[03:01] Hm [03:02] BugSubscription.date_created has tzinfo=UTC [03:02] I thought we mostly used naive UTC datetimes? [03:02] Hm, it seems to vary. [03:51] Morning [03:52] O hai [03:58] wgrant: What do you think of https://code.launchpad.net/~stevenk/launchpad/db-add-audit-spph and https://code.launchpad.net/~stevenk/launchpad/spph-approved_by-model as two thirds of bug 885739? [03:58] <_mup_> Bug #885739: queue and override manipulations should have an audit trail

< https://launchpad.net/bugs/885739 > [04:01] StevenK: Ah, I was going to talk to you about those. [04:01] StevenK: I think this is another patchwork thing. [04:02] wgrant: Oh? [04:02] StevenK: This is the sort of thing that should probably be on PU [04:03] Right. But it's a bit nasty to get at PU from SPPH, isn't it? [04:03] But it's difficult because none of this has been designed since 2005, and there's now 5x more of it. [04:04] I agree PU is probably the better place for approved_by [04:04] overrride_changed_by belongs on SPPH, and I figured out that the date was utterly pointless since a new SPPH is created. [04:05] We probably want something like ArchiveAudit which BPPH and SPPH reference. [04:06] It records all the sponsorship etc. crap, along with the PU if present. [04:06] It also records the details of the operation, for example the fact that the overrides were changed. [04:06] Heh, you mean like SPPH.spondor ? :-) [04:06] Rather than making SPPH even more obese for no good reason. [04:06] Yes. [04:06] That totally doesn't belong on SPPH. [04:06] Agreed [04:08] wgrant: So, can has mumble about ^ ? [04:08] I thought Launchpad had a policy against doing Soyuz correctly. [04:08] lol [04:09] The idea is that $randomperson just adds more columns to random places in the schema until things sort of work. [04:10] StevenK: Anyway, can mumble in a few. [04:10] Just finalising some triggers now. [04:10] Delicious, delicious PL/pgSQL [04:10] Haha [04:11] So, that's what wgrant eats for breakfast. Code. [04:11] My secret is revealed. [04:21] wgrant: A few what? [04:21] :-) [04:24] Hmm, my window decorations have gone missing. [04:30] wgrant: I'm starting to feel like I'm going to need a hardware "unity --replace" button [04:38] lifeless: O hai [04:39] huwshimi: About half the time my kernel hangs, but the other half I just Ctrl+Alt+F1 and do it there. [04:45] wgrant: Yeah, that seems about the same as me [05:04] wallyworld_: Would you mind +1ing https://code.launchpad.net/~stevenk/launchpad/kill-spph-ancestor/+merge/94716 ? [05:04] StevenK: np, looking now [05:06] StevenK: so ancestor is never used anywhere? tests or anything? [05:07] Hmmm, it is referenced in the factory [05:07] Still digging [05:09] ok, i assume you'll update the branch and then i can +1 [05:12] wallyworld_: Hah, turns out both wgrant and I were wrong. [05:13] * StevenK deletes the MP, since it's dangerous to land and would cause us to lose data [05:13] first time for everything [05:21] wgrant: how close is your multipolicy-3 branch to being merged? [05:22] wallyworld_: -50 minutes [05:23] wgrant: cool. i have a branch that i need to merge in your changes before i finish [05:23] wallyworld_: That is, it landed 50 minutes ago. [05:24] ah, i read '-' as '~' [05:24] thanks [05:24] So did StevenK. [05:24] Silly northerners. [05:24] we old farts have poor eyesight [05:41] Hmmmmmm [05:41] Declarative factory next week, I think. [05:45] wgrant: Indicies on ArchiveAudit are going to be ... fun [05:45] Easy enough. [05:45] Oh? [05:45] There's only a single date, and everything else is easy. [05:46] I want an index for every column? [05:46] Aside from date, that is [05:46] So, individual indices aren't immensely useful. [05:46] Indices for individual columns, that is. [05:47] Postgres can sometimes do a reasonable job by ANDing them together. But generally you'd create indices that satisfy the queries that you want. [05:47] Which is, in this case, none. [05:47] I was attempting to think ahead [05:47] Until we want to run queries over that table, single-column indices on id and the person FKs are what you want. [05:48] (just enough to satisfy the foreign key constraints in sensible time, since we don't know what sort of queries we want otherwise) [06:09] StevenK: hi, I'm on leave today; if its urgent please give me a ring, otherwise ping tomorrow [06:09] lifeless: Tomorrow is fine === almaisan-away is now known as al-maisan === al-maisan is now known as almaisan-away [07:18] wgrant: Can I not create a table and add comments to comment.sql in the same patch? [07:20] StevenK: You can. [07:20] psycopg2.ProgrammingError: schema "archiveaudit" does not exist [07:20] You're doing it wrong. [07:21] Indeed [07:21] Did you forget to s/TABLE/COLUMN/? [07:21] Perhaps :-) [07:22] wgrant: And I suspect we can ignore adding perms onto archiveaudit until we start to use it [07:22] No [07:22] The person foreign keys mean that there has to be SELECT and UPDATE on archiveaudit to the person merger. [07:22] fk's [07:23] Which runs as launchpad_main? [07:23] Hell no. [07:23] I see it [07:24] lifeless: Go away, you're on leave. [07:26] Reviewing PL/SQL code doesn't look any more fun after second coffee. [07:26] Haha [07:27] StevenK, wgrant: For person merge, the tests unfortunately still run as launchpad_main I think, and there is a 'person-merge-job' user or similar that is actually used on prod. [07:31] That is disappointing. [07:32] A quick bzr grep agrees, though. [07:38] Hmmm, is bug searching horrible again? [07:41] StevenK: No, should be better than usual, for a couple of reasons. [07:43] wgrant: Keeps timing out for me [07:43] OOPS ID? [07:43] Large listings are certainly not snappy, but then the dynamic stuff never has been. [07:44] wgrant: https://lp-oops.canonical.com/oops.py/?oopsid=OOPS-732cfe7bfe2e69be0875af0907fee4e1 [07:45] 2.6s here [07:47] StevenK: Still happening? [07:47] I've been idle long enough that I'm on a slave. [07:48] You're still on the master. [07:48] Which may be relevant. [07:48] It just worked. [07:49] The karma DoS is in progress. [07:49] May be relevant. [07:50] Heh [07:50] (it's something like 2000 writes per second at times) [07:52] StevenK: Shouldn't archive, distroseries, pocket be NOT NULL? [07:52] Also date. [07:53] Good point. [07:53] And this shouldn't land until we've got code for everything to populate it. [07:54] So we can actually be sure it fits. [07:54] Right [08:53] good morning [09:01] G'mornin' === almaisan-away is now known as al-maisan [09:20] stub: Thanks. [09:21] The DISTINCT ON shouldn't be necessary at all IIRC, except that there's no UNIQUE bugsubscription(bug, person) :/ [09:23] If there are multiple people with multiple results in there, you might get multiple (bug,person) rows intermixed I think. [09:23] oic. [09:24] The join should be unique already, except that the origin table is invalid because of the lack of unique constraint. [09:24] (and there are violations) [09:24] There might have been a time we did allow multiple subscriptions from a person. I wonder if person merge can still create them? [09:25] I believe that's how they all came about. [09:25] I suspect it's an oversight rather than a feature. [09:25] I'd go with the ORDER BY just in case as this is a temporary trigger [09:30] stub: Indeed, wasn't suggesting otherwise :) [09:31] I just added that as a quick hack to workaround the bad data and forgot to make it deterministic. [09:38] 'morning launchpadders [09:40] Morning jelmer :) [10:52] HI jelmer! [12:11] czajkowski: Those bugs are on some project that someone randomly added to launchpad-project. [12:11] czajkowski: We should get a webops to remove it from the group. [12:12] wgrant: context ,ooking at a lotta bugs this morning and trying to work them out [12:13] Ah, it was nearly two hours ago now. [12:13] You triaged three bugs on gazungaos [12:17] wgrant: nods the kubuntu folks [12:18] wgrant: but yes removing it from the group would be good also [12:18] thanks sorry being thick, lotta traiging and back and forthh going on this mornign [12:19] Hm? Not Kubuntu. [12:19] It's some random project unrelated to Ubuntu. === benji changed the topic of #launchpad-dev to: https://dev.launchpad.net/ | On call reviewer: benji | Firefighting: - | Critical bugtasks: 4*10 [14:25] where does the ubuntu bug pull information from for the twitter feed? [14:34] adeuring, https://plus.google.com/hangouts/extras/talk.google.com/orange-standup [14:35] flacoste, lifeless: hi, I wonder if it would be possible to export a Sprint API function that allows registering people by email address and calls ensurePerson() on that email? we'd want this to integrate better with our summit registration form which already collects a bunch of data about people [14:46] adeuring: lp:~abentley/launchpad/lazr.jobrunner [15:02] danilos: well, we don't want to add the ability to register people in Launchpad without their confirmation [15:02] danilos: so that would either need to be limited to a very specific user [15:02] danilos: or that API should email the user to ask for confirmation [15:04] flacoste, right, understood, thanks (that'd be a a simple solution for our integration of registration form/launchpad/summit), but I guess we'll have to fix summit as well :) [15:05] flacoste, basically, we have a requirement that whoever registers for our conference has a Launchpad account, but we could only get to the point of them having SSO account [15:05] flacoste, we'll have to point them to log into Launchpad before we allow them to register [15:12] danilos: that's interesting [15:13] I'm waiting on Launchpad to acknowledge my recent 'push' [15:13] danilos: maybe you could add a ensurePersonFromSSO that would create a LP profile for someone with a SSO account [15:13] I've been waiting for some time now [15:13] jml: interesting, i think abentley landed some change to the branch scanner [15:13] might be related, although i'm not sure if they were deployed [15:13] * flacoste checks [15:14] yep, was rolled out [15:14] yesterday [15:16] jml: what branch? [15:16] abentley: lp:~jml/pkgme-service/directory-already-exists [15:17] jml: So I see pushes from 30 minutes ago, but still "Updating branch". [15:18] abentley: yeah, I'm waiting for r73 [15:22] jml: looking... [15:28] jml: It appears we were just slow. [15:28] abentley: thanks. [15:48] jml: There's been pretty heavy traffic on the branch scanner today. There was a block of 8 jobs that started at 15:05:10 and finished at 15:10:25. I'm guessing you pushed in that timeframe. [15:49] jml: The next block was a block of 30 jobs that ran from 15:11:09-15:13:43. [15:50] jml: then a futher block of 13 from 15:14:09-15:17:54. [15:54] abentley: ah, ok. [15:55] flacoste, that'd work for us, should we go that route perhaps? === matsubara is now known as matsubara-lunch [15:56] danilos: yes, i think so, the complex bit is in finding out if the email address has a SSO account [15:56] flacoste, right, we can't rely on account table anymore [15:57] flacoste, do you have any idea on how we could do that? would we need something on the SSO side as well? [15:58] danilos: yes, there might be an existing api for this on canonical-identiy-provider [16:05] flacoste, ok, I'll check if there's something along the lines === al-maisan is now known as almaisan-away [16:42] * czajkowski glares at launchpad and wonders why it has to misbehave! === matsubara-lunch is now known as matsubara === deryck is now known as deryck[lunch] [17:18] czajkowski: Remember to ask StevenK about how launchpad is made. he has that long line about everything being held together by strings and jaffa tape. === deryck[lunch] is now known as deryck === almaisan-away is now known as al-maisan === matsubara is now known as matsubara-afk === al-maisan is now known as almaisan-away === Ursinha_ is now known as Guest34384 [20:41] gary_poster: any reason why buildout doesn't use the versions specified in setup.py (vs the one in versions.cfg)? [20:43] flacoste: evening, good weekend? [20:44] hi czajkowski, yes, it was great, went to party at a friend's new appartment on Saturday, how about you? [20:47] good thanks, nice weekend went to see a rubgy game, and now working on community council work after dinner [20:50] nice [20:51] yup the weather is actually improving :D [21:04] flacoste, not sure, but I think versions.cfg is applied afterwards, so it wins. [21:05] gary_poster: ah, right, and you usually don't want to pin in setup.py, so that makes sense === almaisan-away is now known as al-maisan [21:32] StevenK: wgrant: jcsackett: the trains are not running today and i have to help out my neighbour by driving her kids to school so i'll miss the standup [21:49] hi folks, how do you assign the on call reviewer? [21:56] flacoste: hi [21:56] lifeless: hang out? [21:56] sending you an invite [21:57] flacoste: it should be in your stream [21:58] lifeless: you invited my canonical.com account? [21:58] possibly not :) [21:58] lifeless: got it [22:03] benji: ^^^ since you're the on call reviewer, ping :) [22:04] cr3: we have a schedule, hold on a sec and I'll find it [22:04] benji: thanks, I'd like to adopt the same process in our team instead of inventing our own :) [22:08] cr3: the schedule is at https://dev.launchpad.net/ReviewerSchedule; you may also be interested in http://bradcrittenden.net/post/358363191/getting-your-code-into-launchpad [22:09] cr3: and this too https://help.launchpad.net/Code/Review === benji changed the topic of #launchpad-dev to: https://dev.launchpad.net/ | On call reviewer: - | Firefighting: - | Critical bugtasks: 4*10 [22:11] benji: thanks for all the info, much appreciated! [22:11] cr3: my pleasure [22:26] lifeless: Why do bugtask inserts take 290ms on prod? [22:26] Is bugsummaryjournal very slow? [22:28] lifeless: i think your ISP crashed [22:31] flacoste: bah my adsl [22:31] lifeless: yep, i need to go, i suggest we resume that conversation tomorrow? [22:32] wgrant: shouldn't be, is it ? === al-maisan is now known as almaisan-away [22:35] lifeless: https://lp-oops.canonical.com/oops.py/?oopsid=OOPS-6e6b4b29ef635939095468528e92c7ec [22:40] wonder if the rollup is honkeydory [22:40] nup 0 [22:40] lifeless: Huh [22:40] Can you try an EXPLAIN ANALYZE in a txn? [22:41] See which trigger takes the time. [22:41] I was worried about my trigger adding 0.4ms sometimes [22:41] I see that won't be a problem :) [22:42] if you prep a test I will run it for you [22:42] today is not a good day [22:42] wgrant: is our heat logic uncontended yet ? [22:43] lifeless: maxheat is no longer updated or even calculated, and garbo-hourly is doing nothing. [22:43] So indeed it should be uncontentious. [22:44] bugtask only has three triggers, and bugsummary is the only one that seems like it could really do much. [22:44] is the bug private with a lot of subscribers [22:45] It's public security. [22:45] 11 subscribers [22:45] 9 trillion contexts [22:45] https://bugs.launchpad.net/ecryptfs/+bug/732628 [22:45] <_mup_> Bug #732628: TOCTOU in mount.ecryptfs_private

Like 14 source packages [22:46] Each with 5 series [22:46] sigh, kernel team abusing packages [22:46] holy smokes [22:48] wgrant: possibly inefficient distro rollup check with each row being in the same distro or something [22:48] wgrant: should be able to reproduce locally. I suspect a bugsummary query inefficiency [22:50] lifeless: Also the kernel team abusing Launchpad [22:51] That bug has 73 tasks :/ [22:51] Maybe I should introduce a limit of 3 package tasks per distro. [22:52] Would only break the kernel team. [22:52] But would also fix Launchpad. [22:53] wgrant: how is that "abuse"? [22:54] elmo: It's an order of magnitude more tasks than Launchpad is designed for, because the kernel is duplicated in 13 source packages. [22:55] 14, sorry [22:55] wgrant: would break transition bugs as well [22:56] poolie: O hai. A user reported yesterday that http://pad.lv/nnnnn => works fine; https://pad.lv/nnnnn => 404 , is that known? [23:02] yes [23:02] i guess making it work but with an invalid cert would be a start [23:04] poolie: Or just get a free StartSSL cert? [23:04] Not sure if IE likes them, but all non-Windows browsers I've run into too. [23:04] s/too/do/ [23:04] oh, can you? [23:04] startssl.com [23:04] i guessed that :) [23:04] Their web UI sucks, but it works. [23:05] is it worth anything? [23:05] i guess the browser vendors have decided there is a modicum of security [23:05] It's probably more secure than most other CAs... [23:05] given they all just do hopeless email validation anyway. [23:05] With the added bonus of being corrupt. [23:06] Ah, they have an IE logo, so looks like even IE likes them now. [23:47] ok i'll see what i can do [23:47] feel free to file a bug against pad.lv if you wish [23:49] are people able to create distributed distributions? I'd like to create a full mirror of beta1 our developers can use to bootstrap their live-build images