[00:02] jml: at least it seems like it's on the decline some [00:03] jml: some really bored people that have probably never been laid and never will be.... [00:08] what's the test plan home page? [00:10] never mind [00:10] (I'd like to know what it is, but at least now I've done my QA item) [00:14] I'm really looking forward to landing my 'hot bugs' patch [00:15] jml: you should perhaps wear your hot pants when you do so? get in the groove of the moment. [00:16] spm, good call. [00:17] spm, I'll take photos for you. [00:17] jml: on this one, I'm prepared to believe *without* photos. a rare exception. [00:17] spm, heh. [00:24] I often want to file multiple bugs on a project at once [00:24] I wish Launchpad made that easier. [00:30] mm [00:37] I'm filing stacks of usability bugs on 'gtg' [00:37] which means filing a bug, hitting "End", looking for the "Report another bug" thing and then clicking it. [00:37] I filed a bug about this years ago, of course. [00:37] since my previous bug tracker was trac, which makes this quite easy. [00:38] * mwhudson noms [00:42] I hate trac... [00:42] jml: just do Back; change details if you know it wasn't reported... [00:45] MTecknology, I hate trac too, but it still does some stuff better than Launchpad. [00:47] jml: You ever try Mantis? [00:47] MTecknology, maybe once, but never for long and not enough to remember it. [01:08] thumper, hey, I thought code review comments would get permalinks for free [01:11] * jml lunching [01:33] jml: what's gtg? [01:34] btw unless it's just me (which it may be) https://bugs.edge.launchpad.net/malone/+bug/471328 seems pretty critical [01:34] Bug #471328: Also affects search error - too many matches [01:34] poolie, "Getting Things GNOME" [01:34] you can't choose bzr or bzr-svn from a project picker [01:34] poolie, it's a todo thing that has potential, a fairly active community and not the greatest code. [01:35] poolie, I'm living in constant tension between writing my own and helping them with theirs. [01:35] "price: Free" :-) [01:35] thanks ubuntu :) [01:37] poolie, I'm not sure how that could be a new bug [01:37] poolie, well, at least new to this cycle [01:38] * poolie plays on staging [01:38] well [01:38] i didn't see it before [01:38] i would have noticed because i do reassign from bzr to plugins pretty often [01:38] maybe it's transient xmlrpc failure? [01:39] poolie, it was filed in early november, so it'd have to affect production. [01:39] * poolie tests [01:39] poolie, anyway, it does seem pretty serious. Given that all of the bugs guys are afk atm, I'm not sure of the best way to escalate it. [01:40] let me try to reproduce it again [01:40] I mean, other than me actually fixing the bug. [01:40] :) [01:41] hm on https://bugs.staging.launchpad.net/bzr/+bug/385879 i don't get that problem but just a stuck spinner [01:41] Bug #385879: EOL filter only applied to files when first checked out [01:41] oh when i tried a second time i did get "undefined" [01:41] so it may be intermittent? [01:42] sounds plausible === adiroiba1 is now known as adiroiban [03:23] mwhudson, hi [03:23] jml: hi [03:24] mwhudson, I was wondering, now that we almost have a fully working bazaar.edge and we have a bzr buildbot... [03:24] mwhudson, how much of a pain would it be to automatically land working bzr updates? [03:24] mwhudson, and do you think that would be a good idea. [03:25] mwhudson, feel free to tell me to go away -- I'm only asking out of curiosity. [03:25] jml: i guess it wouldn't make sense to go on getting bzr from an egg in this case [03:25] mwhudson, but eggs are teh future. [03:26] jml: in fact it flies rather thoroughly in the face of the "use releases" direction buildout consciously pushes us in [03:26] jml: that doesn't mean it's a bad thing though [03:26] necessarily [03:27] jml: having an autoupdated thingy for bazaar.edge but keeping prod using releases would be nice [03:27] a bit harder to arrange though i guess [03:27] mwhudson, I was typing the same thing, more or less :) [03:28] i guess there might be issues with new formats [03:28] you could push a new format to baz.edge but the scanner wouldn't be able to read it [03:28] well, i guess tests fail when a new format is added [03:28] mwhudson, don't we already have that problem? [03:28] jml: yes [03:28] well, modulo what i just said [03:29] mwhudson, yeah, but even without edge, people can still push up formats that we don't support. [03:29] jml: true i guess [03:29] maybe only over sftp [03:29] right. [03:31] mwhudson, anyway, it's something to keep in the back of the mind. getting it set up might save us some boring maintenance work, and give bzr better a tighter feedback loop for new server-side behaviours. [03:32] (maybe we should get poolie to do it!) [03:32] saying the word 'egg' requires fighting robert [03:33] better than fighting a bear i guess [03:33] questionable [03:33] more seriously, we do have nightly debs [03:33] why not use them+ [03:33] s/+/? [03:33] LP has a vendetta against using packages or anything that isn't woefully insecure. [03:34] poolie, I think this is one of those cases where whoever wants to get this done will need to possess a will of iron, regardless of implementation details [03:34] poolie, and that given said iron will, they get to do it however they want. [03:34] the good thing is i do have a will of iron [03:35] the bad bit is it's iron filings [03:35] hahaha [03:35] os [03:35] so [03:36] hm [03:36] i may try to get it running against the system bzrlib [03:36] not your own copy [03:36] poolie, one of the things about using nightlies to consider is that the buildbot doesn't run against them. [03:37] poolie, and we only want to update if the tests pass -- it is still a live system running against production data. [03:41] * jml afk === Ursinha is now known as Ursinha-afk [04:30] ok, i didn't expect this: http://pastebin.ubuntu.com/342383/ [04:30] (i guess i need to run it with ./bin/py or python2.5) [04:33] spm: buildd-manager is broken. [04:34] mwhudson, nice [04:34] * jml does some illicit programming [04:34] don't tell anyone [04:35] https://edge.launchpad.net/~ubuntuone-hackers/+archive/ppa/+build/1129031 is assigned to a builder and utterly broken. I wonder if that is related. [04:35] wgrant: I assume that's not a statement of observation; rather a pls fix? :-) [04:35] spm: Yes please. [04:36] hrm. oddness. going for a bounce on it. [04:36] It might help, but there are some broken build records around... [04:37] i wonder if testcasewithfactory should flush the active stores in it's tearDown [04:37] ahh. k. will chase them too. ta. [04:37] someone just commented that, CACHE_DIRECTORY = os.path.expanduser('~/.launchpadlib/cache') might be a bad idea for windows [04:37] does anyone have any good suggestions for what should be used instead? [04:37] this is for a launchpadlib cache [04:38] wgrant: out of not-idle curisoity, have you considered becomming our nagios system? :-P [04:38] * mwhudson finds that all his tests so far violate database constraints, but none of them actually issued any queries [04:38] anyway, /me eods more or less [04:38] biab [04:38] spm: Ah, that build is fixed now. I wonder if the restart did that. [04:38] haven't been able to get the (*&^(*&^ing thing to stop yet. [04:39] I do so love issuing "stop" scripts and have them ignored [04:39] spm: SIGTERM kills it in not too long for me... [04:39] Hmm. [04:39] Interesting. [04:39] yeah - but there are "right" ways. I suspect the kill and kill with prejudice are in the immediate futuire here [04:40] I don't see why an attempt was made to build https://edge.launchpad.net/~ubuntuone-hackers/+archive/ppa/+build/1129031 at all -- it's five months old, and for a disabled archive. [04:41] 2009-12-16 04:37:04+0000 [-] Received SIGTERM, shutting down. <== LIES!!! [04:42] I' [04:42] I've not worked out why it doesn't just die quickly. [04:42] it's pretty effing annoying... [04:43] 2009-12-16 04:43:08+0000 [-] Server Shut Down. [04:43] awesome. nice and fast. [04:43] Is it back up yet? [04:43] Hum. [04:44] That build oopses sometimes, but not others. That is slightly concerning. [04:44] yup. all looks "normal". for values of... [04:45] Much better. Thanks. [04:45] Any idea what killed it? [04:49] the passage of time... [04:49] I tried 2 'stops', one straight up kill; it actually stopped on it's own a few secs before I was going to -9 it. [04:50] I mean, what broke it? [04:50] either way. wrong. bug 497282 [04:50] Bug #497282: buildd-manager is reluctant on shutting down in a timely manner [04:50] no idea. [04:50] it looked like it was havng issues with one of the hosts. but... [04:51] wgrant: what were the symptoms you were seeing that showed it as broken? [04:51] bohrium was hung as usual, but that happened 10 hours before the breakage. [04:51] spm: No build dispatched lately. [04:51] kk [04:51] https://edge.launchpad.net/~ubuntuone-hackers/+archive/ppa/+build/1129031 was also sitting on /builders, assigned to a builder but with status 'Needs build' [04:51] That doesn't make sense. [04:52] And now it OOPSes intermittently like OOPS-1446ED192 [04:52] weird [04:53] It seems like it was pretty effectively broken, since it also wasn't updating the status of builds that were in progress. [04:54] jml, i think the answer to your 'does lplib follow a fixed release schedule' is 'no' :) [04:54] poolie, it's _really_ frustrating. [04:55] poolie, trunk has some stuff that'll make bzrlib integration much easier. [04:55] well, a little easier. [04:59] That problematic build just got hit again. I wonder if things have jammed up already. [04:59] wgrant: aaaah. this may be a bug we've been tracking [04:59] Yeah, it is. [05:00] For now, kill that build and restart, I guess. [05:00] https://bugs.edge.launchpad.net/soyuz/+bug/496574 [05:00] Bug #496574: buildd-manager fails to deal with "Fault 8002" errors

[05:00] and there was another.... [05:00] Oh, could be that, I guess. [05:01] meh. can't find it. the gist is that something in the soyuz part is borked. this has been *flooding* us with OOPSes. that you saw that oops, (as did I on a 2nd attempt) suggests that could be it. [05:02] That build has hit a builder again, but other builds seem to still be working. [05:02] So maybe the two horrible brokennesses are unrelated. [05:04] mwhudson, where does your lazy-import-supporting pyflakes live? === danilo_ is now known as danilos [07:48] * jml gone [07:54] henninge: hi. do you have 5 minutes to answer a questions? :) [07:54] here is the code http://paste.ubuntu.com/342540/ [07:54] adiroiban: Hi! [07:54] * henninge looks [07:55] i would like to get rid of the permission checking from model [07:55] and only have them in security.py [07:55] also to get rid of permission_helper.py [07:55] as those checks are also implemented in security.py [07:57] adiroiban: I just created permission_helper.py last cycle [07:57] adiroiban: and I actually have a branch that extends it [07:58] adiroiban: I discussed this at length at the Dallas sprint [07:58] henninge: why not have that logic in security.py ? [07:58] adiroiban: because it is needed in the model, too. [07:58] adiroiban: the zope security system will only give you control over access to attributes. [07:58] but we can import the class from security.py [07:59] adiroiban: which class? [07:59] AdminTranslationImportQueueEntry [08:00] like the last example from the pastebin [08:02] but that one still uses is_admin_or_rosetta_expert. Where would that come from. [08:02] ? [08:03] adiroiban: The challange is to control which values the parameters of a method can take in relation to the user that is setting it. [08:03] henninge: true. it just replaces the self.isUbuntuAndIsUserTranslationGroupOwner(user) [08:04] and similar , we can use OnlyRosettaExpertsAndAdmins [08:04] adiroiban: To me, security.py is our interface to the zope policy. But we have more policy than that. [08:04] henninge: yes. but you can do the same in lib/canonical/launchpad/security.py [08:06] but with permission_helper.py, we will end up with code duplicated in security.py [08:06] as we already have a check similar to is_admin_or_rosetta_expert , in security.py [08:06] the idea was to move the code out of security.py to make it usable in model code [08:06] adiroiban: yes, and that check should use is_admin_or_rosetta_expert. [08:06] but then we still need that code for the view [08:07] yep [08:07] ok [08:07] then reference security_helper from security.py [08:07] that is also ok [08:07] sorry, which code do we need for the view [08:07] adiroiban: it does! [08:08] adiroiban: security has just not been converted to use these helpers [08:08] completely [08:08] security.py, I mean [08:08] henninge: yes. I see know. [08:09] adiroiban: may be permission_helpers is the wrong term [08:09] then I should implemenet all the translations security logic in security_helper.py [08:09] and then use that code from security.py [08:09] adiroiban: yes, and call it permission_checkers.py [08:10] adiroiban: I am moving the file out of translations into services anyway. [08:10] adiroiban: I will do the renaming. Can you wait until my branch is done? This should happen this morning. [08:10] henninge: ok. It make more sense now :) [08:11] yes. no hurry [08:11] Great ;) [08:11] adiroiban: actually, danilo wants me to de done with it ... ;) [08:12] henninge: but why is not a good idea to use security.py classes in the model ? [08:13] adiroiban: I have not thought much about that but my first impression is this: [08:13] security.py is our interface to the permissions mentioned in the zcml [08:14] I don't think it would such a bad idea, especially since they define a permission for a user to a model class [08:15] so, if there is a statement on a class that requires a certain permission, there will be a checker or it in security.py [08:15] but, as I mentioned before, that does not give us finer control over method parameters. [08:16] so we have to add checks in the model about parameter values. [08:16] @enabled_with_permission('launchpad.Admin') can only be used in the view? [08:16] henninge, right, but those checks can be gotten by using different privileges... ie. if has_perm(user, 'launchpad.Admin'): allow this [08:16] If we want to use the classes from security.py for this, we'd have to add classes there that have no relation to the zcml [08:17] adiroiban, you can define in ZML what attributed/methods are accessible for each permission, but there are cases where you want to allow only certain values to be passed in by certain users (i.e. henning has an example of setStatus on translation import queue, where only admins can set any status) [08:17] adiroiban: on, it is used on the interface, thus on the model. But it only applies to attributes. [08:17] you know better. I see security.py as the place to store all security logic [08:18] adiroiban: yes, I think it is a question of what we define security.py to be and what is supposed to be in there. [08:18] danilos: yep. I saw that code. and maybe we can have those values in a different attribute [08:18] :) [08:18] or just do the check in security.py [08:19] adiroiban: I thought about that [08:19] that would have meant to have seperate "setStatusApproved", "setStatusFailed" etc [08:19] or "setStatusAdmin". [08:19] I was advised against that at the sprint. [08:20] henninge: ok. np. [08:20] just saying [08:20] then I will add the logic in security_helper.py [08:20] and make use of it from security.py and from the model [08:21] adiroiban: that is how I meant it to be [08:21] adiroiban, danilos: but maybe we should take that discussion to the list as this is a general discussion about the use of security.py. [08:21] henninge, yep, sounds like a good idea [08:22] ok, I will do that [08:26] henninge: i think that is ok to move code away from security.py ... as it can become a big file, but I think we should still use the security.py interface from the model. Instead of security_helper.is_admin_or_rosetta_expert() use security.OnlyRosettaExpertsAndAdmins.checkAuthenticated [08:26] or create a new permission interface [08:27] but I will wait for your branch [08:27] and the discussion from the ml [08:28] adiroiban: yes, the discussion will be of most interest as I don't want some-one else to come along next cycle or so and change it all over again. We need some consensus here across the team. [08:30] henninge: ok. [08:53] adiroiban: I just realised you cannot just use " security.OnlyRosettaExpertsAndAdmins.checkAuthenticated" from model code as checkAuthenticated is not a class method. [08:53] adiroiban: you'd have to instantiate OnlyRosettaExpertsAndAdmins first. [08:54] yes. I'm doing something like this OnlyRosettaExpertsAndAdmins( self).checkAuthenticated(user) [08:55] I see [08:56] I'm running now all tests [08:56] and see if there are problems [09:18] morning [10:07] Hi, I have a trivial lp-dev-deps change, if someone has a moment: https://code.edge.launchpad.net/~maxb/meta-lp-deps/no-pullparser/+merge/16232 [10:35] hi mrevell! you have nag mail. :-) [10:36] jtv, thanks, I see it :) Will get to it today [10:36] * jtv cheers mrevell on [10:36] heh [10:52] henninge: A long, long time ago we specced out a system of crowds, where you would construct a crowd by combining teams and peoples. You could then query the crowd to see if a person was or was not a member of that crowd. Designed to be DB friendly, but also would make nicer things like is_admin_or_rosetta_export() [10:53] stub: sounds cool [10:55] al-maisan_: do you know if there are any current plans to move Build.sourcepackagerelease (and the stuff that relies on it) into BuildPackageJob? [10:55] jtv: on the phone, ttyl [10:55] pl [10:55] ok [10:56] henninge: The crowd was really just a list of Person (or even just their ids), and the 'is in' and 'isn't in' operations just single queries against the TeamParticipation table. [11:04] Morning, all. === matsubara-afk is now known as matsubara [11:13] jtv: AFAIU there are no plans to relocate Build properties into BuildPackageJob [11:13] jtv: why do you ask :) [11:13] al-maisan_: my question was misguided. I don't need a Build for our prospective BuildFarmJob implementation, do I? [11:13] hi deryck [11:15] jtv: you probably don't .. what data/properties do you envision being in your "TranslationFarmJob" ? [11:15] al-maisan_: a reference to the branch it should gets its data from. [11:16] jtv: in that case it would seem there's no need to reference a `Build` [11:48] bigjools, is this the right interpretation of you described on the phone? http://rookery.canonical.com/~jtv/farmjob.dia [11:49] can you turn that into a PNG? [11:49] bigjools: with the power of free software, yes I can! [11:50] bigjools: http://rookery.canonical.com/~jtv/farmjob.png [11:51] bigjools: and then I gather our BuildFarmJobBehavior would receive a BuildQueue entry to work on [11:57] jtv: it's a bit corrupted [11:57] is the << Iwhatever >> part supposed to be an "implements" ? [11:57] bigjools: yes [11:58] although maybe the base class does that... not too important I guess :) [11:58] jtv: ok it's wrong on your translations one then :) [11:58] bigjools: ahhh, they're both wrong actually: meant to say IBuildFarmJob on both [11:59] jtv: that's the basic gist of things, yes. Although you will need content classes and builder behaviours as well. [11:59] bigjools: I knew about the behaviors but left them out here because I'm focusing on db schema right now... what are the content classes for? [12:00] jtv: scratch that you already have one, sorry [12:00] ah, branch :) [12:01] bbiab [12:17] adiroiban, hi, have you done the QA on https://bugs.edge.launchpad.net/rosetta/+bug/406477 as well? [12:17] Bug #406477: Need more permissions on admin template pages for ubuntu-translations-coordinators === jelmer_ is now known as jelmer [12:18] * adiroiban looking [12:22] salgado: Hi. If you're around, I'm going to be doing a lp-deps upload today, if you wanted to put the ssh removal in too [12:29] hi maxb. I'll try to get that in before you upload, but if it's not there by the time you get to it, don't worry [12:29] unless you want to do it yourself and get the blame^Wcredit for it [12:36] adiroiban: need help with that Q/A? I have an unprivileged account for comparison [12:37] jtv: yes. please test this link https://translations.launchpad.net/ubuntu/karmic/+source/kdepim/+pots/kxforms/+admin [12:38] do I need to create a test account on edge or staging is also ok? [12:38] I'm new to LP qa process [12:38] adiroiban: staging should be fine, provided you're sure that your branch has been rolled out there [12:39] i can not access staging right now === Ursinha-afk is now known as Ursinha [12:41] Hi everyone [12:42] to wich user is the svn import branch review going? [12:42] no need to shout! :) [12:42] i cant see ist from the default users site [12:42] ?? [12:42] where did i shout? [12:42] gottmodus: your nick [12:42] ah ok [12:42] adiroiban: same here... and getting a "not found" for that page on edge. We're expecting an "access denied," no? [12:43] jtv: no. I should be ok [12:43] gottmodus: the joke does work better with all caps, but it's a bit noisy nonetheless :-) [12:43] the Navigation is breacking that chain [12:45] jtv: I have tested as an UTC and everything is ok [12:49] adiroiban: it's a bit weird though that it says "there's no page with this address in Launchpad" when it should say that I'm not allowed to access the page [12:50] jtv: that was the same case in the previous implementation [12:50] adiroiban: in that case, no worry here. :) [12:51] jtv: you can try on lp.net [12:52] jtv: i disabled my redirection for edge, and tested on stable and I also got a page not found error [12:52] jtv: there should not be a problem, since the user does not see the admin link [12:53] jtv: actually, neither do UTCs ... as the +templates pages times out [12:53] :( [12:54] bigjools: Do you know if dpkg + buildds are upgraded for tomorrow's rollout? [12:56] adiroiban: from an unprivileged account I see neither the link to that page, nor the same for another (non-deactivated) template. So looks good to me. [12:57] wgrant: in progress, but not yet AFAIK [13:01] danilos: I don't know what to say. I have tested on my computer, and tested on edge as UTC. Everything was ok. Also the test was written to cover an anonymous, a logged in user and an Admin [13:02] adiroiban, right, it was mostly to do some QA, since you haven't marked it as qa-ok in the bug... also, QA is usually done on edge and/or staging, you might be surprised, but many things break only when we hit production database :) [13:02] adiroiban: did D. find something wrong with the branch? [13:03] jtv: I don't know. He approved it [13:03] just confusion over the qa process then I guess? [13:04] jtv: yes. I did not tag it with qa-ok [13:04] but only qa-needtesting [13:04] We're weird that way. :) It's a pretty nice system actually. [13:04] I was thinking that I should mark qa-needtesting, whey I need someone else to also look into it [13:09] adiroiban: it should happen automatically (though we did have some breakage recently) [13:10] jtv: well I don't know how things should be going... so as always... I am very confused [13:10] adiroiban: need a summary? [13:11] jtv: or a wikipage ? :) [13:11] adiroiban: don't think we have one for this, since it's still only on an experimental scale... [13:12] jtv: so when my branch is merged with a fixes=lp:222 the bug report is updated to fix-comminted and qa-needtesting tag is added? [13:12] and then I have to test the commit on devel/staging and set it to qa-ok ? [13:12] adiroiban: the bug status is not updated to "fix committed" iirc, but otherwise, yes [13:13] And you try to Q/A on staging or edge, not just devel because of the unexpected differences that real-life data can make (see what D. said above) [13:13] jtv: I have tested all my bugs targeted for this release and tagged them with fix-ok [13:14] adiroiban: we usually just verb "QA" to avoid confusion with the test suite. [13:15] jtv: ok. thanks for the jargon update :) [13:15] Glad you don't mind; I know I'm pedantic but who knows what confusion it may save later :) === jtv is now known as jtv-eat [13:16] well I still have problems with page tests / stories / functional tests ... etc [13:26] BjornT, bug 497405 [13:26] \o/ [13:26] Bug #497405: Run Windmill tests by default [13:29] beuno: i'm really looking forward not having to monitor the jscheck builder constantly, and chase people to fix things. better to have buildbot take care of that :) [13:30] BjornT, yeah, and I'm happy to never have to stare at flacoste when we mention tests slowing people down to develop awesomeness [13:36] BjornT, can we use utilities/ec2 to run the windmill tests? [13:37] salgado: yes. if you grab my branch, you can. [13:44] BjornT, wow, I was expecting a "not yet" answer. this is great news [13:44] I might even review your branch [13:57] joey, have you changed your default font settings? [13:58] that was fast :-) [13:58] beuno: no [13:58] should I? [13:58] joey, that that I don't trust you [13:58] but [13:58] what's the values? :) [13:59] what are [13:59] serif, 16, monospace, 12? [14:00] joey, Preferences > Content > Fonts / Advanced [14:00] Loading FF. Was in chrome [14:01] can't find the option in chrome [14:01] 16 and 12 in FF [14:01] hm [14:02] and you see all bug pages with the side portlets at the bottom? [14:02] yes [14:02] on both FF and chromium [14:03] started about a week ago [14:03] or more [14:03] I was out for a few weeks [14:03] it was like that when I came back to work [14:03] joey, and your fonts [14:03]