases" which is what the bug report in question is.
[16:26] <xnox> rbasak, the bug here is not "cherrypick high-severety microcode updates only for Skylake"
[16:26] <rbasak> OTOH this particular update we should probably just get through, and for that, I think minimising it to the blob update makes more sense, unless there's a particular reason why that can't be done or there's a specific reason it carries higher risk.
[16:27] <xnox> my request here is for a wholesome backport to latest, not a targeted fix for just this one skylake HT bug.
[16:27] <cjwatson> *wholesale
[16:27] <rbasak> xnox: OK, but if you want to drive the more general bug, it'll end up in the slow lane.
[16:27] <xnox> cjwatson, yes, checked dictionary.
[16:28] <xnox> rbasak, it definately should be in the slow lane, with extra long phasing. I am expecting it will be in proposed for a month of so.
[16:28] <xnox> rbasak, as we need to collect multiple possitive results on multiple CPU SKUs / Families.
[16:29] <rbasak> I agree with extra long phasing and a long time in proposed. I was expecting though to be able to fast track a blob update (only) into proposed so users had somewhere official from which to get it (for opt-in updates).
[16:29] <xnox> there are a lot of updates to a lot of microcodes, and any one of them can regress anything (to the point of graphics not working, or systems not booting)
[16:30] <xnox> rbasak, that would diverge packages and create sru-only delta which i do not want to support, nor maintain, nor debug going forward.
[16:30] <xnox> (delta vs debian)
[16:31] <xnox> imho it is a mistake that we do not routinely update microcode packages, the way we e.g. update the hwe-kernels.
[16:31] <rbasak> It's been done before. I'm not ruling out your longer term plans at all here. If those are successful, then any difficult maintainability point becomes moot.
[16:32] <xnox> rbasak, given lack of dependencies it would be nice to copy down the .deb verbantim. but we will not do that.
[16:33] <xnox> rbasak, please review yakkety/zesty, i think you will have less objections about those.
[16:33] <xnox> rbasak, and i prefer for later releases be fix released, before even accepting xenial package.
[16:34] <xnox> unless you object to s/7/9 in those too
[16:40] <rbasak> AFAICT, either I need to accept the packaging changes on the basis that they are approved under current SRU policy (perhaps with approving whatever I need to that is within my remit), and so I can accept X/Y/Z subject to review this way, or I don't. I don't think it makes sense to accept Y/Z and not X on a policy basis. And if I did, but we later decided to do something different, that'd be even
[16:40] <rbasak> more of a pain to manage in the future.
[16:40] <rbasak> So I think ~ubuntu-sru needs to resolve what to do here first, if you want to include packaging changes in this update.
[16:41] <rbasak> OTOH I'd accept a blob-only update into all three releases as I don't think there's any SRU policy question there, but you've said and given reasons as to why you don't want that.
[16:41] <xnox> rbasak, right, my thinking behind this upload was that this is backport-published-to-updates-via-proposed-testing
[16:42] <rbasak> That's a summary of my thinking right now. I'd like opinions from other ~ubuntu-sru if any are around please.
[16:42] <xnox> i did not prepare these as classic SRUs.
[16:42] <xnox> and backports minimise delta / changes, to only those neccessory to get something building on older release; no changes are required here, as debian too backports this package as far back as is reasonable.
[16:44] <rbasak> The regression risk there though is for example that the delivery mechanism changes to something that older kernels cannot support.
[16:45] <rbasak> OTOH, it seems unlikely to me that just updating the blobs would pose any difficulty for a system on which the delivery mechanism already works (and if it doesn't, that's a matter for a more specific SRU following the usual process).
[16:45] <xnox> the risk of buggy microcodes / microcodes with regressions is much higher. The last delivery change was in the v3.13 kernels, and has not changed since.
[16:45] <rbasak> "the risk of buggy microcodes / microcodes with regressions is much higher" - how so?
[16:46] <xnox> (the requirement around built-in / non-builtin module is imho insignificant)
[16:46] <rbasak> "The last delivery change was in the v3.13 kernels" - only because you specifically know that. What might you not know about that is in the same class of problems?
[16:46] <xnox> rbasak, re:buggy microcode the infamous lock ellision update
[16:46] <xnox> that needed newer glibc on the host too, to prevent it from using the now disabled lock ellision, resulting in crashes.
[16:47] <xnox> e.g. one needed newer userspace glibc and the new microcode.
[16:47] <xnox> that was non-obvious during the microcode update.
[16:47] <xnox> because of lack of disclosures from intel
[16:47] <rbasak> xnox: and how would backporting the latest packaging have avoided that, if that packaging didn't have the versioned dependency because we didn't know about it?
[16:48] <xnox> it doesn't, due to lack of information. Therefore bisecting and cherrypicking individual microcode updates and/or cherrypicking changes does not reduce risk at all. The only way we can reduce risk is extensive in-the-field testing on multiple of microcode families and watch for regression bug reports.
[16:49] <xnox> because we need positive tests that things are not broken horibbly on N latest cpu families. E.g. at least as far back as e.g. ivy bridge today.
[16:49] <rbasak> xnox: fine, but it makes no difference to your example whether we backport the packaging or backport only the blogs.
[16:49] <rbasak> *blobs.
[16:50] <xnox> validation matrix explodes. I do not think we will be able to get comprehensive coverage for every cpu family, on every release series. therefore we will have to trust the validation from multiple releases.
[16:51] <rbasak> I still fail to see how it would be different in your example whether we backport blobs or backport the packaging. The packaging made no difference to that situation.
[16:51] <xnox> which is an argument for not doing these updates, and keep systems knowgly vulnerable.
[16:51] <xnox> rbasak, if the packaging makes no difference, why should i be creating N versions of packaging, instead of maintaining one version of packaging?!
[16:51] <xnox> =)
[16:51] <xnox> and then debugging the case where the packaging did make a difference.
[16:52] <rbasak> Because minimal changes to stable releases minimised regression risk, as the packaging includes interactions with other components (such as the kernel) where there is a non-zero risk of regression that can be mitigated by not changing it.
[16:52] <rbasak> *minimises
[17:03] <xnox> my understanding is that e.g. snapd, firefox, linux-hwe, cloud-archive  use a backport model of complete updates to latest upstream and packaging changes with minimal delta w.r.t. the version being backported. Rather than minimal delta w.r.t. each of the targetted releases.
[17:05] <xnox> i understand not taking 99% of changes where the actual bugfix is a one-lines; i do not understand rejecting 1% of packaging changes, when 99% are all needed.
[17:05] <rbasak> Yes, and each of those have a documented exception and individual policy. We don't currently have one of those for intel-microcode.
[17:06] <xnox> where can i look at that?
[17:06] <rbasak> And each of those has had the pros and cons considered on a case by case basis.
[17:06] <rbasak> https://wiki.ubuntu.com/StableReleaseUpdates#Documentation_for_Special_Cases
[17:06] <xnox> ack thanks
[17:06] <rbasak> Firefox isn't there. Technically it doesn't have an SRU exception AFAIK. It goes in via the security pocket, and the security team have their own polices.
[17:07] <rbasak> Though in practice an SRU would probably be fine because we do it all the time anyway (because security).
[17:12] <rbasak> I've written up what I think is the current situation in the bug. I'm EOD for now - I think from our discussion you're not in a hurry?
[17:19] <slangasek> Depends: libsbml perl (not considered)
[17:19] <slangasek> hmph
[19:14] <slangasek> cjwatson: do you have any idea why auto-sync says dictem_1.0.4.orig.tar.gz content has changed, given that it arrived in Ubuntu via auto-sync?
[19:45] <slangasek> libfile-sharedir-projectdistdir-perl/1.000008-1
[19:45] <slangasek> which are worse, perl package names or perl package version numbers?
[19:58] <infinity> slangasek: Because it did.
[19:59] <slangasek> infinity: so Debian's orig.tar.gz changed?
[19:59] <infinity> slangasek: dak is slack about maintaining the constraint if things are no longer on disk. :(
[19:59] <slangasek> ah, so this was a package removal/readd?
[20:00] <infinity> slangasek: Debian's original orig only existed briefly.  It went (accidentally, probably) native in 1.0.4-2, then back to non-native in -4
[20:00] <infinity> slangasek: But we have a record of -1, and LP won't let you do that:
[20:00] <infinity> slangasek: https://launchpad.net/ubuntu/+source/dictem/1.0.4-1
[20:00] <infinity> slangasek: So, the only solution here would be to fake the sync using the orig from the -1 upload.
[20:00] <slangasek> right
[20:01] <infinity> slangasek: Not that it matters, given that -3 and -4 were no-ops, really.  Just the maintainer asserting his non-demise.
[20:03] <infinity> (Well, they should be no-ops, according to the changelog.... Of course, the new orig might have made them excitingly different!)
[21:57] -queuebot:#ubuntu-release- Unapproved: apport (zesty-proposed/main) [2.20.4-0ubuntu4.2 => 2.20.4-0ubuntu4.3] (core)
[21:58] -queuebot:#ubuntu-release- Unapproved: apport (yakkety-proposed/main) [2.20.3-0ubuntu8.4 => 2.20.3-0ubuntu8.5] (core)
[22:00] -queuebot:#ubuntu-release- Unapproved: apport (xenial-proposed/main) [2.20.1-0ubuntu2.7 => 2.20.1-0ubuntu2.8] (core)
[22:30] <slangasek> Laney, infinity: I'm interested in your thoughts on https://bugs.launchpad.net/britney/+bug/1700668
[22:30] <ubot5`> Ubuntu bug 1700668 in britney "make it easier to reset baseline for autopkgtests that regress in release" [Undecided,New]
[23:06] -queuebot:#ubuntu-release- New sync: ocamlbuild (artful-proposed/primary) [0.10.1-1]
[23:16] <xnox> ^ would unblock ocaml transition going further
[23:18] <slangasek> not an autosync? :)
[23:19] -queuebot:#ubuntu-release- New: accepted ocamlbuild [sync] (artful-proposed) [0.10.1-1]
[23:19] <xnox> slangasek, from experimental
[23:19] <slangasek> ah
[23:21] -queuebot:#ubuntu-release- New binary: ocamlbuild [amd64] (artful-proposed/none) [0.10.1-1] (no packageset)
[23:21] -queuebot:#ubuntu-release- New binary: ocamlbuild [ppc64el] (artful-proposed/none) [0.10.1-1] (no packageset)
[23:22] -queuebot:#ubuntu-release- New binary: ocamlbuild [arm64] (artful-proposed/none) [0.10.1-1] (no packageset)
[23:22] -queuebot:#ubuntu-release- New binary: ocamlbuild [s390x] (artful-proposed/none) [0.10.1-1] (no packageset)
[23:22] -queuebot:#ubuntu-release- New binary: ocamlbuild [i386] (artful-proposed/none) [0.10.1-1] (no packageset)
[23:23] -queuebot:#ubuntu-release- New binary: ocamlbuild [armhf] (artful-proposed/none) [0.10.1-1] (no packageset)
[23:32] <xnox> slangasek, seems to me that all the bugs of failed to install/upgrade are filed against random package that happens to be the unlucky one to call insserv or run of disk space, etc.
[23:32] <xnox> https://bugs.launchpad.net/ubuntu/+source/systemd/+bugs?field.searchtext=install%2Fupgrade&search=Search&field.status%3Alist=NEW&field.status%3Alist=INCOMPLETE_WITH_RESPONSE&field.status%3Alist=INCOMPLETE_WITHOUT_RESPONSE&field.status%3Alist=CONFIRMED&field.status%3Alist=TRIAGED&field.status%3Alist=INPROGRESS&field.status%3Alist=FIXCOMMITTED&field.assignee=&field.bug_reporter=&field.omit_dupes=on&field.has_patch=&field.has_no_package=
[23:32] <xnox> trianging these is not fun.
[23:33] <slangasek> xnox: experiencing the upgrade failures is also not fun; can we get rid of insserv yet?
[23:33] <xnox> slangasek, my understanding is that we have and debian did too. but not in xenial?!
[23:33] -queuebot:#ubuntu-release- New: accepted ocamlbuild [amd64] (artful-proposed) [0.10.1-1]
[23:33] -queuebot:#ubuntu-release- New: accepted ocamlbuild [armhf] (artful-proposed) [0.10.1-1]
[23:33] -queuebot:#ubuntu-release- New: accepted ocamlbuild [ppc64el] (artful-proposed) [0.10.1-1]
[23:34] -queuebot:#ubuntu-release- New: accepted ocamlbuild [arm64] (artful-proposed) [0.10.1-1]
[23:34] -queuebot:#ubuntu-release- New: accepted ocamlbuild [s390x] (artful-proposed) [0.10.1-1]
[23:34] -queuebot:#ubuntu-release- New: accepted ocamlbuild [i386] (artful-proposed) [0.10.1-1]
[23:34] <slangasek> right, it is gone, but still causes upgrade failures in xenial
[23:35] <xnox> hm, my system seems to be broken
[23:35] <slangasek> since insserv in xenial isn't integrated with systemd the way it was with upstart, I think the right fix is to SRU insserv to make it non-fatal
[23:35] <xnox> I'm getting segfaults
[23:35] <xnox> [661832.017428] schroot[30225]: segfault at 500000002 ip 00007fc0023262b8 sp 00007ffde3b7f060 error 4 in pam_cgfs.so[7fc002323000+6000]
[23:35] <xnox> [657710.517534] pcscd[12977]: segfault at 10 ip 00007f7792767d44 sp 00007f778abc8e20 error 4 in libpthread-2.23.so (deleted)[7f779275e000+18000]
[23:35] <xnox> that can't be good.
[23:35] <xnox> slangasek, does systemd in xenial need any of the insserv orderings? and e.g. do we still use insserv orderings for e.g. shutdown?
[23:36]  * xnox thought we didn't
[23:38] <slangasek> xnox: that's what I'm saying, I don't believe it's meaningfully integrated with systemd at all
[23:39] <slangasek> should be verified, but AIUI it's mostly pointless
[23:39] <xnox> ack. let me add a trello card about that
[23:39] <slangasek> (and so shouldn't be allowed to cause upgrade errors)