[00:31] Are we out of testfix yet? [00:33] we were for a bit but I missed the window [00:33] wgrant: possibly. I've kicked/forced a bunch of builds, so we remain optimistic. [00:33] not sure if we're out of textfix again [00:34] hey jelmer! [00:34] 'morning spm! [00:34] Morning spm, jelmer. [00:35] Thanks. [00:36] hey wgrant [00:40] * thumper is getting sad writing the current email [00:41] I'm realising how far behind we are on bzr bits in LP [00:46] thumper, behind in terms of the version you mean? [00:47] jelmer: yeah [00:47] we are very behing in bzr, dulwich, bzr-git, bzr-svn and bzr-hg [00:47] prolly bzr-loom too [00:48] thumper: well, the last three have pretty much stalled for the last couple of months [00:48] there have still been a number of commits though [00:52] rockstar: hey-ho [00:52] mtaylor: I think rockstar is travelling [00:52] thumper: dammit [00:52] thumper: I need him to fix things in tarmac for me [00:53] thumper: it's discarding tag info, even though we thought the other day that it shouldn't [02:31] We're out of testfix, then? [02:32] wgrant: I'm trying to land [02:32] I think the force builds take us out of testfix [02:33] Well, mwhudson landed something a few minutes ago. [02:33] So it seems to work. [02:36] ah cool, that landed [02:38] Hm. [02:38] When did bin/test start spitting out per-test times? [02:38] it's when you have -vvv [02:38] Maybe I just added more -v than usual. [02:38] Ah. [03:17] Can someone please land lp:~wgrant/launchpad/rework-dominator-tests? [03:26] mwhudson: did the email make reasonable sense? [03:26] thumper: yes [03:26] cool [03:27] thumper: although i guess i had a head start on understanding :-) [03:28] :) [03:30] * thumper afk to pickup [03:59] Some of the queries in lib/lp/archivepublisher/domination.py are extended versions of some on ArchiveSet. In a normal application, I'd call the ArchiveSet method and then .find() on the ResultSet to do further filtering. [03:59] But IResultSet doesn't contain .find(), so I can't do that in LP. [03:59] Help? [04:23] wgrant: I think perhaps an updated storm has it... [04:24] not sure though [04:24] could ask on #storm [04:29] thumper: Thanks, I've asked there. [04:29] thumper: Could you be convinced to ec2 https://code.edge.launchpad.net/~wgrant/launchpad/rework-dominator-tests/+merge/29668 and https://code.edge.launchpad.net/~wgrant/launchpad/link-uploaded-ddebs/+merge/29669? [04:29] wgrant: yep [04:30] Thanks. [04:56] moin [04:58] Morning lifeless. [04:59] * wgrant wishes that the 3.0 heading guidelines were complete. [05:00] ...? [05:02] lifeless: Page titles were redesigned for 3.0. [05:03] But they did not consider how it should work when on a page with a non-root context. [05:03] The root context is always shown at the top of the page. [05:03] But say I'm on a page relating to a PPA. It has the owner's name at the top, and no built-in indicator of which PPA it's operating on. [05:03] What should the h1 be? [05:03] 'Administer', 'Administer archive', 'Administer PPA for William Grant'? [05:04] oh [05:04] is this a widely known issue ? [05:04] who was doing the work, can we ask them to finish it off? Or is there a good answer ppl are just doing ? [05:04] I've raised it a couple of times in the last 12 months. [05:04] There's no consistency at the moment. [05:05] when you raised it, what happened; was there some sort of consensus ? [05:05] Nothing really came of it. [05:05] There's also the whole mess around how the tabs work for non-root contexts. [05:05] The issue was identified somewhat before 3.0, but never resolved. [05:05] Resulting in a confusing UI for non-projects. [05:06] is there something you'd like to do about it ? [05:07] actually [05:07] a better question [05:07] do you have a proposal that would improve consistency and be, on the whole, more useful for users. [05:07] and not be terrible to calculate perf wise. [05:07] I don't know what would be best. [05:07] I'm not talking best. [05:07] best is freaking hard man. [05:08] I think some consistency is good - useless consistency isn't. We need to do each page really well, but a sane default can help a great deal. [05:14] Even some pages on root contexts are pretty bad. [05:14] https://code.edge.launchpad.net/launchpad/+activereviews, for example. [05:14] is there some difficulty in fixing them? [05:14] They're trivial to fix, once we define what they should be like. [05:14] uh [05:14] why do we need a pre-fix definition here ? [05:14] *anything* is trivial if you can define it :P [05:14] If we are changing titles to suck less and be more consistent, we need to define what they are going to be consistent with. [05:14] 3.0 tried to do that, but failed. [05:14] perhaps I'm over simplifying [05:14] but - a) suck less, b) be more consistent : seems we could make them suck less by /doing/ and make them more consistent by picking the nicest pattern that emerges and using it as a sane default ? [05:14] Possibly. [05:14] wgrant: if the only consistent thing at the moment is that they suck, surely we should fix that :) [05:16] we have a ui review process, so mistakes should be caught - thats a decent safety net [05:17] You can't make a mistake at the moment. [05:17] Because there are no guidelines to break. [05:18] I think we just need to work out the undefined general cases, and define them. [05:19] This may require a change to the structure of the header. [05:34] wgrant: You seem to be saying folk can't improve things here without a lot of work, which surprises me, [05:34] is that right? [05:35] lifeless: It can be mostly fixed by someone just declaring how much context information should go in titles. [05:35] A complete fix (which fixes tabs for non-root contexts) is harder. [05:36] so if I say 'the right amount to look good on a page should go there', is that useful, or am I *still* missing the point [05:38] wgrant: both your branches playing through ec2 [05:40] lifeless: How much looks good on a page? [05:40] And is that amount enough? [05:40] thumper: Thanks. [05:47] wgrant: that seems to be a matter of judgment and testing [05:47] wgrant: You don't seem to have a hesitation doing that to lp internals; I'm unclear why you have a hesitation doing the same to page templates :) [05:47] lifeless: Precisely. [05:48] lifeless: I am no UI designer. [05:48] LP has a severe lack of UI design [05:48] And it shows. [05:49] are you saying we lack the knowledge to improve at all? [05:50] No. I'm saying that somebody with significant user experience should be consulted before we go and fix all the titles to conform to a new standard that might still suck. [05:51] wgrant: so - +1 on getting more import. -1 on doing /nothing at all/ [05:51] s/import/input [05:51] We have no UI designer. How do we get more input? [05:52] we have a whole design team on call, and they are extremely happy to do small things on-request. [05:53] I'm not at all convinced that this is a problem we lack the knowledge for though. However, I'll grab mpt today and show him this chat and ask his opinion. [05:53] An attempt was made a year ago. [05:53] It did not work. [05:53] wgrant: so try something different [05:53] We can certainly make large improvements easily. [05:53] But it would be nice to get it right. [05:53] great! [05:54] the thing about right. [05:54] is that it lasts for, oh, 2-3 days. [05:54] OK, as close to right as we can easily get. [05:54] :) [05:55] * bryceh goes close to left [05:55] bryceh: any progress on the attachments thing? [05:56] hmm? [05:56] the api call [05:56] in bugs [05:56] is there a partial branch or something like that to make it faster? If not I'll have a stab at same today. [05:58] nope, I'm trying to get another branch finished up before I go on vacation [06:03] ok [06:17] losa hi [06:23] spm: hi ? [06:29] lifeless: hola [06:38] how long till edge rolls out ? [06:38] * lifeless is excited [06:39] lifeless: excited about what? [06:40] lifeless: https://wiki.canonical.com/InformationInfrastructure/OSA/LaunchpadProductionStatus "Edge Updates" 0800 BST. we strikethru if they're not for some reason or another [06:40] thumper: rev 11199 on stable [06:41] spm: kk [06:41] lifeless: which has what? [06:41] a /massive/ search overhead reduction [06:41] ah [06:41] * thumper EOWs [06:43] have a good weekend [07:07] spm: so, another hour right ? :) [07:07] * lifeless bounces [07:10] lifeless: Are you a core-dev in ubuntu? [07:11] lifeless: pretty much exactly, yeah [07:11] bilalakhtar: no, why? [07:12] * bilalakhtar is searching for core-devs to sponsor his bugfixes [07:12] lifeless: ^^ [07:12] put it in the sponsor queue please [07:56] spm: whats the url for monitoring edge rollout progress again ? I didnae bookmark it [07:57] lifeless: we don't have one [07:57] oh [07:57] thats right :) [07:57] we have one for staging or so or something [07:58] yup [07:58] https://staging.launchpad.net/successful-updates.txt <== tho is more "worked/not" vs progress [08:00] so how is edge going ? :) [08:00] <- not keen at all [08:01] OOPSes while trying to log in, a known problem? https://lp-oops.canonical.com/oops.py/?oopsid=OOPS-1665ED841 [08:01] from memory, it'll be around 5-10 mins before it gets beyond simply building/syncing the tree [08:02] danilos: ugh [08:02] spm: we may have to abort - benji's code may be bust [08:02] spm: don't stop it yet [08:03] danilos: does your username have utf8 in it ? [08:03] spm, did we get a new openid rollout by ISD guys as well? (just to confirm we only have to look at our code for the problem) [08:03] lifeless: not really; if it's a hard fail; edge1 will break and the rollout will stop [08:03] lifeless, username doesn't, but my name does [08:03] if it's more subtle than that; rolling back is trivial [08:03] spm: actually, thats already on edge. I was thinking a prod stop-point [08:03] <- EUNCAFFEINATED [08:04] * danilos disables edge redirection and tries again [08:04] lifeless: Is it complaining about the XSRF token sent by SSO? [08:04] I can't see the OOPS, but there's already a branch to fix that. [08:04] wgrant: no, unicode [08:04] Ah. [08:07] danilos: file a bug on foundations [08:08] lifeless, right, my name is the culprit, I've changed it and can log in [08:08] lifeless, prod also works, so it's edge code [08:08] its a classic 'this code wasn't QA'd' scenario :- [08:09] which the new deployment process should address [08:11] lifeless, I'd also lean on marking this one as critical, but will leave that to whoever gets a chance to take care of it [08:11] anyway, bug 609029 [08:11] <_mup_> Bug #609029: UnicodeDecodeError OOPS on login on edge [08:12] danilos: I agree [08:18] spm: hows edge, and how is the db load looking [08:20] lifeless the-impatient ;-) fwiw, your email will probably know before I do, unless something breaks hard. error reports: Subject: Cron /home/pqm/bin/rollout_edge.sh will have the details [08:21] spm: yes, its live [08:22] load looks normal, as a one off snapshot ish. [08:24] looks like edge4 is/has just been updated [08:24] has [08:26] ok, https://bugs.edge.launchpad.net/ubuntu/+filebug does seem better on edge now [08:27] Howdy [08:33] spm: ok, still looks ok ? [08:34] lifeless: I assume so, not getting any alerts; which is the usual 1st line [08:35] k, thanks === almaisan-away is now known as al-maisan [09:35] Does PQM really take half an hour? [09:45] FYI, I'm getting 'Internal Server Error' here: http://bazaar.launchpad.net/~jkakar/storm/resultselect/files [09:54] jkakar: that branch is 1.9 format, but is stacked on trunk which is 2a [09:54] jkakar: upgrade it and it'll be fine [09:55] spiv: Yeah, I figured as much... [09:55] spiv: Regardless, I shouldn't see "Internal Server Error". [09:55] Oh, certainly. [09:55] I *think* there's a bug about that already, feel free to click "affects me too" on it. [09:56] Cool. [10:37] bigjools: Are there really no tests of NascentUpload that don't use a real upload? [10:38] I can't see any :( [10:38] I guess I'll work it out. [10:40] wgrant: probably not :( I know there's some unit tests for dscfile [10:40] that's about it [10:41] but it's the direction we need to go so that soyuz tests don't take 35-40 minutes. :/ [10:41] Yeah, I've started shuffling things around so it's more possible. [10:41] archiveuploader is really ugly. [10:43] ugly with boils [10:58] wgrant: ok, mpt review done, checklist being fixed. [10:59] its like the title but less so :P [11:06] lifeless: Ooh, thanks. [11:15] lifeless, https://dev.launchpad.net/UI/Reviews?action=diff&rev2=34&rev1=33 [11:17] mpt: Thanks. [11:17] So we're to include context in the h1, even if it's the root context? [11:17] (which is already displayed at the top of the page) [11:17] mpt: thanks! [11:18] wgrant: mpt said verbally that folk ignore the global nav stuff, universally [11:18] perhaps thats worth adding [11:19] In the root context case, it's in the breadcrumbs and header already. [11:20] wgrant, remind me what "the root context" means? [11:21] mpt: The pillar or person containing this page. [11:21] The context may be a source package, but the root context is the distribution. [11:22] It's displayed at the top of every page. [11:22] (I believe the concept and term were introduced with 3.0) [11:22] wgrant: the point is people don't read that stuff [11:23] wgrant: mpt: anyhow: if we can transfer the info so that wgrant is unblocked, I will be ecstatic. I think the page improvement so far is very useful. [11:24] It looks like a good idea. [11:25] Ah, I should have mentioned another thing [11:25] More specific stuff should go earlier in the heading and the title [11:25] less specific stuff later [11:26] That's ensured in the title, since it's reversed breadcrumbs. [11:26] Although that does result in oddities like "10.04 : Ubuntu", which I'm not really a fan of. [11:27] yeah, reversed breadcrumbs are an unfortunate case of automation over sensibility [11:27] I think they are better than what came before them. [11:27] But not better than a fully redesigned set of custom titles. [11:28] They may be better technically, in that pagetitles.py was hilarious [11:28] There is at least some amount of consistency now. [11:28] pagetitles.py is almost gone. [11:28] But at least pagetitles.py let me eyeball inconsistencies. [11:28] lifeless: Have you looked at garbo.py? It is a framework suitable for many of our hourly and daily tasks - the ones that don't need external resources besides the database. [11:30] lifeless: So for instance I'll be suggesting to jtv that the cachesuggestivetranslations.py cronjob should be a task in the garbo, avoiding yet-another-cronscript-to-manage as well as bonuses such as aborts if it takes too long and fewer notifications for people to monitor. [11:31] lifeless: It needs better plugin architecture, and it isn't a replacement for a general job handler, but I think a step in the right direction. [11:38] stub: nice [11:38] stub: +1 [11:39] stub: I do think jtv's particular case would be better served by events, but the existing event queuing things are not nice enough to draw in adopters yet [11:40] I don't know enough of the task system or the suggestions to know if that is a suitable fit. [11:41] stub: making garbo replace most of the existing cron jobs would help the sysadmins a lot I think [11:42] Yup. Also makes scheduling nicer - we don't end up with batch jobs running simultaneously causing load spikes. [11:46] anyohe that has oopstools working - could you see if https://bugs.edge.launchpad.net/oops-tools/+bug/608914 has a working patch ? [11:46] jelmer: wb, did you see my reply? [11:48] <_mup_> Bug #608914: Bugs not shown in the Timeouts section of oops summaries === al-maisan is now known as almaisan-away [12:05] lifeless: yep, thanks [12:12] james_w, istm that (for example) ought to link to the packaging branches [12:12] if any [12:12] or tell us there are none [12:13] yes [12:13] "Code", but yes [12:22] Morning folks! [12:22] hey rockstar [12:30] bigjools, what's the status of the maverick buildd stuff? [12:33] rockstar: still buggered as far as I know. Are you referring to the aptitude thing? [12:33] bigjools, yeah, [12:33] bigjools, is there anything I can do to help that along? I have some other tasks, but that's one of the blockers for jorge getting upstreams to use it. [12:34] (I'm going to work on the other one today) [12:34] rockstar: I don't know, I am not dealing with it. Were you expecting me to be? [12:35] it's a problem with the recipe stuff in the slave code [12:35] bigjools, yeah, I thought that you were going to get abentley's patch applied to the buildd to fix it. [12:35] I didn't know there was a patch :) [12:35] we're currently having problems getting any recipe build to go through dogfood - they all hang. [12:36] bigjools, okay, lemme chat with abentley when he's around, and we'll sort it out. We need to make that a priority. [12:36] bigjools, :( [12:36] rockstar: see this for example: https://code.dogfood.launchpad.net/~abentley/+recipe/bazaar/+build/263/+files/buildlog.txt.gz [12:36] "bzr: ERROR: No previous changelog to take the package name from, and --package not specified" [12:36] wtf? [12:37] rockstar: ok when you get know where the patch is, we can ask lamont to deploy it on the DF builders. [12:37] bigjools, yeah, I was just about to point that out. [12:37] bigjools, ack. Thanks. [12:37] it takes 10 minutes to print that error out as well [12:38] bigjools, since you said they were hanging, I wondered if that was an artifact of killing the build. [12:38] james_w, ping [12:38] no, I didn't kill it [12:39] bigjools, we also have some sourcecode/ updates that need to be applied to fix some more bugs. [12:39] bigjools, do you have any comments on fixing https://bugs.edge.launchpad.net/launchpad-code/+bug/583403 [12:39] <_mup_> Bug #583403: Sourcepackage recipe builds need much more logging [12:41] rockstar: yeah I think the logging needs to come from bzr [12:41] bigjools, okay. Are those logs just capturing stdout then? [12:41] yes [12:41] bigjools, great. I'll get a patch together and have james_w review it. [12:42] nyshe [12:42] rockstar: we're not going to get far until we figure out that long hang + the error above though :( [12:42] bigjools, yeah, what resources do you need in order to do that? [12:43] rockstar: james_w :) [12:43] it's running stuff I know nothing about unfortunately [12:43] bigjools, rockstar: That recipe is buggy. [12:44] well that's a good start. [12:44] It should be a merge, not a nest. [12:44] bigjools, okay, well I'll try and reproduce it here. james_w is kind of going into maintenance mode on some of these udd tasks, and I offered to pick up maintenance for them. [12:44] it should fail a bit quicker though [12:44] (Of course, that means I need to learn them first) [12:44] How long does it hang? [12:44] 10 minutes [12:44] Interesting. [12:44] * bigjools fixes the recipe [12:45] bigjools: Just s/nest/merge/, and drop the 'debian' from the end. [12:45] That might be a bzr stupidity. [12:46] Like bzr builds a whole working tree in memory just to get the single folder, or something. [12:47] dispatching the new recipe now [12:47] wgrant: btw I have a new buildd-manager on DF. [12:47] should be fun to see if it breaks production :) [12:47] Heh. [12:47] As in the new properly parallel one? [12:47] testing so far shows no issue [12:48] yes [12:48] Now we just need to convince it to dump uploads in a queue, and the problem is mostly fixed! [12:48] I do like Twisted. I don't like the b-m code so much. [12:48] yes, jelmer is working on that :) [12:48] Excellent. [12:51] I've also ported lp-buildd to use a modern system version of sbuild, so we can drop our unmaintained ancient buggy fork once Soyuz does ddebs. [12:51] So we should have a much healthier, more maintainable build farm soon, on multiple fronts... [12:52] losa ping [12:52] wgrant, this, my friend, is most excellent news. [12:52] http://pastebin.com/e9D6x3hX <- I'd like timing data for that query on each DB host please. [12:52] lifeless: hi [12:52] wgrant: \o/ \o/ \o/ [12:52] Although I had to destroy buildrecipe in the process. [12:52] wgrant, now, if only we could have test coverage for the build farm. [12:52] james_w, https://bugs.edge.launchpad.net/launchpad-registry/+bug/609116 [12:52] <_mup_> Bug #609116: package page should point to package branches (or not) [12:53] rockstar: Translations made a good start on testing their part of it. [12:53] it the literal query running on filebug queries on edge; its timing out a lot - I'm interested in figuring out if: [12:53] - there is one bad host [12:53] - there are bad query plans being generated [12:53] mthaddon: oh, and hai [12:53] :) [12:54] bigjools: Um, where did that recipe go? [12:54] It has disappeared. [12:54] I wonder if you ran into a form bug and accidentally reassigned it yourself. [12:54] Heh, yes, there it is. [12:54] the log was there briefly [12:55] wtf [12:55] * rockstar waits for the expansion of "form bug" [12:55] The owner