-queuebot:#ubuntu-release- Unapproved: ubuntu-release-upgrader (disco-proposed/main) [1:19.04.3 => 1:19.04.4] (core) | 00:06 | |
-queuebot:#ubuntu-release- Unapproved: accepted lxc [source] (xenial-backports) [3.0.2-0ubuntu4~16.04.1] | 00:11 | |
-queuebot:#ubuntu-release- Unapproved: accepted lxcfs [source] (xenial-backports) [3.0.2-0ubuntu1~16.04.1] | 00:11 | |
-queuebot:#ubuntu-release- Unapproved: accepted lxd [source] (xenial-backports) [3.0.2-0ubuntu1~16.04.1] | 00:13 | |
mwhudson | so what's the python3.7-defaults story? | 00:18 |
---|---|---|
mwhudson | ruffus needs bad-testing or marking as a "big" test | 00:18 |
mwhudson | and then it's gotten entangled with the perl transition? | 00:18 |
xnox | mwhudson, yes https://code.launchpad.net/~xnox/autopkgtest-cloud/+git/autopkgtest-cloud/+merge/358468 | 01:04 |
doko | mwhudson: disentangled, perl went in. ruffus is the last blocking one for python3-defaults | 05:08 |
doko | ahh no, looking at notest ... | 05:09 |
=== blackboxsw is now known as blackboxsw_away | ||
doko | graphviz is blocking perl as well via the failing python-ruffus test | 05:13 |
-queuebot:#ubuntu-release- Unapproved: automake-1.16 (disco-proposed/main) [1:1.16.1-3 => 1:1.16.1-4] (core) (sync) | 05:40 | |
mwhudson | doko: we need to point Laney at https://code.launchpad.net/~xnox/autopkgtest-cloud/+git/autopkgtest-cloud/+merge/358468 then? | 06:09 |
mwhudson | or sil2000 but he's on leave | 06:10 |
mwhudson | maybe apw | 06:10 |
doko | Laney: please update for bionic: laney:force-badtest libreoffice/1:6.0.3-0ubuntu1/i386 | 08:25 |
doko | to libreoffice/1:6.0.6-0ubuntu0.18.04.1 | 08:25 |
doko | still fails on i386 with the same issues | 08:26 |
doko | infinity: for bionic you have: adconrad:unblock ocrmypdf/6.1.2-1ubuntu1 please could you mark this test as failing as well? seen triggered by python3-defaults | 08:27 |
Laney | doko: https://code.launchpad.net/~laney/britney/hints-ubuntu-bionic-glib2.0/+merge/358449 | 09:03 |
Laney | mwhudson: thanks | 09:03 |
Laney | (I do get the emails though) | 09:04 |
-queuebot:#ubuntu-release- Unapproved: nautilus (disco-proposed/main) [1:3.26.4-0ubuntu7 => 1:3.26.4-0ubuntu8] (ubuntu-desktop) | 09:20 | |
-queuebot:#ubuntu-release- Unapproved: plasma-discover (cosmic-proposed/universe) [5.13.5-1ubuntu6 => 5.13.5-1ubuntu6.1] (kubuntu) | 11:07 | |
-queuebot:#ubuntu-release- Unapproved: plasma-discover (disco-proposed/universe) [5.13.5-1ubuntu6 => 5.13.5-1ubuntu7] (kubuntu) | 11:08 | |
Laney | xnox: can you do ruffus retries with the right triggers pls | 11:15 |
doko | Laney, xnox: looks like LocutusOfBorggave those back | 11:43 |
Laney | coordination is hard | 11:44 |
-queuebot:#ubuntu-release- Unapproved: zeroc-ice (disco-proposed/universe) [3.7.1-4ubuntu2 => 3.7.1-5] (cli-mono) (sync) | 11:51 | |
-queuebot:#ubuntu-release- Unapproved: accepted zeroc-ice [sync] (disco-proposed) [3.7.1-5] | 12:02 | |
doko | python3-defaults migrated \o/ | 12:29 |
LocutusOfBorg | lol somebody generated a trigger wrongly | 12:33 |
LocutusOfBorg | libhtml-parser-perl/3.72-3build2 ibmoose-perl/2.2011-1build1 libnet-ldns-perl/0.75-3build1 perl/5.28.0-3 | 12:33 |
LocutusOfBorg | forgetting an "l" :) | 12:33 |
LocutusOfBorg | this should make perl a little better if libzonemaster-perl tests are good | 12:33 |
LocutusOfBorg | qaplà doko! | 12:33 |
-queuebot:#ubuntu-release- Unapproved: zeroc-ice (disco-proposed/universe) [3.7.1-5 => 3.7.1-5ubuntu1] (cli-mono) | 12:59 | |
-queuebot:#ubuntu-release- Unapproved: accepted zeroc-ice [source] (disco-proposed) [3.7.1-5ubuntu1] | 13:05 | |
xnox | mwhudson, i am wondering if clisp is missbuilt on 32bit platforms | 14:08 |
-queuebot:#ubuntu-release- Unapproved: gnome-initial-setup (cosmic-proposed/main) [3.30.0-1ubuntu3 => 3.30.0-1ubuntu3] (ubuntu-desktop, ubuntugnome) | 14:25 | |
-queuebot:#ubuntu-release- Unapproved: rejected gnome-initial-setup [source] (bionic-proposed) [3.30.0-1ubuntu3.1] | 14:26 | |
=== blackboxsw_away is now known as blackboxsw | ||
santa_ | dear release wizards, | 15:50 |
santa_ | not sure if this is relevant for the pre-release work but systemd has a higher version in cosmic than disco because of a security update: https://launchpad.net/ubuntu/+source/systemd | 15:52 |
santa_ | that got into my way when building packages for disco, not a big tradegy since I could workaround it | 15:53 |
cjwatson | I could copy it forward but I'd rather wait until disco is properly open before doing stuff like that | 15:59 |
cjwatson | Just in case it gets in the way of some transition that's blocking disco opening | 16:00 |
xnox | santa_, cjwatson - well there is one more security update coming | 16:01 |
xnox | (in security proposed ppa) | 16:02 |
santa_ | no prob, just wanted to notify the issue, just in case | 16:08 |
xnox | santa_, yeah, cool thanks. | 16:14 |
infinity | doko: unblocks have nothing to do with tests. That test has very much regressed from the looks of it. | 16:47 |
infinity | doko: (re: ocrmypdf) | 16:47 |
ginggs | infinity, doko: i think that ocrmypdf failure is due to ghostscript | 17:14 |
infinity | ginggs: The last two definitely seem to be addressed by: | 17:14 |
infinity | https://github.com/jbarlow83/OCRmyPDF/commit/517b385fe5cb2195023100a807e6f18dc7e6faea#diff-e00b5801e6b2703e3d116bb5e533a81d | 17:14 |
infinity | Not sure about the first. | 17:14 |
vorlon | vim hinted through; maybe that's enough to get perl in | 17:17 |
vorlon | ah no, clisp-module-gdm needs separate sorting | 17:17 |
infinity | vorlon: And the two bizarre rdep failures for clisp itself... | 17:18 |
ginggs | infinity: i think this https://git.spwhitton.name/ocrmypdf/commit/?h=debian&id=6b5934ff4ef974613474fa69077777d2d2f338a5 | 17:18 |
xnox | vorlon, infinity - should fix livecd-rootfs tests [ubuntu/disco-proposed] livecd-rootfs 2.544 (Waiting for approval) | 17:23 |
infinity | xnox: That change is silly (IMO) | 17:28 |
infinity | xnox: snap-seed-parse isn't some third-party binary we have no control over, it's *from* livecd-rootfs, and there's no reason for it to be logging informational messages to stderr. | 17:29 |
infinity | xnox: Ignoring stderr because we're dumb isn't the right answer. | 17:29 |
xnox | infinity, i'm not ignoring stderr. | 17:29 |
xnox | infinity, i'm logging it to the full log. and there are messages there, printed always. | 17:29 |
infinity | xnox: You're sending stderr to stdout, so yes, you're ignoring that there's stuff on stderr. | 17:29 |
xnox | on my books, ignoring would mean 2>/dev/null | 17:30 |
infinity | xnox: From the POV of autopkgtest trapping stderr as "maybe there were errors" (which is a sane default), redirecting it to stdout is ignoring. :) | 17:30 |
xnox | infinity, i agree this is not nice, but this unblocks the world. making that script better is/was out of scope for me. | 17:30 |
xnox | infinity, i did flag this up previously, but there is no tracktion on this. and all of these people are gone this week. | 17:31 |
xnox | there are multiple things wrong here =) people really should run livecd-rootfs adt before upload, locally. it works. | 17:31 |
xnox | and cpc sauce should not have started to use a newly introduced functions before livecd-rootfs migrates and gets published in their ppa. | 17:32 |
infinity | Yeah, I guess the real problem here is that the result is on stdout, so the author decided to log to stderr. | 17:32 |
infinity | Taking a second argument for output file, then logging to stdout (except for actual errors to stderr) would be saner. | 17:32 |
xnox | i'm happy to open that as a bug report | 17:33 |
infinity | Please do, and assing to the original committer with prejudice. :) | 17:33 |
infinity | But yeah, we can take your hack for now. | 17:33 |
infinity | I'm okay with temp hacks, as long as they're actually temp. | 17:34 |
infinity | Knocking on all the wood. | 17:34 |
xnox | cause like doko has been pointlessly retrying livecd-rootfs tests for days now =) he clearly wants to keep the planet warm via our scalingstack ;-) | 17:34 |
Laney | :/ | 17:38 |
vorlon | anyone else already working on clisp? | 18:58 |
infinity | vorlon: I only scraped the surface enough to say "WTF", feel free to dig deeper (please). | 19:19 |
infinity | vorlon: If the cl-asdf and cl-unicode failures end up being the only perl/gdbm blockers, I'm tepted to badtest them (even though that's a bit of a lie) because hey, two out of three lisp interpreters still work. | 19:21 |
infinity | vorlon: I'd be more comfy about that call if I knoew what the most popular common list interpreter was, mind you. | 19:22 |
coreycb | doko: can you reject my latest openstack-pkg-tools upload to disco? it is fixed instead by your latest upload of python3-defaults. | 19:23 |
vorlon | infinity: well good news, I can't get cl-asdf/armhf to pass at all on the porter box | 20:07 |
infinity | vorlon: That's news? | 20:08 |
mwhudson | NOTAREGRESSION i assume? | 20:09 |
mwhudson | also not a regression is my python-tornado SRU failing to build in cosmic | 20:09 |
vorlon | infinity, mwhudson: I mean that I can't get a clean baseline | 20:09 |
vorlon | it passes on autopkgtest | 20:09 |
mwhudson | ah | 20:09 |
vorlon | then fails on the porter box | 20:09 |
infinity | Why have I forgotten the name of the porter? | 20:09 |
vorlon | rugby? | 20:10 |
infinity | Okay, better question, why can't I see it in DNS? :p | 20:10 |
* infinity restarts the VPN. | 20:10 | |
mwhudson | the arm64 porter has always had oddball dns | 20:10 |
infinity | Well, it's rugby.inernal, if I recall, but irritatingly not resolving for me. | 20:10 |
mwhudson | well modulo the typo, yes | 20:11 |
infinity | Though, neither is ftpmaster.internal, so... I blame systemd-resolved. | 20:11 |
vorlon | infinity: yes, see discussion internal #is right now | 20:12 |
mwhudson | those are fun looking errors | 20:12 |
vorlon | nm something systemd-resolved something something broken | 20:12 |
vorlon | anyway, cheating and using my access to autopkgtest infra :/ | 20:12 |
infinity | vorlon: So, you're saying that on rugby, it fails the same way with the release pocket version of clisp? Or fails differently? | 20:14 |
vorlon | infinity: fails differently | 20:14 |
vorlon | (at an earlier stage) | 20:14 |
infinity | Fun. | 20:14 |
vorlon | ugh and launching a disco autopkgtest container gave me no network by default | 20:30 |
tsimonq2 | vorlon: Would that happen to be an LXD container? | 20:30 |
vorlon | yes | 20:30 |
tsimonq2 | wxl ran into the exact same thing with Cosmic images last cycle and we just could not figure it out... | 20:31 |
tsimonq2 | (It worked on my machine, so it was just me suggesting ideas, but... :P) | 20:31 |
vorlon | it's not generally speaking a problem | 20:31 |
vorlon | I don't know what's up with this particular disco image | 20:31 |
teward | vorlon: does a manually launched disco image in LXD work without issue with networking? | 20:32 |
vorlon | cloud-init missing from the image, that might be it | 20:32 |
teward | just curious :P | 20:32 |
teward | ahhhh, cloud-init, that evil thing... | 20:32 |
vorlon | cloud-init, that essential piece of infrastructure | 20:32 |
* tsimonq2 hands teward another gallon of coffee. | 20:32 | |
* teward dumps it over tsimonq2 and goes back to determining whether he should propose an SRU or just a backport of partclone due to the major bug he found in it today | 20:33 | |
vorlon | ok, and an autopkgtest container gives me the same result as on rugby, and not the behavior seen in the autopkgtests | 20:40 |
infinity | ... | 20:41 |
vorlon | https://paste.ubuntu.com/p/DzvTdhT9KC/ if someone wants to explain this | 20:41 |
infinity | vorlon: export AUTOPKGTEST_TMP | 20:42 |
vorlon | oh. it's caused by the runtests.lisp being sensitive to $AUTOPKGTEST_TMP being set in env. :P | 20:42 |
vorlon | yeah | 20:42 |
vorlon | aaaand now it passes with clisp from -proposed | 20:43 |
infinity | ... | 20:43 |
infinity | With clisp from proposed? :( | 20:43 |
infinity | WTF. | 20:43 |
infinity | I mean, yay, you can badtest it with "passes with manual test", but also, WTF. | 20:43 |
infinity | clisp is a flaming heap. | 20:44 |
vorlon | well, I also wasn't very granular, let me see about --all-proposed | 20:44 |
infinity | I'm not sure how all-proposed would change much with those two cl* tests. | 20:44 |
infinity | Did anything really change other than clisp itself being rebuilt? | 20:44 |
vorlon | infinity: point being that when I pulled in clisp from -proposed I also pulled in a bunch of other stuff (e.g. perl) and I don't know *why* clisp would care, but maybe it does | 20:47 |
infinity | vorlon: Right, I wasn't questioning the "you had a bunch of taint" argument, just that I too can't see why it would matter in this case. :P | 20:48 |
vorlon | k | 20:48 |
infinity | Since clisp's dependency chain is approximately nil. | 20:48 |
vorlon | ITYM 'NIL' | 20:48 |
infinity | And cl-asdf even less so (literally just wants *some* lisp interpreter) | 20:48 |
vorlon | dependencychainp? | 20:49 |
infinity | (( stahp)) | 20:49 |
teward | vorlon: because i'm tired: ITYM = ? | 20:50 |
infinity | teward: I Think You Mean. | 20:50 |
teward | ah, of course, thanks. :) | 20:50 |
infinity | vorlon: Okay, publisher run to remove those two things is about to go. | 21:21 |
infinity | ... in 2 minutes. | 21:21 |
infinity | vorlon: Of course, that removal will break all the cl-* autopkgtests that test-depend on clisp and attempt to run it... | 21:24 |
infinity | Unless they're somehow magically conditional. | 21:24 |
vorlon | ¯\_(ツ)_/¯ | 21:24 |
infinity | Yeah, I don't think there is a magical "only run this test if the deps exist" conditional in autopkgtest is there? | 21:25 |
infinity | That might be useful. | 21:25 |
vorlon | there is now | 21:25 |
vorlon | recently added. not sure we support it yet. | 21:25 |
infinity | As would be a Test-Architectures field, maybe. | 21:25 |
vorlon | and I don't remember what it's called | 21:25 |
vorlon | but I did see it show up as a restriction in a diff recently - skip-not-installable or so? | 21:26 |
infinity | Oh, nice. | 21:26 |
vorlon | https://launchpad.net/debian/+source/autopkgtest/+changelog tada - well-named | 21:26 |
infinity | Although, not-avilable would be more what I'm after. | 21:26 |
infinity | vorlon: Yeah, that looks like it's what it implies (it'll skip if the installation pass fails), which is icky. But in this case, would work alright. | 21:30 |
infinity | vorlon: Still gross, though, cause it means tests with that restriction will now skip/"pass" if the dep chain is temporarily in an uninstallable state, when what I think the submitter really wanted was "if the deps don't exist". | 21:31 |
infinity | Oh well. | 21:31 |
vorlon | infinity: alright, I can't reproduce the cl-unicode/i386 segfault either | 21:38 |
vorlon | so maybe we should just badtest those? | 21:38 |
infinity | vorlon: Fine with that too, it'll just sail back in. | 21:38 |
infinity | (but also, wtf) | 21:38 |
vorlon | of course, this means those packages' autopkgtests can no longer be used to catch regressions in either sbcl or ecl | 21:39 |
infinity | vorlon: Just one test of one CL package on one arch each, it's probably acceptable collateral damage. | 21:39 |
infinity | We certainly have much more scary/dangerous badtests (*cough*systemd*cough*) | 21:40 |
vorlon | and what should we make of the fact that in all cases, the bad addresses are 0xaf[...]? | 21:40 |
infinity | 0xaffluenza? | 21:41 |
vorlon | looks like clisp doesn't have any ASLR. maybe that's reason enough to kill it | 21:42 |
infinity | I have no strong opinions about how many lisp inerpreters/compilers we need, nor which ones. | 21:43 |
infinity | It feels like Debian's CL people have more or less settled on sbcl as the One True C Lisp. | 21:44 |
mwhudson | it's also the only common lisp i have drunk beer with two of the authors of! | 21:53 |
infinity | mwhudson: I've had beer with several of the systemd authors, I'm not sure that's helped me hate it any less. | 21:54 |
infinity | SUCCESS (631/8) | 21:54 |
infinity | I think I'm about to get a lot of email. | 21:54 |
mwhudson | infinity: fair | 21:56 |
infinity | doko: Do you have an archive opening mail prepped? | 21:59 |
tsimonq2 | >_> | 22:00 |
xnox | infinity, i acknoledge systemd continious flakiness. but also note that it's like a leaky poo.... cause it's constant improvements and constant new regressions =/ never ending battle. | 22:09 |
infinity | "It's like a leaky poo". | 22:09 |
infinity | Thanks for that. | 22:09 |
xnox | the current flakiness around activating / starting / that weird gdm is the latest puddle. | 22:10 |
infinity | This metaphor isn't getting any more pleasant. | 22:11 |
* xnox ponders how many publishing cycles it will take, given that like delete is done; yet publish is not yet there. | 22:18 | |
infinity | xnox: Patience. LP's copy interface is async, I'm making sure everything went through before I publish again and turn reports back on. | 22:19 |
xnox | nice! thanks for doing it without havoc. | 22:19 |
infinity | I wish there was a restricted copy_package_sync() that britney could use. | 22:21 |
infinity | Better yet, move_package_sync(), so it would copy/delete atomically, and get a pass/fail return. | 22:21 |
infinity | Sadly, there is not. | 22:21 |
xnox | well. imho we should only do copy_package; not move. and have an separate pass to notice duplicate/equal things published in both -proposed and release, and clean that up later. | 22:22 |
infinity | That's probably what we should be doing today, to avoid things accidentally going missing. | 22:23 |
xnox | cause having identical thing published in both -release and -proposed is harmless; removing a thing from -release without a new one publish and -proposed one removed as well, is bad. | 22:23 |
infinity | But move would be cleaner. | 22:23 |
vorlon | infinity: still waiting for everything to publish? | 23:59 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!