=== Ursinha-afk is now known as Ursinha [00:23] thanks python 2.7 [00:23] String or Integer object expected for key, unicode found [00:27] What for? [00:27] Some fake dict? [00:27] DBM [00:28] https://bugs.launchpad.net/testrepository/+bug/775214 [00:28] <_mup_> Bug #775214: On python 2.7: String or Integer object expected for key, unicode found < https://launchpad.net/bugs/775214 > [00:29] Ah :( [00:35] wgrant: the fun of it is it that one can expect python 3.x to be diametrically opposed [00:35] lifeless: Surely 3.x expects bytes/int? [00:37] (if not I may cry) === micahg_ is now known as micahg [00:49] wgrant: for a dict interface? [00:49] wgrant: I haven't started a 3.x port of testrepository yet. [00:49] lifeless: Yes. [00:49] I may tonight or something; right now am just yak shaving my way back up to a working desktop [00:49] Since Unicode makes very little sense in DBM, surely... [00:50] wgrant: no less than it does in the sql interface postgresql uses [00:50] lifeless: Hm? postgresql does Unicode... [00:51] wgrant: yes [00:52] dbm doesn't do collation [00:52] so implementing unicode for a user is simply encode on the way in, decode on the way out. [00:52] no good reason not to support that [00:52] Mmm [00:56] the only reason not to that I can think of is that its slow [00:56] but thats python's choice of string model more than anything else [02:34] lifeless: https://lp-oops.canonical.com/oops.py/?oopsid=1944DQ278 [02:35] Heat updates are expensive. [02:35] (although this seems like a particularly inefficient way to do it) [02:35] There was a change for that in the deployment report I noticed [02:36] Yeah, but different. [02:36] Of course it is. :-) [02:43] wgrant: https://lp-oops.canonical.com/oops.py/?oopsid=1944DQ278#repeatedstatements should be helped by abels change? [02:44] lifeless: I thought that looked specific to Bug.markAsDuplicate. [02:47] wgrant: I only read the cover letter [02:47] so it may need an equivalent change [02:49] lifeless: Except this is somewhat deeper :( === lifeless changed the topic of #launchpad-dev to: PQM in RC mode | https://dev.launchpad.net/ | On call reviewer: - | https://code.launchpad.net/launchpad-project/+activereviews [03:47] lifeless: How well do you know Storm's codebase? [03:49] I am considering digging around to see why inherited References don't have the correct class, while normal properties do. [03:49] But the descriptor/metaclass stuff is not entirely clear. [03:57] I can generally find what I am looking for [03:57] not far enough along to have a sense of how to tackle things tastefully [03:57] Ah. [03:57] k [03:57] so you have a reference on class A [03:58] and a concrete class B [03:58] Yes. [03:58] and the reference cls handle when you access B.reference is A [03:58] The Property descriptor handles that nicely. But Reference does not. [03:58] or [03:59] is it that you have a reference into the class hierarchy ? [03:59] PackageBuild.archive is inherited by BinaryPackageBuild.archive. [03:59] 'BinaryPackageBuild.archive == foo' creates a query for PackageBuild instead. [04:00] BinaryPackageBuild.archive_id == foo works fine. [04:00] Because the descriptor recreates the column. [04:00] With the correct class. [04:01] now [04:01] this is with sqlobject sugar right [04:01] ? [04:01] No. [04:02] Raw Storm. [04:02] PropertyPublisherMeta looks interesting [04:02] Well. BinaryPackageBuild is SQLBase, but PackageBuild and SourcePackageRecipeBuild are native. [04:03] And I tested it with native non-LP Storm. [04:03] PropertyRegistry [04:03] Mm, it's not that interesting. [04:03] I think that's mostly for string property lookups. [04:04] Yes. [04:04] class PropertyRegistry(object): """ An object which remembers the Storm properties specified on classes, and is able to translate names to these properties. """ [04:04] So not relevant here. [04:04] It's just the descriptors. [04:04] lifeless: Compare Property.__get__ and Reference.__get__ [04:05] Property uses _get_column, which creates a new PropertyColumn based on the right class. Reference just returns self. [04:05] (where obj is None, obviously) [04:09] yes [04:14] Project windmill build #227: STILL FAILING in 1 hr 3 min: https://lpci.wedontsleep.org/job/windmill/227/ [04:19] lifeless: So, BuildFarmJobSet.getBuildsForBuilder [04:19] This method is The One. [04:20] wgrant: you have a fix for storm ? [04:20] Did you envision any particular solution to it? [04:20] No, I just worked around it by using _id. [04:20] wgrant: what file is that in ? [04:20] wgrant: did you file a bug ? [04:20] lib/lp/buildmaster/model/buildfarmjob.py [04:20] I didn't. [04:20] wgrant: please file one [04:21] I will. I spent most of Friday afternoon dealing with this so was pretty displeased by the end, so didn't end up filing a bug. [04:22] Bug #682989 is probably the same thing. [04:22] Or rather similar. === Ursinha is now known as Ursinha-afk [04:26] wgrant: so that query [04:26] wgrant: one option is to union [04:26] That's what I was thinking. [04:26] But the sort may be slow. [04:27] I was going to half-perform the migration in a transaction on DF and see what happens. [04:27] But populating BPB takes forever. [04:29] lifeless: Although I guess key-based offsets with a union could be done rather cheaply. [04:29] s/rather/very/ [04:31] yes [04:32] the key question is whether sql will sensibly do partial iteration on each input [04:32] UNION ALL should let it do this [04:32] I'm looking at uses [04:32] builder.getBuildRecords is one [04:32] its binary_only flag will permit a single-table lookup [04:34] I believe that should be the only user. [04:35] looks like it [04:35] sadly getBuildRecords is used as a name for other things [04:35] Yes. [04:36] It's on like 5 or 6 other classes. [04:36] I don't think its sensible to report back to the start of time on builders [04:36] It's not essential, no. [04:37] I'd considered adding a threshold if the sort was too slow. [04:37] But it may not be. [04:37] If we consider dynamic provisioning of builders [04:37] it becomes actively undesirable [04:41] hmm, no wallyworld today [04:41] This week. [04:41] Or next. [04:41] He's in Budapest. [04:41] oh right [04:45] anyone know how to land stuff on oops-tools ? [04:48] pqm-submit? === almaisan-away is now known as al-maisan [05:58] grrr [05:58] PATCH -> no idea what was changed [05:58] Oh? [05:58] in oops-tools [06:02] * wgrant vomits a bit. [06:02] The old bug subscription is awful :/ [06:02] Not obvious how to fix without reverting and deconflicting. [06:03] The old bug subscription *JS* is awful, that is. [06:03] // Determine if this is a dupe. [06:03] var is_dupe; [06:03] var icon_parent_div = icon_parent.get('parentNode'); [06:03] var dupe_id = 'dupe-' + person.get('css_name'); [06:03] if (icon_parent_div.get('id') === dupe_id) { [06:03] is_dupe = true; [06:03] LPCIBot: I'm also not sure about the recent Jenkins picture change [06:03] } else { [06:03] Doh [06:03] is_dupe = false; [06:03] lifeless: ^ [06:03] } [06:03] StevenK: the new logo? [06:03] Ignoring the whole if (foo) bar = true; else bar = false; thing, that's still a bit bad. [06:03] lifeless: Yah [06:04] I like it [06:04] The old one was mostly one colour, this is ... too red [06:04] StevenK: Agreed. [06:05] And French, rather than British. :-P [06:05] lifeless: I don't think this needs reversion. [06:06] Its only effect is minor. [06:06] Everything else is still fine. [06:06] And it doesn't revert cleanly. [06:06] Although that's fixable... [06:06] are you familiar with curtis question work? [06:06] the last remaining qa item [06:07] untestable, I believe, but let me check. [06:07] Oh dear. The last 55 windmill builds on Jenkins have failed. [06:07] StevenK: Yes, there is that one test which fails only in Windmill :( [06:08] lifeless: It is unused, but there are QA instructions in the bug. I will grab spm. === al-maisan is now known as almaisan-away === almaisan-away is now known as al-maisan [06:43] wgrant: thanks for checking out the latest generate-contents-files... did it all work then? [06:44] jtv: Yup! [06:44] All looks good. [06:45] What a relief. [06:47] wgrant: maybe we should just schedule that bugger then [06:47] ah no, downtime [06:51] :( [06:51] Ja. [06:52] Hmm that means about the same thing in all 3 Indo-European languages I can think of that have that word, so I get your drift. [06:53] In Spanish, however, it could be interpreted as you laughing at me which would be less than kind. [07:00] :( [07:01] By the way, wgrant, isn't this a holiday for you? [07:01] Grrr this edge connection isn't very suitable for actual work. Lightning seems to have fried my dsl modem because I was too lazy or worried about signal quality to loop the phone line through the UPS. I'll try once more and if not successful, get a new one. [07:02] jtv: No, not a holiday here. [07:02] Should it be? [07:02] Labour Day. [07:02] In the UK, maybe. [07:02] Ours was in March. [07:02] NSW's is in October. [07:03] QLD is today, though. [07:03] Here, certainly. The reasoning seems to be that there must be a day without labour specifically for the occasion of labour day, and just having the day off because it's Sunday isn't enough. [07:04] It's "May Day" in the UK, isn't it? [07:04] Quite possibly. [07:05] Northern Terrority also have a 'May Day' public holiday. [07:05] May Day is when pilots stop working and just let their planes drop to the ground, right? [07:05] NT doesn't exist. [07:07] wgrant: Wha? [07:08] This could be bad news for any inhabitants the NT may have. [07:09] Meanwhile, how can it be a national holiday for Graham and Julian but not for Gavin? [07:09] It has a population of like 200000. It doesn't exist. [07:09] jtv: Swap day? [07:09] Maybe. [07:09] Project windmill-devel build #1: FAILURE in 1 hr 4 min: https://lpci.wedontsleep.org/job/windmill-devel/1/ [07:09] wgrant: no, the list I'm looking at would still show a national holiday. [07:10] wgrant: By your reckoning, the small towns dotted around Australia don't exist either. [07:10] Perhaps he just forgot. [07:10] StevenK: They're not a whole semi-state. [07:10] It's a terriority! [07:10] Right, it's like a state except more pathetic. [07:11] We should be trying to emulate them! They have less politicans. [07:11] True, true. [07:14] And windmill-devel failed, surprise. === al-maisan is now known as almaisan-away [08:10] StevenK: I'm looking into the "synchronize updates" button for the packages diff page. Do we have code landed to synchronize a package? [08:11] The code has existed for ... eons [08:12] Æons? [08:13] night all; [08:14] StevenK: damn, connection bounced again... where does this code live? [08:14] Project windmill-db-devel build #228: STILL FAILING in 1 hr 5 min: https://lpci.wedontsleep.org/job/windmill-db-devel/228/ [08:15] * jtv traces through the call chain [08:16] jtv: I'm not what the button calls, but likely IArchive.syncSources() [08:16] Ah, I thought that was a recent development. Thanks. [08:42] What review type is for a release critical? rc or release-critical or what? [08:43] release-critical [08:43] I think [08:46] https://code.launchpad.net/~stub/launchpad/pending-db-changes/+merge/59640 if the RM is in the house. I'm out of here to enjoy this holiday :) [08:49] Hm. We don't have RMs anymore? [08:49] lifeless is the perma-RM. [08:52] Heh [08:52] good morning [08:57] Hi wgrant. If you get a chance, would you take a look at https://code.launchpad.net/~rvb/launchpad/change-perm-sync/+merge/59341 ? I've modified the packages copier method carefully but since the modification is rather sensitive, I'd be happy if you could take a look at this. You will see that I've solved the problem about passing a None person to the copier method by avoiding this case with code in the view code. Please also take a l [08:57] ook at adeuring's remark in the MP's comments. [09:03] rvba: Looking. [09:03] wgrant: great, thank you. [09:05] 405 + :param strict_component: True if access to the specific component for [09:05] 406 + the package is needed to upload to it. If False, then access to [09:06] 407 + any package will do. [09:06] Access to any *component* will do. [09:06] wgrant: right ... so I'll have to change the doc in the other method as well (where it came from) [09:06] Ah. Indeed. [09:07] rvba: Hmm, so anonymous users are prevented from doing everything by line 19 of the diff? [09:08] wgrant: yes, this protects the call to sync_source. [09:09] rvba: That's a bit riskier than I'd like :/ [09:09] wgrant: yeah, I confess I'm not really happy with that either. [09:09] rvba: Could you make checkCopy take a check_permissions arg in addition to the person? [09:09] and set this to False by default? [09:10] No, True. But having check_permissions=True and person=None will forbid the copy. [09:11] You will have to change the existing callsites to use check_permissions=False, but then at least the dangerous action is explicit. [09:11] And this is the very definition of dangerous action. [09:11] ok, I've done it this way to avoid having to change the call sites ... but your solution is bar better ... [09:12] If there is one security hole we don't want in Launchpad, it's probably in this method. [09:13] all right. I'll do this. [09:13] Thanks. [09:13] wgrant: Thank _you_ ;) === henninge changed the topic of #launchpad-dev to: PQM in RC mode | https://dev.launchpad.net/ | On call reviewer: henninge | https://code.launchpad.net/launchpad-project/+activereviews === almaisan-away is now known as al-maisan [09:56] hi together, I want to setup a intranet launchpad is there a howto? with public/private projects, account management etc [10:15] MonkZ1: I am not aware of such a howto. It certainly is not supported "out of the box" but will require some work. [10:16] MonkZ1: I know that others have done it but I don't know how sucessful they were. [10:19] henninge: there is no need for "out of the box" (mach ich gern). just need a light guideline [10:21] MonkZ1: I am not aware of anthing like that, at least not on offcial LP pages. === al-maisan is now known as almaisan-away [10:23] henninge: ok thanks === almaisan-away is now known as al-maisan [10:30] henninge: the "private project"-feature is free available? [10:30] MonkZ1: AFAIK we have private bugs, private branches and private teams. [10:30] henninge: ok nice [10:31] MonkZ1: these are available to commercial subscriptions. [10:31] :( [10:32] MonkZ1: https://launchpad.net/+tour/join-launchpad#commercial [10:33] henninge: so i need to subscribe to get the possibility of a private * in a intranet environment on a selfhosted launchpad? [10:33] MonkZ1: no [10:33] MonkZ1: sorry, misunderstood you [10:34] MonkZ1: you can use your own instance as you like it. [10:34] MonkZ1: you get the exact same source tree that Launchpad.net is using. [10:34] MonkZ1: so all these features are available. [10:34] henninge: aaah k this was the bit i need [10:35] henninge: thx [10:35] bitte schön ;) [10:35] henninge: ach dann hätten wir auch auch deutsch quatschen können ;) [10:36] bye [10:40] MonkZ1: no, this is an English language channel ... :-P [10:40] Tschüß [11:59] jtv: So ArchiveDependency would have a series at both ends? [12:17] wgrant: hadn't thought of that--it'd be a bit of a misnomer then wouldn't it? [12:17] Then again, with component & pocket already in there... [12:18] jtv: You propose it has at least one series, though? [12:18] And if there is one there must surely be two. [12:19] Then yes. [12:19] Otherwise how would it work? [12:19] ArchiveDependency on Ubuntu Natty... for which series? [12:20] Yes, I see that now. [12:20] Great. [12:20] Then I mostly agree with your idea of merging them. [12:21] Phew. I thought that was going to be a stumbling block. :) [12:21] I haven't run through all the cases yet, but it seems like it should work... [12:29] wgrant: the operative question remains of course: would it make our lives any better? There's probably enough flexibility in that schema to cause us a few headaches. [12:31] jtv: I'm not sure it adds significant complication. [12:31] It mostly makes things simpler. [12:31] And a little less inflexible. [12:31] Very, very relieved to hear that. [12:32] Apart from the fact that it'd take actual work to implement, of course. [12:33] I'm trying to think how the implicit primary archive dependency would work. [12:33] I guess it would be implicit per-series. [12:33] What are the problems with making it explicit? [12:33] ie. if there's not an explicit one for that series and the primary archive. [12:34] Well, maintaining that automatically is difficult. [12:34] Added a new series to Ubuntu? Need to add 20000 new ArchiveDependency rows. [12:35] That is rather a lot, yes. What are all those archives? [12:35] PPAs. [12:35] Lots of PPAs. [12:37] Part of the idea though is that for the general case, an archive dependency doesn't _have_ to be restricted to a series. [12:37] That's where the headaches come from. [12:37] Right. [12:37] But it might just help avoid that per-series cost. [12:39] My original proposal or yours will both work fine. Yours is probably a bit cleaner. But maybe more expensive. [12:39] But...; ArchiveDependency.component and ArchiveDependency.pocket really need to be series-specific anyway. [12:39] So we might as well merge AD and DSP and solve everything. [12:45] henninge: saw the request to open Oneiric translations? I'm on feature rotation and rather busy... are you on maintenance by any chance? [12:45] Where does that go? And what's involved? [12:45] wgrant: I'd suggest you reply to jtv's mail with how to merge AD and DSP? [12:46] StevenK: I need to think about it. [12:46] The trivial solution is sort of obvious. [12:46] But it's not clear that it's perfect. [12:47] wgrant: you're asking about the translations opening? It's a matter of running a script, much like init-from-parent, and it should be automated similarly. [12:47] wgrant: My thought was DSP at *least* needed to grow pocket [12:47] StevenK: Yes. [12:47] StevenK: Julian disagrees. [12:47] Or we create a Series table that maps all this crap [12:47] But he's not here today, so stuff him :) [12:47] jtv: Right. [12:47] jtv: It's fairly cheap nowadays with sharing and stuff? [12:48] y [12:48] jtv: Which list/address was notified? [12:49] wgrant: Okay, will you at least agree to replying with "Oh, here's this trivial solution I thought of." [12:49] wgrant: the former translations team still has a mail alias [12:49] StevenK: Add two series fkeys to ArchiveDependency. Adjust code to cope. Done! [12:50] (delete DistroSeriesParent too) [12:50] (bonus points for making DSD point to an ArchiveDependency) [12:50] jtv: Ah. That's a bit sad. [12:50] wgrant: Bah, we just added it [12:50] StevenK: Yes, and that's why I argued about it. [12:51] Because it was premature and ill-thought. [12:51] And I wrote a 2,000 line branch to *use it* [12:51] Sure. But it's mostly reusable. [12:51] Just need to change the class it looks at. [12:51] It's the ArchiveDependency code that needs to change significantly. [12:51] DSP-using code just has to use the other class. [12:53] wgrant: I'm surprised to hear we were thinking along such similar lines all the time. [12:53] (including the "premature and ill-thought" bit) [12:53] Heh. [12:54] Your combined solution is better than the one I had come up with. [12:54] But I knew the existing one was entirely inadequate :) [12:56] This is why I like non-trivial discussion before making large changes :) === benji changed the topic of #launchpad-dev to: PQM in RC mode | https://dev.launchpad.net/ | On call reviewer: henninge, benji | https://code.launchpad.net/launchpad-project/+activereviews [13:00] jtv: I was pinging you to ask you about that ... ;-) [13:00] Ah! [13:01] jtv: do we have it documented somewhere? [13:01] Errr yes somewhere I think... [13:01] How hard is it? [13:01] Can NRCP just say to poke a LOSA to do it? [13:01] wgrant: not hard at all. [13:02] It used to be a big deal where lots of things could go wrong and bugs would surface with merciless savagery. [13:02] Yeah. [13:02] But nowadays, yes, a LOSA could do it or better yet, a job runner. [13:02] And you had to take down LP for 12 hours. [13:02] Those were the good old days. [13:03] 12 hours? you kid had it easy [13:03] jtv: also, I am wondering if "Mail rosetta@..." is still a good instruction for the Ubuntu guys. [13:03] henninge: we were just talking about that--you're right, it isn't. [13:04] right [13:04] I have my cursor on the NRCP edit link :) [13:05] Is there a LOSA doc I can link to? [13:05] It should be a simple matter of running scripts/copy-translations-from-parent.py -d ubuntu -s and, once it's done, enabling imports for the series. [13:05] Nothing under LPHowTo like there is for branch-distro? [13:07] jtv, wgrant: I can see to it that it happens this time around and wirte up a doc for it. [13:07] I'm searching the wikis atm [13:07] (which would mainly include those two instructions) [13:08] jtv: cool. I have to relocate right now. Please let me know what you find ... ;-) [13:08] OK [13:16] jtv: https://answers.launchpad.net/launchpad/+question/154984 <- is that just a matter of changing the series on +admin? [13:17] wgrant: the great irony of the translations opening is that the code relies on DistroSeries.parent_series. [13:17] wgrant: looking... [13:17] jtv: Of course :) [13:38] Anyone around that might be able to help me figure out why I can't get Launchpad running locally? [13:42] huwshimi: What does it do? [13:42] rvba: What's this -updates thing? [13:42] wgrant: I can't make a connection between the browser and the local server. At least that's what I suspect. [13:43] wgrant: launchpad.dev just times out [13:43] wgrant: the card says exaclty this "Synching to a released series should put packages in -updates" ... but I haven't talked to anyone yet (and Julian is only back tomorrow) [13:43] wgrant: and the server does not log any events [13:43] wgrant: https://code.launchpad.net/~rvb/launchpad/sync-to-updates/+merge/59653 [13:43] huwshimi: ping launchpad.dev [13:43] Which address is it trying? [13:43] rvba: I saw the diff, and I think the idea of hardcoding it like that is silly. [13:44] It is probably what Julian wants. [13:44] But it is still silly. [13:44] wgrant: :) I'll be glad to hear about any improvement you would care to suggest. [13:44] wgrant: Ah, it's trying to use my ip address at home [13:45] wgrant: I am a long way from that router :) [13:45] wgrant: I should have noticed that before [13:45] wgrant: not that the destination pocket was previously hardcoded to "release". [13:46] s/not/note/ === Ursinha-afk is now known as Ursinha\ === Ursinha\ is now known as Ursinha [13:51] rvba: Right. We need to think how pockets are going to be handled. [13:51] rvba: At the moment we handle !Release for source or destination. Hardcoding this is probably not the solution. [13:51] Er. [13:51] We *don't* handle !Release for source or destination. [13:55] StevenK: wgrant: I'm not RM; RM role is dissolved. [13:55] lifeless: But you are the effective RM, aren't you? [13:55] StevenK: wgrant: rc exceptions can be handed out by reviewers as long as it *is* rc (that is we can't deploy without it) [13:55] Oh. I didn't hear that. [13:56] wgrant: I don't have any permission bits that I'm aware of that haven't been delegated out to be team responsibilities [13:56] wgrant: what do you mean exactly "we don't *handle* release for source or destination" ...? [13:56] wgrant: I'll be closely involved, this is our single highest risk activity [13:56] rc also means less than it used to, since most things can go out shortly after. [13:57] wgrant: did you get spm cycles for qastaging ? [13:58] rvba: Different distros will want to do different things with pockets. At the moment we are hackishly hardcoded to ignore -backports and -proposed when generating DSDs. Now we're adding more hacks to map release into -updates when something has released. This is just going to become an ever-increasing stack of hacks. [13:58] ugh [13:58] lifeless: For which? The questions QA? [13:58]