[00:34] benji: thanks for your review of https://code.launchpad.net/~cjwatson/launchpad/api-build-builder/+merge/101162; would you mind landing it for me? [00:40] StevenK: enterpriseid probably needs a fixup approach [00:40] StevenK: otherwise it will become a source of circular imports [00:41] cjwatson: I'm loving your recent stuff :) [00:41] cjwatson: are you finding the LoC constraint problematic ? [00:42] wgrant: StevenK: wallyworld: one of you reprobates should be around and able to land cjwatson's branch ? [00:42] lifeless: I'm not sure; I alternate between finding it a tedious source of busywork and thinking it's a useful way to make the codebase less of an inaccessible monster [00:42] lifeless: sure, even though i'm not sure what a reprobate is [00:43] lifeless: so far I've had enough sources of crap to remove that it's not been too bad [00:43] (in practice) [00:43] cjwatson: which branch do you need landed? [00:43] wallyworld: https://code.launchpad.net/~cjwatson/launchpad/api-build-builder/+merge/101162 [00:43] wallyworld: thanks [00:43] np [00:43] wallyworld: lossely, scoundrels [00:44] lifeless: clearly i need to have a thesaurus handy when talking to you :-P [00:44] lifeless: Francis muttered something a few months back about possibly giving me PQM access; what oaths would I need to swear in blood for that? [00:45] cjwatson: an rt asking for it, webops can talk you through the bits [00:45] this branch will take me to joint second non-LP Canonical contributor :) [00:46] lifeless: OK, great, I'll put that in tomorrow [00:46] cjwatson: it's tomorrow, today. [00:46] * spm messes with colin's mind [00:46] my family are about to go away for a few days, so I fully expect to be on weirdo timezones [00:48] spm: that took a bit to understand... future ma. [00:48] *man [00:48] * spm bows, todays troll achievement, unlocked. *ding*. [00:49] spm: don't forget to do it daily; tis classic grinding [00:50] lifeless: indeed. this is the cross I bear [00:57] lifeless: If you mean fixup like _schema_circular_imports, I object. [00:57] StevenK: I imported enterpriseid_to_object from lp.services.mail and the world ended [00:58] lifeless: Sure, so it needs a solution, but I *hate* _schema_circular_imports as a pattern and do not want to repeat it. [00:58] sure [00:59] wallyworld: thesaurus : [00:59] https://www.google.com/search?q=reprobate&client=fs [01:00] He is calling us villianous [01:00] lifeless: i was being rhetorical before :-) [01:00] wallyworld: :P [01:00] lifeless is an expert at answering questions you weren't asking. [01:01] lifeless: I am curious whether you expect the LoC constraint to actually result in a net reduction of code by some point, I must say :-) [01:01] cjwatson: if it stops the bleeding, I will be content [01:01] * cjwatson nods [01:01] cjwatson: I expect the cultural values to shift over time [01:01] StevenK: perhaps he knows they need answering before we do [01:01] cjwatson: and once that happens, we've won [01:02] cjwatson: I see rules as, at most, a way to flag things that are likely problematic from the desired values [01:02] lifeless: How do you suggest we fix enterpriseid? [01:04] StevenK: there are three general patterns I know of; one is to have the module provide framework but no registrations - and then register into it [01:04] StevenK: the second is to use something lazy - e.g. lazy_import, or local imports in factory-functions, to let the module not import the world but still know about it [01:05] StevenK: the third is to put all the needed imports at the end of the module after the things it exports [01:05] e.g. [01:05] imports [01:05] public thing [01:05] public thing [01:05] more imports [01:05] I dislike the imports at the end as a pattern too. [01:05] my first inclination is to use the solution we have used elsewhere [01:06] and IFF that works worse for this case than those cases, hunt for something else. [01:06] but I'm lazy [01:06] * StevenK stabs _schema_circular_imports and then twists the knife. [01:08] wgrant: lp:~lifeless/launchpad/notificationapi is my sketch [01:09] aaaa [01:09] phew, it's not scanned yet [01:14] * lifeless goes hunting for test case code matching 'please run jobs' [01:18] yay, go's next stupid package manager reimplementation has debconf questions [01:19] there was a prior one? [01:20] I mean "next" as in "next in the line of programming language designers ignoring nearly 20 years of prior art and doing it all again themselves because fuck every other language" [01:20] Sounds like Go. [01:21] Those who don't understand history are doomed to repeat it. [01:22] (Is my bitterness about Go showing yet?) [01:26] StevenK: we (Canonical) seem to be jumping on the Go bandwagon? [01:27] Only because Gustavo is pushing it. [01:28] sure one person doesn't make policy though [01:31] * StevenK pulls the messaging indicator out, and stabs it through the heart. [01:32] The message it considers unread is OPEN in Thunderbird [01:33] wgrant: StevenK: https://bugs.launchpad.net/launchpad/+bug/30496 seems half-or-more-done, could I impose on one of you to either translate whats remaining for me, or do so in the bug ? [01:33] <_mup_> Bug #30496: [feature-request] status pages for uploaded packages

< https://launchpad.net/bugs/30496 > [01:33] StevenK: and Mark is too [01:34] dunno exactly how the wagon is structured. [01:34] lifeless: I'm trying to contain my bitterness since this is a public forum. [01:34] StevenK: indeed [01:36] lifeless: It wants something like http://qa.ubuntuwire.org/ftbfs/ but with more queue integration etc [01:36] And expiring old entries [01:37] That does expire old entries :) [01:38] so, can you edit the summary to reflect this ? [01:38] right now, my eyes glaze and roll back in my head [01:39] Welcome to Soyuz [01:39] heh [01:41] well, its more the description is awkward [01:41] I realy need to setup a launchpad qdb [01:41] and, of course, put everything spm says. [01:41] Haha [01:41] it should either describe the symtoms - the things users can't do - or if its going to ask for something, ask directly rather than in semi-passive phrases [01:42] :-) [01:42] are you saying that everything that spm says is a joke? [01:42] nigelb: but he's so entertaining [01:42] bigjools: everything he says in launchpad-dev. I suspect he does work in other channels ;) [01:42] ajmitch: Indeed! [01:42] wgrant: StevenK: will you guys be drive-by fixing https://bugs.launchpad.net/launchpad/+bug/249112 ? [01:42] <_mup_> Bug #249112: Changing security status requires clicking "This report is public"

< https://launchpad.net/bugs/249112 > [01:42] "work" [01:42] ;) [01:42] lifeless: Yes [01:43] StevenK has it fixed. [01:43] "Fixed" [01:43] its fixed? Can close ? [01:43] ... in a branch [01:43] lifeless: It's in progress. [01:43] Assign to me and mark it so if you wish, I'm working on it this week. [01:50] done [02:24] Oh [02:25] Awesome [02:25] We're still using YUI 3.0.0's cssgrids [02:25] Rather than 3.3's. [02:25] Deliberately. [02:25] I wonder why [02:55] jelmer: bzr daily builds have been failing for a while. Looks like 04_serve_address is in bzr.dev, so it can be dropped from the packaging branch? [02:55] hi wgrant [02:55] bit later even for jelmer in cet, i think [02:55] //hope [02:56] Hi poolie [02:57] I hope it's a bit late, yes :) [02:57] *late [04:54] wgrant: do you know, if i'm defining a new ws collection using export_as_webservice_collection, what version string do i use for export_as_webservice_entry and export method operation_for_version in order for it to all work? [04:55] atm, wadl generation fails because the collection adaptor generated by the ws infrastructure has no entry_schema defined [04:55] wallyworld: devel [04:56] lifeless: tried that :-( [04:56] i double check though since i've tried a few things [04:56] wallyworld: beta [04:56] It has to be the earliest version [04:56] tried that too :-( [04:56] diff? [04:57] one sec [04:58] wgrant: https://pastebin.canonical.com/64011/ [05:05] wallyworld: looking [05:06] wgrant: thanks. i need to rejig the web service stuff for services so that services can be exposed nicely via launchpadlib. hence i need a service collection which i didn't need before since i relied on traversal [05:07] yep [05:10] the LP api stuff is a lesson in how not to design an API, right from how it's coded to how it's presented [05:10] wallyworld: Are you sure this is going to work at all, given the different types that can be returned? [05:11] It'll work if you hardcode .services in launchpadlib [05:11] But not if you use API methods on IServiceFactory [05:12] wgrant: the return types will all implement IService [05:12] (saying lp.services.getByName('sharing') in launchpadlib will probably just return an IService, without the implementation-specific methods. [05:13] i've added .services to launchpadlib but it doesn't work without a collection being defined [05:14] wallyworld: You'll need to do something like PillarSet [05:14] Not PersonSet or BugSet. [05:15] Because of the different types. [05:15] i'll look at that. does the PIllarSet pattern allow something like lp.services['sharing'].foo() to be used? [05:15] That's exactly what it allows. [05:16] what's the collection name? 'pillars' is assume? [05:16] (note that these are client-side *Set definitions, independent of the server ones) [05:16] No collection. [05:16] bigjools: well, it had a set of specific goals, which it reached. [05:16] They're client side utter hideousness. [05:16] But they're the only way to do this sensibly. [05:16] bigjools: sadly, some things we consider very important weren't there as goals, metrics - nor deliverabels [05:16] wallyworld: Can't get past this: [05:17] LocationError: (, 'top_level_resources') [05:17] wallyworld: Even after removing the ServicesLink [05:17] lifeless: it's what happens when you leave someone alone to do something important [05:17] wgrant: that's exactly my problem, and i too removed the link [05:17] Ah [05:17] I thought you'd got past that [05:17] So I will dig more [05:17] lifeless: it's just crap on so many levels I despair [05:17] wgrant: so would i be wrong is assuming we cannot defined a new collection atm? [05:17] It's pretty elegant, pretty and nice. [05:17] It's just completely ineffective at anything ata ll :) [05:18] bigjools: nah, it wasn't because it was one person; it was because of fundamental approaches - like the choice to expose internal model objects [05:18] bigjools: which was, AIUI, mandated [05:18] anyhow [05:18] we have it now [05:18] and need to migrate gracefully [05:18] wallyworld: We can define a new collection. And lp.services.getByName('sharing') will work. But the object you get back will have no methods. [05:18] wallyworld: With the client-side hack, lp.services['sharing'] will work and have methods. [05:19] bigjools: it's one of the reasons i did the services stuff - to try and keep internal model things hidden [05:19] wgrant: client-side hack = pillarset? [05:19] I am wondering if the pragmatic approach may be to register all the services as top-level entries [05:19] wallyworld: Yes [05:20] wallyworld: fuck yes [05:20] lifeless: well I am sure that was the case but there's other human factors involved which I don't want to discuss on a public channel [05:21] bigjools: note though that it does use the restful infrastructure so there are some concessions [05:21] lifeless: anyway I'd love to see us migrating to something else. I doubt we'll find the resources though :( [05:21] wallyworld: restful is fine if used correctly [05:22] Mmm [05:22] wgrant: not sure about exposing all services as top level entities [05:22] It's not clear what "correctly" is [05:22] For our needs, correctly RESTful is difficult. [05:23] wgrant: services aside, i still wonder why defining a new simple collection of entities didn't work for me [05:23] wallyworld: It shouldn't fail to build, and I'm trying to work out why it is. [05:23] But it also won't be useful once it does build. [05:24] wgrant: is that because it checks the ws.op client side and complains if it doesn;t think there's a server side end point to handle it? [05:24] wallyworld: I believe so. [05:24] It will respect the collection's resource type [05:24] well, what it believes it to be [05:25] What you declared it to be :) [05:25] so it's static [05:26] Basically. [05:26] You have to explicitly downcast. [05:26] By saying lp.load('someuri') [05:26] which i'm trying to hide from the user [05:27] And it's not possible with the current implementation. [05:27] Unless you do something like PillarSet [05:27] Which implements client-side URL generation [05:27] Which it then spits into lp.load directly. [05:27] so i wonder if it's worth doing this then. i guess so to make it user friendly [05:34] wallyworld: Oh, the error message was obvious all along. [05:34] i didn't know how to fix it though [05:34] or why the entry_resource was None [05:35] which was the root cause [05:35] You are truly ignorant of the world's cultures. "LocationError: (, 'top_level_resources')" is TALESlish for "IService isn't imported by lp.registry.interfaces.webservice" [05:35] Everyone should know that. [05:35] (I suspect it's hiding an AttributeError or KeyError under that LocationError, as TALES is wont to do) [05:36] it wasn't before either, but i guess adding a collection made it matter [05:36] It wasn't involved in the root before. [05:37] I suspect only the root relies on the webservice registrations without anything else. [05:37] Also, fuck TAES. [05:37] TALES [05:38] wgrant: right, will try that. thanks for looking. i wasted too much time on that :-( [06:13] lifeless: is there any support in testtools for re-running tests under different setup conditions? === almaisan-away is now known as al-maisan [06:20] bigjools: scenarios [06:20] eg lib/lp/poppy/tests/test_poppy.py [06:21] I knew someone would say that [06:21] Although that uses bzrlib's, not testscenarios [06:22] there's a clone_test_with_new_id thing, but no explanation of how to use it [06:22] wgrant: turns out i don't need to do any client side hacking - it all just works without [06:24] ah, it depends [06:24] launchpad.services['sharing'].foo() works [06:24] launchpad.services('sharing').foo() doesn't [06:25] that's pretty sad [06:25] Why would services('sharing') work? [06:28] wgrant: it doesn't since it checks that the method exists and it doesn;t as far as the client knows [06:28] but why does services['sharing'] work [06:28] it must not do the same client side checks [06:28] bigjools: testscenarios [06:29] bigjools: (python-testscenarios) [06:29] lifeless: any online docs? [06:29] bigjools: what test-poppy is using is an earlier edition of that [06:29] http://pypi.python.org/pypi/testscenarios [06:30] thanks [06:32] bigjools: obviously if you have q's I'll be happy to answer [06:32] lifeless: doc seems good so far thanks! [06:43] lifeless: ok question: the scenarios list isn't explained very well, what is supposed to be in it? [06:43] a list of (name, dict-of-things-to-apply) tuples [06:43] e.g. ... test.scenarios = [ [06:43] ... ('scenario_1_0', {}), [06:43] ... ('scenario_1_1', {})] [06:43] the dict maps attribute names to values [06:43] e.g. [06:44] ,lifelessall, [06:44] grear [06:44] {'foo': 'bar'} [06:44] lifeless: apply to what? [06:44] will make self.foo be 'bar' once the test is setUp [06:44] ok [06:45] see the dynamic scenarios example [06:45] that makes two scenarios, one that sets self.hash=md5 and one that sets self.hash=sha1 [06:45] right [06:46] wgrant: so setting collection_of=None and forcing it to lookup what the resource entry provides seems to do the trick. but a trap that [] and () both return service resources which behave differently [06:46] wallyworld: () isn't documented and really shouldn't work at all [06:46] I don't know why it does [06:47] I'm surprised collection_of=None works [06:47] Wait. [06:47] collection_of=None where? [06:47] i'll keep using [] in my test code and at least launchpadlib will work in both case [06:47] in the ServiceSet class [06:47] Oh [06:47] I thought you were doing this without the client-side hack :) [06:47] 16:22:38 < wallyworld> wgrant: turns out i don't need to do any client side hacking - it all just works without [06:47] i am [06:47] confuse [06:48] d [06:48] but i need to define a client side set [06:48] but only the one - not one for sharing, one for ... [06:48] That's the client-side hacking we wanted to avoid :) [06:48] there's also BugSet too etc [06:48] Since it requires a client upgrade, which is less than ideal. But it's probably unavoidable. [06:48] Sure. [06:49] Precedent doesn't make it less nauseating. [06:49] so i thought that having a ServiceSet was kosher (like BugSet etc). and that avoiding SHaringServiceSet, FooServiceSet like was done for pillars was not using any hacks [06:50] It's accepted, but still evil since it means we can't evolve the webservice without client changes. [06:50] Which sort of destroys the entire purpose of WADL [06:50] yes. i didn;t think we could anyway [06:51] Hm? [06:51] As we saw before, you can add new types, methods, top-level entries, top-level collections without changing the client. [06:51] But not this sort of top-level collection [06:51] given our current implementation (ie BugSet, XXXSet all defined in the client side lib) i thought we needed to make that compromise [06:51] 'all' [06:51] Aren't there only 3 of them? [06:52] Bug, Pillar, Product, Person, Distribution etc [06:52] Those are only needed because we wanted foos['somefoo'] to work [06:52] Collections with normal methods don't need one of those client-side hacks. [06:53] so it's really a limitation of our infrastructure [06:53] since all the client side classes do is define _get_url_from_id [06:53] and that could be put in the wadl [06:53] as an attribute [06:53] In this case we want services['someservice'] to work, partly because due to lazr.restfulclient limitations the ServiceFactory's methods will only return objects with IService's methods [06:53] Yes. [06:53] there is a bug [06:53] about allowing arbitrary interfaces on each api object [06:54] rather than predicting the interface for collection elements [06:54] But in this particular case we only want that workaround to get that functionality to workaround another limitation [06:54] lifeless: Right, that would work too [06:54] so for now i'll do a new version of launchpadlib === mabac__ is now known as mabac [08:00] good morning [08:10] bigjools: it making more sense now? [08:11] lifeless: yes thanks - rvba is on it :) [08:11] Hi lifeless. [08:12] good morning [08:16] cool [08:19] lifeless: was this about the testscenarios thingy? [08:27] yah [08:32] lifeless: I've been bitten by 872887. Any advice on how to use testscenarios with nose? [08:38] rvba: use load_tests with nose2 [08:38] rvba: or don't use nose, its terrifying [08:39] rvba: or figure out how to make nose do a custom callback before test execution, it probably has a hook somewhere for that, it just won't be compatible with anything else [08:43] lifeless: ok, I'll try to do that thanks. (too late, we're already using nose extensively. It's worked fine so far to easily integrate django tests, twisted tests, etc…. But maybe it's terrifying inside ;) ) [08:48] rvba: it is; are you using nose's 'twisted' thing, or testtool's twisted async module ? [08:49] rvba: (it is terrifying inside) [08:49] * lifeless *far* preferrs using testtools.run / subunit.run to nose. [08:50] lifeless: testtool's [08:51] rvba: phew :)( [08:51] s/(// [08:51] testtools* === jam1 is now known as jam [09:02] lifeless: fwiw load_tests (module level function) is called without any argument by nose. [09:03] yay, totally incompatible [09:04] indeed. [09:05] I guess I'll have to resort to nose-specific hooks. [09:05] * rvba starts to hate nose. [09:05] http://pypi.python.org/pypi/discover - see load_tests on that page [09:05] rvba: welcome to my world [09:06] I guess nose stinks :) [09:12] lifeless: in fact, nose is simply treating load_tests as a module level test function. [09:14] yup [10:05] jam: Hi (I'm from the Launchpad team). I'm struggling quite a bit trying to get nose working with testscenarios and I see that you've been trying to do that as well (https://bugs.launchpad.net/testscenarios/+bug/872887). Do you happen to remember what you ended up doing to generate the scenarios? [10:05] <_mup_> Bug #872887: TestwithScenarios incompatible with nose 1.1.2 < https://launchpad.net/bugs/872887 > [10:24] argh doctests with sampledata. hate hate hate [10:24] things I do not like include reading current-dev.sql to discover exactly what should be guaranteed in a given test [10:26] also, the "bob the builder" joke in the test suite gets very old after (a) you've read it a few hundred times (b) your child watches enough of it that you get the theme tune in your head every time you see it [10:27] cjwatson: current-dev.sql has bad data in it [10:29] indeed, but I'm not quite sure of your point :) [10:30] whats guaranteed in a test isn't well demonstrated by current.sql [10:30] (and current-dev.sql isn't used at all for tests ;)) [10:31] uh, current, ok. but I meant doctests that are basically pretty-printing objects straight out of the sampledata. [10:31] okies [10:31] kill crush destroy [10:32] quite. need to make this branch pass for now, though :) [10:34] Is this webservice doctests for build.builder? [10:34] Yes [10:34] I've pushed a fix [10:37] http://bazaar.launchpad.net/~cjwatson/launchpad/api-build-builder/revision/15073 is no more evil than the crap around it, I think === al-maisan is now known as almaisan-away [10:54] cjwatson: A few hundred, you say? I was sighing about it after the tenth time, and I have no young child. :-) [10:55] I don't know whether to blame Julian or Celso for that particular piece of Soyuz. [10:58] you sure it's not a Daniel thing? [10:58] I believe it was Daniel [10:58] most of the puns are his fault [10:58] Oh. [10:59] Is he still in Melbourne? [10:59] I have a few ... debts to settle. [10:59] Silverstone, not Stone [10:59] Oh, bah. [10:59] AFAIK he was never in Melbourne [10:59] arguably transitively my fault for feeding him salami while he wrote chunks of Soyuz at my house, I guess [10:59] Daniel Stone was, though. [11:17] rvba: basically, I stopped using nose w/ testscenarios... [11:17] Specifically, nose doesn't seem to support the python 2.6+ test feature of "load_tests" [11:18] if you look here: http://bazaar.launchpad.net/~jameinel/u1db/c-oauth/view/head:/u1db/tests/__init__.py#L304 [11:19] That is the adapter function I wrote, which allows you to do: "from ... import load_with_scenarios as load_tests [11:19] and then python 2.6/7 unittest and testtools itself [11:19] will load the tests in your module and parameterize them with testscenarios [11:19] nose doesn't support having "load_tests" defining how to extract tests cases from a module [11:20] and I never figured out how nose does that sort of thing [11:21] jam: thanks. I've been trying to use nose-specific stuff to generate the scenarios but no luck so far… [11:22] rvba: testscenarios lets you do the parameterization by either inheriting from its TestCase or by using a load-time hook [11:22] I tried the test case approach, but it was broken for a lot of cases [11:22] a lot of runners, I should say [11:23] specifically, there are ones that assume the number of tests that will be *run* with one call to .run() is always going to be 1 [11:23] but the test-case based approach creates them on-the-fly [11:23] so I recommend going with the load-time version [11:23] but I don't know how to get nose to not try and load your tests normally, and only accept them through your loader [11:23] Yeah, but an assertion in nose makes that approach impossible to use with nose. [11:23] rvba: exactly [11:24] I even tried to monkey patch nose to remove that assertion but then there is something wrong with how the tests are run. [11:27] jam: anyway, thanks for sharing the result of your experiments. I'll try again for a few hours and I guess I'll not use testscenarios if I can't find a proper way to use it with nose. [11:27] rvba: have you tried to file a bug/hit up the mailing list? I know Jason's been working on putting out some nose2 releases. Helpful guy. [11:28] rvba: yeah, I didn't need much from a test runner. I ended up writing a plugin to hook into bzr's test infrastructure, as it is the best I've found yet :) === salgado-afk is now known as salgado [11:28] rick_h: I haven't tried that. I'll look into it thanks. [11:28] (partial loading, progress bars, regex matching, etc, etc.) [11:35] jam: the only runner I know that assumes one TestCase == 1 test is nose [11:35] jam: everything else i"ve see, however awful, honours the protocol [11:36] lifeless: sure, though everything else is subtle shades of broken in different ways :) [11:36] jam: I'd love bug reports; I commented on all your current ones I think [11:36] lifeless: yeah, its mostly problems with the actual runners [11:36] * lifeless goes back to bed. [11:36] like not great support for --verbose, etc [11:38] I love the nose --with-id [11:38] use that one all the time [12:28] hey [12:29] I've seen a fair bit of chatter about testtools & friends over the last couple of days [12:29] I haven't followed it all closely, but I'd like to know: how can I help? [12:45] cjwatson: i've merged your branch directly since the fixes you did were just test fixes so you can update locally and your changes will be there [12:51] wallyworld: excellent, thanks [12:51] np [12:52]