[01:17] wgrant, i kinda hoped for a +1 on my process-one-mail announcement [01:17] not all that externally exciting but i found it much more comfortable [01:17] maybe someone will use it in future [01:17] and, traditional email is kind of lacking a cheap non-noisy +1 [01:17] poolie: Sorry, meant to +1 that last night, but was very busy. [01:17] np [01:18] And then had far too much mail this morning. [01:18] not really fishing :) [01:18] It is a really handy helper. [01:18] Often I avoid testing mail stuff because it's just so awkward :/ [01:18] But not any more! [01:18] Thanks for JFDI. [01:18] yeah, for me it was a minor mental breakthrough [01:18] "this is so hard to test.... oh, but it doesn't actually inherently have to be" [01:47] hello losa? i'd like some help testing email on staging [01:48] poolie: What are you wanting to test? staging processes mail automatically. [01:48] oh does it now? [01:49] and qas does too? [01:49] for some reason i thought it needed to be manually prodded [01:49] Ah, qas, true. It used to do it less frequently than staging, and may no longer do it at all.. [01:49] That's a good point. [01:50] hloeung, spm? [01:50] poolie: hi, how can I help you? [01:50] also i'm a bit curious what software ends up delivering the mail to lp's mailbox [01:50] hloeung, does qas automatically receive and process mail or do i need to get you to thump it [01:51] wgrant, re bug 813349, are you reopening it because it's flaky? [01:51] <_mup_> Bug #813349: hard to get from mp to per-revision diffs

< https://launchpad.net/bugs/813349 > [01:51] i see the point about it not being done done [01:51] poolie: I'm not too sure [01:51] poolie: It's not done-done, needs more work, and was one of two bugs that were cluttering up my queue. [01:51] poolie: I decided to finally get rid of it. [01:51] it would be easier if i just try it [01:51] Now that the queue can be otherwise clear. [01:52] lifeless: https://lp-oops.canonical.com/oops.py/?oopsid=OOPS-2084H53 [01:52] lifeless: timeouts now break lazr.restful :/ [01:53] I guess we want to capture sys.exc_info and use that. [01:53] Rather than assuming __traceback__ [01:53] Isn't that new in 2.7/3.0? [01:53] wallyworld_: http://paste.ubuntu.com/690450/ [01:53] wallyworld_: is what I recommend [01:53] did ec2 land change so that it no longer detaches from the vm? [01:53] poolie: No. [01:54] *intentionally? [01:54] because it does seem to be not detaching now [01:54] What's it said? [01:54] it's running 'python .... ec2test-remote.py', just printed some bzr hpss counts and now it's just sitting there connected [01:54] wallyworld_: that query will do unstacked branches first [01:54] i can see the tests are running [01:54] lifeless: so the change is you added "ORDER BY stacked_on ASC NULLS FIRST" ? [01:54] when this happened yesterday, it did eventually complete [01:55] poolie: Ah. It should detach right after ec2test-remote.py [01:55] poolie: I haven't seen that happen in months. [01:55] But when it does you can normally just ctrl+c [01:55] that's two out of two so far [01:55] After it starts ec2test-remote.py [01:55] wallyworld_: and then in increasing branch id, which will (except where folk have restacked) keep the depth of recursion low [01:56] wallyworld_: I think thats all I did yes [01:56] lifeless: that change somehow helps the optimiser ? [01:57] sinzui: There is something you can help me with, in fact. [01:57] lifeless: how did you know that's what you needed to do? [01:57] wallyworld_: it doesn't help the optimiser [01:57] StevenK, yes? [01:57] wallyworld_: but it will stop us looping 6 times through [01:57] ah ok. i'll look at the explains in more detail [01:58] they're still a bit hard for me to read [01:58] so we have a branch A->B->C->D->E->F [01:58] sinzui: I'd like a header and prose for the pop-up when the user wants to unsubscribe themselves. [01:58] and grok [01:58] until we update F, anything stacked on any of B,C,D,E,F, will recurse all the way to F [01:59] this tweak makes it be more likely to calculate F before A, as F will be unstacked, E will have a lower branch id than D, in the usual course of htings. [01:59] StevenK, I writing job? I am on it. Header: Hey, I see you are shooting yourself in the foot. I can help [01:59] there is one more step we can take [01:59] Hahaha [01:59] which is to use transitively_private and short circuit when it is set. [01:59] lifeless: but we don't limit the recursive query to take that into account [02:00] yes, so we add the short circuit [02:00] that's a nice optimisation [02:00] stevek prose: By unsubscribing from this bug you are trading lots of exciting spam for 404's Click Oops to stop this. [02:01] hloeung, wgrant, it's not obviously responding to my mail, so i think i do need process-mail run by hand [02:01] i'm pretty sure this was needed last time too [02:01] robert said something abotu it creating too much load to run it automatically [02:01] poolie: We should set up a regular cron job, and bump staging's frequency. [02:01] poolie: Since as of like 12 hours ago we have a new staging scripts server. [02:02] wgrant: file an RT to have that done. In the meantime, I'll run process-mail by hand [02:02] so should you file an rt for that or something? [02:02] thanks [02:03] * wgrant RTs [02:04] #47933 [02:04] <_mup_> Bug #47933: [Feisty] system requires acpi=off noapic nolapic

< https://launchpad.net/bugs/47933 > [02:05] poolie: ok, done [02:05] wgrant: ta [02:05] woo hoo [02:06] that's so awesome [02:06] thanks [02:06] poolie: Oh, it worked? :/ [02:06] only a wry face? [02:06] poolie: gandwana doesn't seem to be able to talk to the mailserver. [02:06] File "/srv/staging.launchpad.net/staging/launchpad/lib/lp/services/mail/mailbox.py", line 119, in open [02:06] raise MailBoxError(str(e)) [02:06] MailBoxError: [Errno 111] Connection refused [02:06] it did work [02:06] https://bugs.qastaging.launchpad.net/launchpad/+bug/800275 [02:06] <_mup_> Bug #800275: boot splash not branded < https://launchpad.net/bugs/800275 > [02:06] hloeung: Did you run it on asuka? [02:07] hloeung, can it ever send outgoing mail? [02:07] wgrant: yeah, I ran it on asuka [02:07] also, can you tell me what software is involved in delivering the message to lp? [02:08] poolie: We can run send-bug-notifications.py (on asuka, not gandwana yet, or we will be in big trouble) and it will end up in the staging mailbox. [02:08] lifeless: btw, thanks for all the xecellent input on getting that query optimised. i'm really pleased with the result [02:08] does it really (as that traceback suggests) read it over pop/imap? [02:08] Yup. POP3. [02:08] Really we should have a handler in the MTA which sticks it in rabbitmq or something. [02:09] Where it's picked up by a daemon. [02:09] Rather than being a */3 cronjob. [02:09] wallyworld_: http://paste.ubuntu.com/690457/ [02:09] wallyworld_: this is more of a rewrite [02:09] wallyworld_: it implements the recursion short circuit when you encounter a calculated transitively private [02:09] wallyworld_: it can do 800 a loop [02:09] in < 2 seconds [02:10] lifeless: i was in the middle of writing something similar myself :-) but you beat me to it [02:10] can't predict behaviour once you're 30/40K in trivially. [02:10] wallyworld_: happy to help [02:11] wallyworld_: I'm satisfied this will do reasonably well; I'm going to leave it here until/unless you ping me again [02:12] lifeless: np. it's all good. i'll check the unit test passes and then do a db-devel branch with the index [02:12] hloeung, process-mail again please? [02:12] poolie: sure [02:15] and once more? [02:16] poolie: done [02:16] super [02:16] looks like it all works [02:17] thanks very much [02:17] np [02:20] StevenK, http://pastebin.ubuntu.com/690460/ [02:21] hloeung: at some stage, no hurry, there's a new feature flag to go on production documented on LaunchpadProductionStatus [02:21] sinzui: Thank you, that is excellent. [02:23] wallyworld_: added [02:23] ok now some more real work [02:24] hloeung: legend! thanks! [02:54] what does 'orphaned' mean in the deployment report? that there's another branch for that bug? [03:10] poolie: There is no open bug related to the landing. [03:11] poolie: This could be because it was landed as testfix or a rollback, so has no bug, or the associated bug(s) are closed. [03:18] mwhudson: branch tokens - I think we need to talk about them :) [03:19] mwhudson: I have some fears [03:19] mwhudson: if you can make some time next week, that would be great [03:19] lifeless: yeah sure [03:19] lifeless: can you articulate your fears at all? [03:19] i.e. conceptual problems, security problems, performance...? [03:19] performance impact, user model impact, bzr ui impact [03:20] auditability too [03:20] and long term interaction with oauth/openid for bzr [03:20] (I want to ditch ssh keys altogether in some ways) [03:21] really? [03:21] tell me more? [03:21] poolie: we've spoken about this before [03:21] i have an awful memory [03:21] can you prompt it [03:21] poolie: I haven't advanced any plans, but ssh key management is a real headache for new users [03:21] doing an API style login once on the client would be much nicer. [03:22] i would certainly agree with making them a lot less manually maintained [03:22] right [03:22] lifeless: would you be less concerned if i just targeted removing the puller? [03:22] mwhudson: I think jelmer will have that done soon anyway :) [03:22] mwhudson: won't he ? [03:22] lifeless: no [03:22] Still needed for imports. [03:22] the import puller will still be required [03:22] mwhudson: so, we should talk :) [03:22] ok [03:23] lifeless: i suspect your diary is massively more cluttered than mine, feel free to do something involving calendaring applications and michael.hudson@linaro.org [03:25] mwhudson: i'll ping you monday :) [03:25] lifeless: ok [03:25] lowtek ftw [03:28] relying on memory << * in my case though :) [03:28] kk lunch biab [03:37] is there a private xml-rpc server method to check for a feature flag? [03:37] No. [03:38] It's not quite trivial, since you'd need some way to work out scopes. [03:40] what are scopes again? things like projects and teams? [03:40] https://launchpad.net/+feature-info [03:40] " [03:40] Scopes" section [03:40] ah nice [03:42] well, as nice as it would be to restrict anonymous ssh logins to members of a team ... :) [04:33] wallyworld_: \o/ landed that branch last evening :-) [04:37] wgrant: mwhudson: pass in a dict of literal scopes you're interested in [04:37] as a first-pass approximation [04:38] mwhudson, most obviously you could limit by ip [04:38] maybe also ssh client signature [04:38] that will let some obvious ones like page-id be passed in [04:38] lifeless: I considered that. [04:38] Maybe. [04:39] for a v1 lets get unblocked thing [04:39] its fine. [04:39] internal-only. [04:47] o/ huwshimi [04:47] * mwhudson contemplates the fun that is adding xmlrpc methods to lp [04:47] poolie: Hey! [04:48] mwhudson: xmlrpc it self is "fun" :-) [04:48] also true === almaisan-away is now known as al-maisan === al-maisan is now known as almaisan-away [05:10] wgrant: what are the '53' [05:11] lifeless, sorry to re-ask something i asked a month or two ago, but [05:11] 42 [05:11] once revisions are qa'd on stable, they'll deploy within a few days? [05:11] ideally ;) [05:11] No. [05:11] Ideally within 24 hours. [05:12] unless something else fails on production, or fails to qa? [05:12] it all depends on someone putting up a deployment request, other revisions not being in the way etc. [05:12] Unless stuff goes wrong, they're generally out in 24 hours. [05:12] the goal time is < 1 hour from qa [05:12] but we are far from that yet [05:12] lifeless: Ah, that email is mostly directed at Julian. https://devpad.canonical.com/~wgrant/2011/soyuz-expiration-blargh.html is not light reading. [05:13] lifeless: But, in summary, there are 53 BPRs published since hardy that are illegaly expired. [05:13] Because they're still published in the primary archive. [05:13] righto [05:13] Ugh, Tal. [05:13] We need to recover them from the archive. [05:14] I've read it now [05:14] thanks [05:15] jamesh: hi [05:15] jamesh: got 5 ? [05:15] sure. [05:15] Does anyone here use Identi.ca? I have an issue identical to this post but I don't understand what they're talking about: http://forum.status.net/discussion/873/badge-identica-badge.js-not-working-with-my-statusnet-installation/p1 (I'm asking here because I'm trying to set up something for Launchpad). [05:16] i use it but not the badge feature [05:16] idnetical how? [05:17] jamesh: so, I think: timeline wsgi injector should be a separate timeline-wsgi project; storm can grow a timeline_factory that timeline-wsgi talks to. python-oops-timeline should exist and have the context -> report code in it. [05:17] jamesh: so 2 new trivial projects, a patch to storm, and we're done [05:18] poolie: I have the same issue, as in, no posts show up on the badge. That post seems to say something about only friends posts showing up, but I don't understand the identi.ca language enough to know what they're saying [05:19] oh you want launchpadstatus's badge to be somewhere? [05:19] i thought you were trying to implement lp badges or something [05:19] lifeless: so what would python-oops-timeline hold exactly? Just the code to add the timeline to environ['oops.context'] and some on_create handlers? [05:19] or something more? [05:19] o/ jamesh [05:19] poolie: yeah I'm trying to show it on a status page [05:20] huwshimi, i suspect 'friends' is a hippy equivalent of twitter 'following' [05:20] jamesh: oops-timeline would be on_create handlers [05:20] statik: LOVE the reply you got from DEVOPS_BORAT :-) [05:20] and launchpadstatus has no friends :) [05:20] let's see if that's true [05:20] or I suppose it could do without copying to oops.context, since you could go context['wsgi_environ']['timeline.whatever'] [05:20] jamesh: thats timeline-esgi :) [05:20] *timeline-wsgi* [05:21] it's true: no friends [05:21] jamesh: I'm trying to avoid having specific projects have code for sibling projects in them [05:21] lifeless: okay. So the timeline-wsgi module knows about oops-wsgi. That sounds fine. [05:21] i wonder who DEVOPS_BORAT is [05:21] or alternatively the other way around [05:21] perhaps call it timeline-oops-wsgi; then it clearly can know about oops variables [05:21] sometimes it feels like it must be someone i know [05:22] mwhudson: Heh, yes. [05:22] huwshimi, i wonder what happens if lpstatus follows itself [05:22] jamesh: or as we touched on the other day, environ['timeline.timeline'] could be the dedicated variable [05:22] poolie: Can we make that happen? [05:22] lifeless: if the only connection is that it does "if 'oops.context' in environ: environ['oops.context']['timeline'] = timeline", then I don't think an extra package is worth it [05:22] apparently no [05:23] jamesh: So I've found, when I dig into some java stacks, that having *tiny* packages is actually very useful [05:23] poolie: OK, I might give up and use the twitter feed instead. [05:23] jamesh: it keeps the dependency chain very lean, when you go to reuse. [05:23] poolie: Thanks for your help :) [05:23] i have an idea [05:23] jamesh: I know storm doesn't do this, but I wish more code did. [05:24] it would be -awesome- if the overhead of doing a single python module (not package) was lighter. [05:25] jamesh: it won't have a practical difference for you though, will it ? [05:25] huwshimi, what happens if you just make a copy of the js, and change 'statuses/friends' to 'statuses/user_timeline'? [05:25] i reckon there's a fair chance that will work [05:25] lifeless: I guess not. It is one more thing to package though [05:26] poolie: I can do that (and it will probably work), but I'm trying to limit the number of working parts as this will appear when LP is down. [05:26] [we haven't drunk the buildout kool aid] [05:32] jamesh: time to give pkgme a test drive then :) [05:33] jamesh: these should be ideal cases for fully automated packaging [05:40] wgrant: welcome to the dark side [05:40] stub: yo [05:41] yo [05:41] lifeless: Heh. [05:41] stub: time for a catchup ? [05:41] I indeed finally gave in :( [05:41] lifeless: sure. Gimme a sec to boot the lappy [05:41] and steal my mike back [05:41] you want to be like mike/ [05:41] ? [05:48] When the database goes down at the moment do we currently display anything? What's involved to put a [new] page there? [05:48] stub: Matthew said you might know something about this...^ [05:50] huwshimi: Bug #840098 [05:50] <_mup_> Bug #840098: Librarian should return 503 pages during database outages < https://launchpad.net/bugs/840098 > [05:50] huwshimi: We just display an Oops page now. [05:50] huwshimi: We can put whatever you want there, basically... as long as it doesn't involve DB access, obviously. [05:52] wgrant: OK, I have (most of) a static HTML page (a little js and some images too) that will go there. Do you know how I can make this work? [05:53] wgrant: dark side? [05:53] nigelb: I created a Twitter account. [05:53] OHMYGOD [05:53] @wgrant? [05:53] wgrant: We can no longer be friends. [05:54] nigelb: I need to try to claim that using the 9 month rule. @wgrant_ for now. [05:54] Nasty common names :( [05:54] StevenK: No, but you can be followers [05:54] huwshimi: Boo, hiss. [05:54] wgrant: _ is the new style on twitter [05:54] huwshimi++ [05:54] huwshimi: It should ideally use normal LP JS/image infrastructure. [05:54] huwshimi: As it's rendered by LP, like any other view, except without the DB. [05:55] huwshimi: Just like the current OOPS page. [05:56] poolie: could you look at https://code.launchpad.net/~mwhudson/launchpad/feature-flag-xmlrpc/+merge/75673 briefly? [05:57] +"""XXX: Module docstring goes here.""" [05:57] or not :) [05:57] well yes, not that stuff: [05:57] heh [05:57] """Not ready yet (docstrings need writing etc) but would like feature flag manipulation looked at please :)""" [05:57] wgrant: Can we have a tumblbeast? :) [05:57] i know i'm just being snarky [05:58] yes, please [05:58] or, that's more of huwshimi's department, beasts [05:58] Heh [05:58] Was that the one TheOatmeal created? [05:58] yeah [05:58] its free to use now [05:59] * mwhudson goes to the pub [05:59] oops [05:59] mwhudson, it's not obviously wrong but i don't really understand it [05:59] poolie: ok, let's talk next week then :) [05:59] ok [06:00] I'm having brunch and mwhudson is headed to the pub. [06:00] the intention is good [06:00] timezones.sigh. [06:00] cheerio mwh [06:00] nigelb: Then move to Australia? [06:00] haa. [06:00] nigelb: what bothers me is when i'm going to the pub, and my colleague in POLAND hasn't stopped working yet [06:01] heh [06:01] mwhudson: It probably bothers him more :P [06:01] him/her [06:01] what bothers me is when i come home from the pub and robert is still working [06:01] StevenK: Only if it were that easy. [06:01] poolie wins [06:02] poolie: It's only when I'm in another timezone when I realise how many people here work *very* late [06:03] Oh wgrant doesnt have a concept of time. [06:03] Neither does StevenK. [06:04] Last night was particularly bad. [06:04] Since a script I expected to take 3 hours actually took 13 hours. [06:05] Could someone help me with this? http://paste.ubuntu.com/690527/ [06:06] nigelb: You have a tag mismatch somewhere. [06:07] I've been trying to trace it without much luck :( [06:09] http://paste.ubuntu.com/690530/ <-- pt file [06:09] Which table are you trying to close on line 68? [06:09] wgrant: StevenK: do you see an issue with lowering the minimum picker search term length from 3 chars to 2? perhaps if term is < 3 chars do not do fti search, just exact match to lp name [06:09] That table seems to be opened in the previous metal:widgets. [06:10] for bug 768148 [06:10] <_mup_> Bug #768148: person picker does not handle exact matches well - including not permitting exact matches for usernames < 3 characters long. < https://launchpad.net/bugs/768148 > [06:10] the one in line 25 [06:10] nigelb: I think 68 is trying to close the table on 25, but they are in separate macros. [06:10] Oh. I can't do that? [06:11] This is approximately XML. [06:11] It's not text-based like Django markup. [06:11] :D [06:11] You guessed where I'm coming from. [06:12] Argh. Need to ask sinzui for help. [06:12] I'm not sure how to get this workign without it being ugly. [06:14] nigelb: cani help at all? [06:15] wallyworld_: ohai, I'm trying to fix bug 80660 without killing kittens and without it being ugly. [06:15] <_mup_> Bug #80660: GAIM2.0b5 crashes on quit < https://launchpad.net/bugs/80660 > [06:15] err [06:15] 806660 [06:15] bug 806660 [06:15] <_mup_> Bug #806660: "Add a new address" in e-mail settings does the wrong thing when pressing Enter < https://launchpad.net/bugs/806660 > [06:15] but i hate cats. they all must die [06:15] * wallyworld_ looks at bug [06:16] wallyworld_: heh, on that you should look at Amber's (Graner) facebook photos (if you are her friend on fb) [06:16] wallyworld_: http://cheezburger.com/kenneystudios/lolz/View/1802695936 [06:16] nigelb: i hate fb more than i hate cats. and i hate twitter too. total waste of time [06:16] heh [06:18] StevenK: don't know where you find this stuff [06:18] StevenK probably has subscription to icanhazcheesburger [06:19] I do as well. BUt to one of their smaller sites. [06:24] nigelb: so did you try splitting the form into two like suggested in the bug report? [06:25] I did. [06:25] It was ugly [06:25] no luck? [06:25] What I did was split the table [06:25] Because semantically, I cannot have one form in another form. [06:25] And a form must either include the entire table or be only in one td. [06:26] yes [06:27] and that didn't work? having two tables? [06:27] It worked! [06:27] BUt it was ugly with lots of padding [06:27] Can we fix that? [06:27] s/padding/margin/ [06:27] can i see what you did? [06:27] sure [06:28] sec, let me restore yesterday's changes [06:28] Oh gah. [06:28] Give me 5 minutes I'll have to do it from scratch. Shoulda commited at safe point. [06:29] ok, just ping me [06:29] Shelve is handy. [06:29] Except when I end up with 60 shelves in devel. [06:29]