[00:28] rockstar: can you ping me if/when you might have a moment? [00:29] rockstar: so what was up with your template [00:51] Project devel build (166): ABORTED in 1 hr 39 min: https://hudson.wedontsleep.org/job/devel/166/ [00:51] * Launchpad Patch Queue Manager: [r=adeuring][ui=none][no-qa] New utility script, [00:51] massage-bug-import-xml, [00:51] that attempts to repair some commonly seen problems with bug import [00:51] XML. [00:52] * Launchpad Patch Queue Manager: [r=allenap][ui=none][bug=655242][bug=505304][bug=504126][bug=606866] [00:52] Fix OpenGPG documentation. [00:52] * Launchpad Patch Queue Manager: [rs=me][ui=None][no-qa] Slow the rate of polling the buildd-manager [00:52] does to reduce the load on the database and cesium (the [00:52] buildmaster host). [00:52] LPCIBot: Shhh, don't tell everyone. [00:53] thumper: quick question... ping? [01:09] wallyworld_, what's up gangster? [01:09] rockstar: can you skype? [01:13] wallyworld_, I can't, actually. :( [01:13] ok. i just wanted to touch base about your merge queues email [01:14] one question - while i wait for your branch to merge in to db-devel, can i merge from your lp copy and later merge from db-devel with no problems? [01:14] that way i can start using your stuff now [01:14] wallyworld_, yeah, you certainly can. [01:14] wallyworld_, yeah. deryck and I have learned some icky edge cases that don't play well with feature flags. [01:15] anything i need to know? [01:15] also, do turn on the merge queue feature flag, do i need to insert a record into my local db? [01:16] wallyworld_, just do it the way I did it, and you should be good. [01:16] ok. i'm also unsure what you meant by merge queue vocabulary and the comments around that [01:18] actuallt, to clarify my previous question, i know how to wrap stuff inside a feature flag, but how do i enable that flag locally for my testing? [01:18] inprod, there would be a db record required, no? [01:20] wallyworld_, see my branch: lp.code.browser.tests.test_branchmergequeue [01:21] rockstar: np [01:21] Project db-devel build (108): STILL FAILING in 3 hr 55 min: https://hudson.wedontsleep.org/job/db-devel/108/ [01:22] that is fine for unit tests, but if i want to run up a system to try out, is there a preferred way for that? [01:22] wallyworld_, you don't need to worry about the vocabulary stuff. I basically just need to be able to get a set of BranchMergeQueue items for a given criteria. You're currently working on a solution for that. [01:23] rockstar: yes. the only methods currently supported are visibleBy() and ownedBy() but the infrastructure is there to extend as required [01:23] wallyworld_, yeah, I can extend it. [01:25] rockstar: ok, i'll grab your branch off lp. you can grab my working copy too i guess if you need to. i'll email you a status later today [01:26] wallyworld_, yeah. It'd be great if you could get it landed though. Pipes don't like to have merges from elsewhere. [01:27] rockstar: ok. the main work left is the visisbleBy() and the tests and merging in your latest once it lands into db-devel [01:28] wallyworld_, cool. I've got LOTS of other things to work on, so don't feel too much pressure. [01:28] wallyworld_, I'm going to be busy all weekend. [01:28] ok. np. thanks. [01:43] hey [01:44] straw [01:44] I've made it to the shared work space [01:44] very interesting [01:44] and free for where I'm at :-) [01:44] so now everyone else there has to put up with you too and not just us :-) [01:46] heh, yeah [01:47] thumper, cool. I'm interested to see what your experience with it is. [01:47] rockstar: sure [01:47] rockstar: how's UDS going for you? [01:47] thumper, I'm REALLY tired. :) [01:47] heh, not surprised [01:47] deryck and I are watching football and spending a quiet evening in. [01:47] :) [01:48] * thumper needs to focus on actually doing something :) [02:39] rockstar: hey [02:39] rockstar: so what was up with the request.features thing the other day [02:40] lifeless, complications in using page fragments. [02:40] It seems that in some cases, page fragments result in entirely separate request objects than the original request, even though it's technically in the same request. [02:41] I think page fragments are a potential performance problem. [02:41] Maybe not the biggest performance problem, but something we may need to evaluate. [02:42] That's why we would get those separate "requests" halfway through a pdb of a page: it was getting another page fragment. [02:42] yeah, they've been on my radar already. [02:42] however why weren't the request *objects* getting a start_request event ? [02:43] lifeless, I think it's because it wasn't an actual request, but a page fragment. [02:43] They'd technically already had a start_request event. [02:44] When you left, we were checking for "features" in self.__dict__ when we should have been checking for "_features" in that pdb setup we made. [02:45] When I fixed that, I found that we weren't ever getting without setting. [02:45] Although the request object did some funky things. [02:45] so, you've a work around ? [02:46] lifeless, no, the actual issue was another type poolie found. [02:47] I was missing a semicolon in my tal:define. [02:47] lifeless, so I got it to work when I defined features correctly. [02:47] lifeless, the pdb quest was very eye opening though. [02:47] -ah [02:47] coolio [02:48] The branch index page traverses zcml at LEAST 8 times... [02:48] yeah [02:48] we have a lot of inefficiencies [02:48] what we need to do, once we have the coarse performance headaches out of the way [02:48] is to start streamlining the stack, [02:53] rockstar: what do you mean traverses zcml at least 8 times? [02:54] thumper, every time we go get a page fragment, we have to look it up in zcml. [02:54] rockstar: you mean humans? or machines? [02:54] rockstar: because the machine doesn't touch the zcml [02:54] rockstar: it is parsed once [02:54] thumper, for context, lifeless and I went on a pdb adventure yesterday trying to sort out a feature flag problem. [02:55] thumper, yeah, but it's stored in an object tree, which gets traversed. [02:57] Surely the adapter lookup doesn't suck too much, though... [02:57] I thought that was pretty much optimized at the zope level [02:58] thumper: its about 40 function calls to do a notify() [02:58] make no assumptions [02:58] heh [02:58] 40? [02:58] hmm.. [02:58] thumper, easily 40. [02:58] that's kinda shit [02:58] conservative [02:58] theres a profound degree of indirection in the zope core [02:58] some bits are highly tuned [02:59] lifeless, yeah, stepping through a whole request life was VERY eye opening for me. [02:59] but not all, not by any stretch [02:59] rockstar: ;) [03:03] lifeless, kapil also said he made a really good debug WSGI middleware that gets good SQL tracing and such. [03:03] sadly we really don't have a wsgi stack [03:03] we're glued on the end [03:03] OTOH we have damn awesome sql tracing already [03:04] * wgrant likes the Django Debug Toolbar. [03:04] wgrant, I have tried to duplicate that since day 1 at Canonical. [03:05] web debugger? [03:05] (It hooks into WSGI in a couple of places, adding a toolbar on the right of every page which lets you easily see template contexts, seeing SQL statements and running EXPLAINs and all sorts of other awesome stuff.) [03:05] should be broadly straight forward. [03:05] lifeless, no, just adds some HTML with information about query times and templates used and performance times. [03:05] rockstar: oh, just make the comment at the bottom visible via a feature flag. [03:05] lifeless, we're not using a WSGI stack? I thought gary said we were. [03:06] I've been meaning to do it, but after edge going [03:06] rockstar: FSVO stack. [03:06] lifeless, no, you'd never enable this on production. It'd be a devserver thing. [03:06] rockstar: we're 95% custom publication code incompatible with wsgi [03:06] wgrant, FSVO? [03:06] For Some Values Of [03:06] rockstar: oh, we have a bunch of stuff that I'd be fine showing to you on prod. [03:07] you could expand it for lp.dev [03:07] lifeless, yeah, but the query middleware makes things pretty slow. [03:07] AIUI it's not so much a WSGI stack as some Zopeish Launchpaddery with a WSGI frontend glued on. [03:07] right [03:08] wgrant, yeah, that's what I was assuming, but I guess I assumed there was more WSGI-ish stuff than what's there. [03:10] Is crowberry still unwell? [03:10] This remains to be seen [03:11] wgrant: Why? Do you have an issue? [03:11] Hard / Soft Page ID [03:11] 283 / 17 Archive:EntryResource:getBuildSummariesForSourceIds [03:11] 180 / 30 Person:+commentedbugs [03:11] 140 / 203 CodeImportSchedulerApplication:CodeImportSchedulerAPI [03:11] 100 / 251 BugTask:+index [03:11] 98 / 3 MailingListApplication:MailingListAPIView [03:11] 69 / 270 Distribution:+bugs [03:11] 16 / 0 Sprint:+temp-meeting-export [03:11] 13 / 1018 Archive:+index [03:11] 13 / 44 Archive:+packages [03:11] StevenK: Just wondering about the frequent downtime lately. [03:11] 13 / 1 ProjectGroup:+milestones [03:12] lifeless: That top one is a bit odd. [03:12] wgrant: We are clearly only trying to make you think there is an issue. === Ursinha is now known as Ursula [03:14] * rockstar shuts laptop and concentrates on football [03:17] wgrant: whats odd aboot it ? [03:19] lifeless: I don't recall seeing that one high up before. [03:19] But I may be wrong. [03:20] https://bugs.edge.launchpad.net/soyuz/+bug/662523 [03:20] wgrant: I've been nagging you to stab it for weeks [03:20] Hmm. [03:20] well [03:20] 11 days :P [03:21] StevenK: Has there been any discussion at UDS about deleting partner? [03:21] some [03:21] wgrant: Indeed. We have a plan. It will just take 4 years. [03:21] StevenK: ie. until Lucid EOL? [03:21] Yes [03:22] Can't we just repurpose commercial-compat to rebrand a PPA as partner? [03:22] * StevenK shrugs [03:22] Then terminate ArchivePurpose 4 with extreme prejudice? [03:23] wgrant: Does this mean you have a branch that removes every bit of ArchivePurpose.PARTNER and you're looking for an excuse to land it? [03:23] StevenK: No. [03:23] Pity. [03:23] Unlike lucilleconfig. [03:25] Has that all trickled through? [03:25] * thumper cries quietly while looking at blueprint code [03:26] thumper: i wonder if you look at the code every week, say, it gets less surprising how awful it is [03:26] mwhudson: I'm betting not [03:26] (i doubt it, the brain doesn't remember pain) [03:26] mwhudson: as part of my blueprint work I have a herd of yaks to shave [03:26] Which particular herd is it? [03:27] i suppose if you spend an hour improving it each week, eventually it will be ok [03:27] in like 150 weeks [03:27] Which is 3 years [03:27] yeah, sounds optimistic [03:28] Optimistic at best. A nightmare at worst. [03:28] thumper: Put the rope down. [03:28] * thumper might start soonish with the shaving [03:28] * thumper shakes the shaving foam can [03:29] is it worthwhile renaming Specification to Blueprint? [03:30] thumper: It's been through three names already... leave the poor creature alone. [03:30] that would confuse people [03:31] what, renaming it? [03:31] wgrant: three names? [03:31] thumper: what's the heritage of the blueprint code> why is it so bad? [03:32] s/>/? [03:32] wallyworld_: that is a long story [03:32] perhaps we can talk about it on Monday [03:33] thumper: Features, Blueprints, Specifications. [03:33] ok. sounds like something best discussed with a case of red at close hand [03:33] But it was renamed to Blueprints long before most people here appeared. [03:33] the word Specification is only mentioned 2699 times in the code [03:33] simple find and replace :) [03:33] Heh. [03:33] Was there a Feature to rename Features to Blueprints? [03:34] StevenK: no, it would have been Feature to Specification [03:34] and Specs to Blueprints [03:34] I think it was originally specification. [03:34] the UI changed, but not the code [03:34] Then to feature, then specification, the blueprint. [03:34] But that first transition may be wrong. [03:34] We changed it *back*? [03:34] I just find it a tad annoying in the code [03:36] and 100 files would need to be renamed [03:37] but that is a small price to pay I think for some form of consistency [03:37] And a 17,000 line diff [03:37] sure, whatever [03:38] as long as it doesn't conflict :-) [03:38] ... more than 3 times [03:38] I'd be against the rename. Most other objects in Launchpad now have sane names. [03:38] Specification -> Blueprint is the opposite of sane. [03:39] The URL is blueprints.l.n [03:39] It is. [03:39] i've never used the term Blueprint wrt software. for me, it's always been a Specification Document [03:39] But most people call them specs. [03:39] blueprint are something use use to build a !$%@$ house [03:39] Heh. [03:39] So now they're blues? [03:39] i just try to alternate [03:39] As in, we have them? [03:40] wallyworld_: No, blueprints are used for a *bikeshed* [03:40] :-D [03:41] so my vote would be to use the industry accepted term, namely Specifications [03:41] why reinvent terms that are in common usage? [03:41] Because five years ago it was cool. [03:41] Or something. [03:41] wgrant is just bitter because five years ago he was eight. [03:42] how could calling something a non-standard term that no-one else uses in that context have been considered cool? [03:42] ever? [03:42] wallyworld_: Don't ever go back before r3000 or so. [03:42] * thumper snorts [03:43] what's lurking there? anything interesting? [03:44] FAAARRRRKKKK! [03:44] buildbot is all read [03:44] or red even [03:45] maybe it's feeling Blue :-) [03:45] lp.registry.windmill.tests.test_distroseriesdifference_expander.TestDistroSeriesDifferenceExtraJS.test_diff_extra_details_blacklisting [03:45] lp.registry.windmill.tests.test_product.TestProductIndexPage.test_title_inline_edit [03:45] that's devel [03:45] thumper, yes, rs=me on deleting all windmill tests [03:46] :) [03:46] wallyworld_: Names were a fair bit less sane back then. [03:46] db-devel: test_uploading_collection (lp.soyuz.tests.test_binarypackagebuildbehavior.TestBinaryBuildPackageBehaviorBuildCollection) [03:46] rockstar: beep, denied. [03:46] thumper: you saying those tests are failing? [03:46] wallyworld_: failed on buildbot [03:46] what's the error? [03:46] wallyworld_: did you want to take a look? [03:47] thumper: in terms of names, I think we should do serious user research before doing another rename. [03:47] lifeless: ack [03:47] lifeless, I've said this before, but I think their cost outweighs their value. [03:47] *cough* Branches *cough* [03:47] thumper: tell me where to look - do i log into devpad to see the logs? [03:47] lifeless: there is enough other yaks to shave in blueprints [03:47] thumper: exactly. [03:47] wallyworld_: https://lpbuildbot.canonical.com/waterfall - click on summary [03:48] Blueprints mean something very valuable to the core users of blueprints: Ubuntu. [03:51] Also, of the three things I've sent to ec2 and the two that deryck sent to ec2 today, we have no emails to show for it. :( [03:52] rockstar: I tossed two branches at ec2 and got 2 mails [03:53] StevenK, lucky you. [03:53] rockstar: i've had stuff go missing lately too :-( [03:53] StevenK, still, 2 for 7 isn't great. [03:53] rockstar: Better than 0 for 5 :-P [03:53] wallyworld_, yeah, it really affects my productivity, because now I have to go back and babysit. [03:54] StevenK, not for me it isn't! :) [03:54] We should get out of testfix first, though [03:54] rockstar: i hear you. i've lost a lot of time as well, also chasing down "failures" which weren't [03:54] StevenK: Better than 0 for 9 :( [03:55] The last time it happened to me, there was a bug which I had introduced which was killing stuff [03:55] One of these branches has support for soupmatchers, which I want to use everywhere, so it'd be nice it that landed... [03:56] StevenK, so, there's some layer teardown stuff that's causing hangs in windmill. deryck and I finally reproduced it locally with my lazr-js branch. [03:57] Ah, excellent news. [03:58] thumper: well, there's no dog's ball anywhere that i can see. and the code in question hasn't changed recently [03:59] rockstar: interesting [03:59] rockstar: tried with trunk? landed a layer teardown change today [03:59] wallyworld_: does it pass locally for you? [03:59] trying now, updating as we speak [04:00] s/updating/building [04:00] have to stop my other db-devel system first and switch schemas @#@^^!$@ [04:02] thumper: hey, could you perhaps https://code.edge.launchpad.net/~lifeless/launchpad/edge/+merge/39594 ? [04:04] lifeless: there seems to be some waste there [04:04] * thumper comments [04:05] oh? [04:05] thanks [04:05] lifeless, well, ec2 merges into trunk, right? [04:05] rockstar: yes, but it would depend on the revno at the time you started ec2 [04:06] lifeless: done [04:07] thumper: thanks [04:07] lifeless, yeah. I'm sending them back to ec2. [04:07] lifeless, also, it seems that ec2 is still running postgres 8.3 [04:07] /o\ [04:08] thumper: >< our coding standards make your change _longer_ [04:09] :) [04:09] * lifeless ignores the nuts wrapping rules and just does the bzr way. [04:10] hmm, I've added another 80 lines of budget :) [04:13] thumper: apart from a LayerIsolationError, WFM [04:13] hmm... [04:13] wondering if we should just force a build to see if it works this time [04:13] that would be my plan of attach. at least the problem will be deffered for a little while :-) [04:14] i gotta learn to type better [04:16] I've forced devel [04:16] let's hope it works [04:16] wondering about the db-devel failure though [04:17] * wallyworld_ looks [04:18] something just doesn't look right with that build server. those errors don't appear coding issues do they [04:19] How can I have a webservice method that I can invoke in a stories test but not with launchpadlib against launchpad.dev? [04:21] I thought the one implied the other. [04:22] night all [04:23] jtv: Are you using launchpadlib in your test? [04:23] no, webservice [04:23] Your WADL is out of date. [04:23] make build [04:23] ! [04:24] * jtv tries that [04:24] wgrant: then again, I don't see the method on staging or qastaging either, and the revision's been on there for a while now. [04:25] jtv: You're not checking a frozen version of the webservice? [04:25] wgrant: I tried version='devel' [04:25] Which method? [04:26] IHasTranslationImports.getTranslationImportQueueEntries [04:27] Tested in lib/lp/translations/stories/webservice/xx-translationimportqueue.txt [04:28] Yippie, build fixed! [04:28] Project devel build (167): FIXED in 3 hr 36 min: https://hudson.wedontsleep.org/job/devel/167/ [04:28] Launchpad Patch Queue Manager: [r=thumper][ui=none][bug=662552] Re-use POFile edit permission info [04:28] across message sub-views. [04:35] wgrant: what I don't get is how that test can pass, yet I can't get at that method using launchpadlib interactively. I haven't played with launchpadlib much so maybe I'm just doing something wrong there. [04:38] jtv: Old-style webservice tests do not use WADL. [04:38] So they don't care how broken it is. [04:38] wgrant: so I neglected to update the wadl? [04:39] jtv: It's meant to be autogenerated. [04:39] grrr [04:39] But surely there is a way of getting our methods into the API, or else why do we have any [04:39] It is in the WADL, though. [04:39] Or in my WADL, at least. [04:39] And in my apidoc. [04:40] So maybe I'm just doing something stupid. [04:40] When did it land? [04:40] Is it deployed? [04:40] Days ago. It should be on staging. [04:40] Here's what I do locally: [04:41] l=Launchpad.get_token_and_login('gaga', service_root='https://api.launchpad.dev/', version='devel') [04:41] (Authorize in browser) [04:41] jtv: It's on qastaging's apidoc. [04:41] l.me.getTranslationImportQueueEntries() [04:41] AttributeError: 'Entry' object has no attribute 'getTranslationImportQueueEntries' [04:41] jtv: [04:41] grep -r getTranslationImportQueueEntries lib/canonical/launchpad/apidoc/ [04:42] Nothing. [04:43] Does 'make apidoc' take forever and say it's generating stuff for all three versions? [04:44] I haven't tried that. [04:44] (Didn't know it existed) [04:44] build depends [04:44] ... on it [04:44] Well, "make build" does take forever now, after a "make clean." [04:45] But that's probably buildout automatically sacrificing various items of wildlife that if finds around you. [04:45] Not the WADL generation. [04:45] wgrant: it's building that now [04:45] Aha. [04:47] And now my wadl shows the method. Wonder why "make build" didn't do that before. [04:47] Are these files under revision control? [04:47] They're not. [04:48] But this may change soon. [04:48] They take forever to generate. [04:48] Ah, I did this in an old branch. [04:48] So 'make build' probably won't regenerate them every time. [04:48] I guess the dependencies aren't complete. [04:48] Or make run would take even longer than it does already. [04:48] s/aren't/can't be without taking 15 minutes to start an appserver/ [04:50] Anyway, hopefully that will all work. [04:50] I need to run off and submit my final project. [04:50] wgrant: thanks as always! [04:50] Final… university project? [04:50] That's the one. [04:51] * thumper closes up. [04:51] wgrant: good luck—ace it like I did. :-) [04:51] bye thumper [04:51] Heh. Thanks. [04:51] Night thumper. [04:51] thumper: quick quest before you go [04:51] ====================================================================== [04:51] ERROR: lp.codehosting.vfs.tests.test_transport.TestLaunchpadTransportImplementation.test_win32_abspath (subunit.RemotedTestCase) [04:51] ---------------------------------------------------------------------- [04:51] _StringException: Text attachment: garbage [04:51] ------------ [04:51] [] [04:52] does that mean there's something screwed on ec2 wrt code hosting or something [04:52] an ec2 test run just failed with 1000 odd errors all like this [05:03] ? [05:05] wallyworld_: ah.... [05:05] not seen that before [05:06] sounds like something screwy [05:06] thumper: no matter. looks like there's something environmental going on [05:06] * thumper nods and leaves [05:06] i'll resubmit later. not having much luck with ec2 lately :-( have a good w/e [05:42] losa fyi: I'm trying to reproduce a bzr/codehosting bug, and I think I may have caused a python process to spin in an infinite loop on codehosting. [05:42] impressive [05:43] spm: if you're not too busy, I'd be quite curious to know if I really have done that... [06:51] Is anyone here on first-name terms with the webservice API? [06:51] What has it done to you now? [06:53] Well my method is documented now [06:53] !! [06:53] I mean, it's in the wadl and apidoc now, [06:54] _but_ [06:54] it doesn't appear on the objects that implement it and the interface it's in. [06:54] Instead, it appears in a separate API class named after the interface. [06:55] Which I suppose means that, instead of making objects implement the interface, I need to derive the objects' interfaces from my interface. [06:55] That's right. [06:55] Pretty unfashionable strong static typing, that. [06:56] It is unfortunate, but fortunately not difficult. [06:58] Unless it goes against the direction of all other dependencies and so is likely to introduce massive amounts of circular imports. [06:58] Well, true. [06:59] And guess what? That appears to be the case. :) [06:59]