[00:01] sinzui: stupid mic [00:02] wallyworld: I had to toggle mute on Sound Preferences > Input. Even though it said it was on, the live check clear said it was not [00:03] * wallyworld tries that [00:05] sinzui: here but mic still broken [00:08] * sinzui is now deafened [00:18] we should do a ndt today [00:22] lifeless: I plan to, once I attack staging. [00:22] importds need QA. [00:22] oh cool [00:22] Then I need to remerge db-stable. [00:26] WebCat sounds like a fun project [00:32] utilities/shhh.py bin/sprite-util create-css [00:32] Traceback (most recent call last): [00:32] ... [00:32] KeyError: u'../images/mute.png' [00:32] * StevenK pouts. [00:35] wallyworld: so, what's the bit that needs dealing with picker wise? [00:35] jcsackett: sent you an email :-) [00:35] ah, fantastic. [00:35] just move a coupl eof line sof js [00:36] * jcsackett wades through codeimport emails to find the relevant one. [00:36] jcsackett: i have code impor temails go to /dev/nul :-) [00:38] StevenK: rm ./lib/canonical/launchpad/icing/icon-sprites.positioning [00:39] wallyworld: that definitely fits within the scope of what i'm working on now. i'll incorporate it into my personpicker-fixes branch. [00:39] jcsackett: thanks :-) [00:39] wallyworld: you're welcome. [00:40] wgrant: Thanks [01:02] And another 113 code import mails :-( === Ursinha is now known as Ursula [01:20] StevenK: That's what mail filters are for [01:21] Yes, I just fixed sieve [01:21] I think it's working, too. [01:40] jelmer: Still around? [01:40] wgrant: barely [01:40] jelmer: http://staging.launchpadlibrarian.net/72500040/vcs-imports-virt-manager-trunk.log [01:41] Something to worry about? [01:41] wgrant: somewhat - I tagged one of the bugs associated with the newer-bzr branch qa-bad for this reason [01:42] jelmer: qa-bad meaning 'must rollback' [01:42] At this stage it is more "must panic" [01:43] yup [01:43] so, rollback that rev. [01:43] jelmer: How critical is this? [01:43] Can we just roll back bzr-hg, or do we have to do the whole thing? [01:43] I guess the whole thing. [01:43] rollback the lot, its simplest. [01:43] known good. [01:43] wgrant: the whole thing, the older bzr-hg won't work (at least isn't tested) with the older bzr [01:44] OK, rolling back. [01:44] Thanks. [01:44] buildbot is even waiting for us. [01:44] \o/ [01:44] Hah [01:44] Although we'll just miss the following db-devel run. [01:44] So QA will be this evening. [01:45] What else can go wrong... [01:45] Build breakage, qastaging breakage, staging breakage, three RC revs, two bad revs, major QA backlog [01:45] 8000000 estimated plans make me cry [01:46] + deploying a big new feature at the same time [01:46] well [01:46] shortly after [01:46] I agree, it would be nice not to have it bundled in [01:47] wgrant: you redid the tag portlet queries recently right ? [01:47] lifeless: Yes. [01:47] They still suck. [01:47] https://bugs.launchpad.net/launchpad/+bug/793809 [01:47] <_mup_> Bug #793809: Distribution:+bugtarget-portlet-tags-content timeouts < https://launchpad.net/bugs/793809 > [01:47] Will need bugsummary. [01:47] (which is in!) [01:47] yes [01:47] I'm hacking on that now [01:47] Fantastic. [01:47] just gthering data on th ecurrent behaviour [01:47] jelmer: Do you have a bzr-hg bug for that? [01:49] (actual time=0.164..0.164 rows=1 loops=323890) [01:49] what is 0.16ms * 300K :P [01:50] wgrant: bug 793812 [01:50] <_mup_> Bug #793812: doesn't lock target repository when reading tags during fetch < https://launchpad.net/bugs/793812 > [01:51] jelmer: Hm, weren't tags not supported before? [01:51] So this will only affect branches that were already broken? [01:51] wgrant: tags were ignored before, and *most* of these branches were not working earlier [01:52] lifeless: 48 seconds? [01:52] Hmmmm. [01:52] lifeless: Perhaps we should not roll back. [01:52] Given that we can roll out to the importds later fairly easily. [01:52] wgrant: (hence my answer "somewhat" to your question earlier; the risk is that there's a couple of branches out there that were previously working that have tags) [01:53] StevenK: yes, and thats what bug privacy costs us [01:54] jelmer: wgrant: I'm fine with saying that this only affects previously broken branches [01:54] but in that case its not qa-bad [01:54] its qa-ok [01:54] lifeless: It looks like it will break some existing branches. [01:54] fsvo some [01:54] But bzr-hg is still beta. [01:54] Well. [01:54] It looks like there are 15 succeeding bzr-hg branches. [01:54] In all of LP. [01:54] So it can't break more than 15 branches. [01:54] some of them may have tags [01:55] lifeless: "Orsum" [01:55] jelmer: when do you think this will be fixed ? [01:55] lifeless: tomorrow morning [01:55] StevenK: see the plan in the bug I linked if you are interested [01:55] I say we deploy with this, and rerelease to the importds on Friday or Monday. [01:55] jelmer: wgrant: FWIW I wouldn't rollback given this analysis [01:55] Right. [01:55] Yeah, I agree. [01:55] I was trying to convince you of that, but it seems it wasn't necessary. [01:55] Great. [01:56] * mwhudson too, if anyone cares [01:56] jelmer: You've tested the others? [01:56] Sorry, I probably should've communicated that a bit better than just saying qa-bad and "somewhat"... [01:56] bzr-git seems happy, but I've not tried bzr-svn. [01:56] Heh. [01:56] It's late :) [01:56] I've tried bzr-svn and bzr-git [01:57] jelmer: qa-bad today means zomgpanic, since we have downtime in <48 hours. [01:57] wgrant: I realize that, but I also didn't want to say qa-ok while it could break existing branches (I hadn't bothered looking at the graph yet) [01:57] Yep. [01:57] Heh. [01:58] jelmer: Should we qa-ok all the bugs? [02:00] wgrant: yep [02:01] Hopefuly 13162 is good. [02:04] wgrant: btw, a window function eliminates the union [02:04] lifeless: Oh? [02:04] How? [02:05] tag in (official) or rank() < 10, or something like that [02:05] I won't need to do it [02:05] Hmm, true. [02:05] just noting another technique we could use to tune further [02:08] (49 rows) [02:08] Time: 595.781 ms [02:08] Nice. [02:09] thats for distro=1 [02:10] https://bugs.launchpad.net/launchpad/+bug/793809/comments/3 [02:10] <_mup_> Bug #793809: Distribution:+bugtarget-portlet-tags-content timeouts < https://launchpad.net/bugs/793809 > [02:27] wgrant: thanks for looking at my qa-needstesting's [02:28] cjohnston: Hah. wgrant looks at *everyone's* qa-needstesting's revisions :-) [02:28] lol [02:29] in that case, there's still one pending.. lol [02:29] cjohnston: He couldn't possibly steal all of your fun. [02:39] cjohnston: I'm waiting on a script to run on qastaging. [02:39] :-) [02:39] hloeung: How's send-bug-notifications.py going? [02:39] wgrant: Still going... and going... and going... [02:40] hloeung: Is branchmail eating the CPU or something? [02:40] I had no idea send-bug-notifications.py was the energizer bunny [02:40] I see nothing new in the inbox. [02:41] wgrant: err, I'm seeing it print out lots of "Notifying about bug XXXXXX" being displayed on my screen [02:42] Ah, yes, Thunderbird was just being slow. [02:44] It's not just the branchmail, there's quite a few jobs running right now [02:44] so yes, the load is pretty high on that server [02:44] :( [02:48] hloeung: Define pretty high? 10? 50? [02:49] StevenK: anything over 10 is high for me, and it's at 12 [02:52] Project windmill-db-devel build #369: STILL FAILING in 1 hr 8 min: https://lpci.wedontsleep.org/job/windmill-db-devel/369/ [03:48] StevenK: Can I have mawson? [03:48] wgrant: Take it, I'm done with it until after lunch. [03:48] Thanks. [03:51] lifeless: We are green. [03:51] Shall we unfreeze? [03:57] wallyworld: The pickers are healthy when the flags are all disabled? [03:57] wgrant: afaik [03:58] wallyworld: Great. [03:58] i last tested that a day or two ago [03:58] but there have been code changes since [03:58] Yeah,. [03:58] I'll check once hloeung's back. [03:58] By disabling them on qas. [03:58] i think prod has an earlier version deployed that is fine [03:58] Yeah. [03:58] But that's about to change. [03:59] Well, the earlier version thing is about to change. The fineness hopefully won't. [03:59] i tested when i put in support for the result ordering to be behind a flag [03:59] But we've had lazr-js changes and such since. [04:00] yes, but the lazr-js changes don't use a ff [04:00] So they are more likely to break even when the FF is off. [04:00] lazr-js will use json attributes if they are there, which they won't be if the ff is off [04:01] Sure. But this is Launchpad. [04:01] I'm still going to test it :) [04:01] ok [04:01] wgrant: fully, really ? [04:01] lifeless: Hm? [04:02] Oh, QA? Yes. [04:03] Both are fully green. [04:04] \o/ [04:04] daaaaploy [04:04] Already requested. [04:04] love your work [04:04] Will do it and unfreeze and remerge once hloeung is delunched. [04:04] Heh. [04:06] garrrr [04:06] storm reference columns paaaain [04:07] Yes. [04:16] ~/launchpad/lp-branches/devel$ bzr diff -c13163 | wc -l [04:16] 20211 [04:16] D: [04:26] lib/lp/bugs/doc/bug-tags.txt [04:26] Ran 1 tests with 0 failures and 0 errors in 0.679 seconds. [04:26] \o/ [04:34] wgrant: (48 rows) [04:34] Time: 165.065 ms [04:34] lifeless: Where'd the 49th row go? [04:35] And the other 340ms. [04:35] I had forgotten we aggregate up by sourcepackagename [04:35] so it was counting a few hundred K rows too many [04:35] Ahh [04:38] the extra row was apport-collected [04:38] not sure why its gone, looking [04:41] is there with spn is null [04:41] my theory - it was 11th with the duplicates fixed. [04:41] because its /all/ source package targeted bugs it had 100% inflation [04:42] Ahh [04:42] That could do it, I guess. [04:43] 3626 [04:43] (null) [04:43] and [04:43] 3869 [04:43] (not null) [04:43] discrepancy handwaved away by multi-task bugs [04:43] that takes its count to 7K [04:44] and that was 10th place [04:44] hmm,t hats ever [04:44] * lifeless checks with open [04:45] 2100 [04:48] rank | tag | sum [04:48] ------+------------------------+------- [04:48] 1 | i386 | 34609 [04:48] 2 | apport-crash | 29323 [04:48] ... [04:48] 14 | apport-collected | 2003 [04:50] Hey, the 'Packaging/SourceBuildsdaily builds knowledge base' link at the bottom of 'https://help.launchpad.net/Packaging/SourceBuilds/BzrBuilder' is broken. I can't seem to log in to fix it... [04:55] can has review ? https://code.launchpad.net/~lifeless/launchpad/bug-793809/+merge/63635 [04:55] michaelh1: login attempts seem to hang for me :/ [04:55] sso is having trouble probably; we have a bug open on that === lifeless changed the topic of #launchpad-dev to: Performance Tuesday! | devel is release-critical until r13164 deployable | https://dev.launchpad.net/ | On call reviewer: - | Critical bugs:212 - 0:[######=_]:256 [04:56] Ah, finally worked. [04:58] michaelh1: Fixed, thanks. [04:59] Ta [05:09] wgrant: You called? [05:10] hloeung: Hi! Three things: [05:11] okay, hit me [05:11] 1) Could you please bring PQM out of RC mode (I think both devel and db-devel are RC at the moment)? [05:12] 2) Could you remove the disclosure.picker_enhancements.enabled feature rule from qastaging, so we can test how it will behave on production without it tomorrow? [05:12] 3) There's a nodowntime deployment request on LPS, to deploy all the stuff before the db-stable merge. [05:17] lifeless: Reviewed. [05:20] wgrant: dunno whats wrong with the indentation [05:20] wgrant: if I have a PackageUpload pu, and I do pu.addSource(…), should that be enough to make contains_source become True? [05:21] jtv: Yes. [05:21] jtv: Minus caches, but there might not be any. [05:21] lifeless: Multi-line function calls are meant to be indented like this: [05:21] something( [05:21] foo, bar) [05:21] Not [05:21] something(foo, [05:21] bar) [05:22] And multi-line function definitions are to be indented like: [05:22] def something(foo, [05:22] bar) [05:22] I forgot a space. [05:22] But you get the idea. [05:22] problem is both of those are fugly. [05:23] you'll find indenting like I did already present in the code base [05:23] -> vet, back soon [05:25] lifeless: Hm, I think the postgres planner needs a good killing. [05:25] Nice argument there: the same person who won't listen to his reviewer and created optional reviews says "I can do it the way I like because we already have places where it's done that way." :-) [05:26] Heh. [05:29] What's the problem with the postgres planner? [05:29] jtv: See the last comment in the MP. [05:29] https://code.launchpad.net/~lifeless/launchpad/bug-793809/+merge/63635 [05:31] The answer to that is not killing the planner. [05:31] It would help. [05:32] No it wouldn't — then we'd have to hand-tune _every_ query at the cost of legibility. [05:32] What Robert mostly uses WITH for AFAICS is as a new optimization barrier. [05:32] Yeah. [05:32] I was tempted to read the slides about the query planner from the last Postgres meetup, but I chickened out [05:34] It's just an ever-shifting battleground: we find we can get a query faster (at least for the cases we're looking at) by making it too hard for the optimizer to restructure. [05:34] Then the optimizer gets smarter, and most queries get faster — but it breaks that trick. [05:34] The fights over planner hints are endless. [05:36] Even though the decision is already made: we're not going to have them. [05:36] Heh. [05:36] Yay, I can self-review now. [05:36] But you have been for months with rs=wgrant [05:37] Or was I not supposed to point that out? [05:37] That was always for testfixes/merges. [05:37] No new code :) [05:37] But yes, shh. [05:39] * jtv worriedly pretends he had not heard that [05:41] Hmm. [05:41] Pickers with dozens of pages look a bit bad. [05:41] But it'll do. [05:41] eg. search for 'launchpad' in the product picker on https://bugs.qastaging.launchpad.net/launchpad/+bug/1234 [05:41] <_mup_> Bug #1234: Gina is an unmaintainable mess of command line options, environment variables and shell scripts < https://launchpad.net/bugs/1234 > [05:42] This is a bit of python syntax I'm still missing: the "if" clause from list comprehensions, but without the "for" [05:42] Names = [self.name if self.name is not None] [05:43] Yes :( [05:44] filter()'s deprecation upset me. [05:44] Same with map(). [05:44] Those have been deprecated!? [05:44] What jtv said?! [05:44] Surely that's only in MS Python.NET enterprise? [05:45] Or some other thing that we don't care about? [05:45] Hm, looks like they weren't actually killed from py3k. There were going to be at one point. [05:45] Phew. So their obsolescence was deprecated? [05:46] map/filter appear to be iterators in py3k, though. [05:49] That makes more sense. [05:50] * jtv dreams for a moment about the cool vectorizations we could have if this wonderful language were a compiled one [05:50] Then again, modern compilers can do that without it being explicit in the code can't they [05:51] Yes :( [05:54] -> vet, back soon]]]]]({{{{{kkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkK0;2~")Y9;2~ Also, calling that "store" is a bit too much of a lie for my tastes. [05:54] Did you forget to take the cat? [05:54] Hahaha [05:55] wgrant: Perhaps that was cat clawing despertly in a vain attempt to not be scooped up [05:55] We can only hope! [05:55] jtv: uhm, I do listen to reviewers - a lot. [05:55] Shhh you're spoiling it [05:56] Never let the facts get in the way of a good punch line [05:57] hloeung: PQM is indeed un-RC'd. Thanks. === wgrant changed the topic of #launchpad-dev to: Performance Tuesday! | https://dev.launchpad.net/ | On call reviewer: wgrant | Critical bugs:212 - 0:[######=_]:256 [05:58] wgrant: How did you check? Just so I know in the future how to check [05:58] hloeung: I landed something :/ [05:58] Oh hahah [05:59] so, if you look at the query plans *we get*, the planner often treats exists / subqueries as things to run multiple times [05:59] lifeless: Sure. But it should be smart enough to not do that. [05:59] jtv: if you can find a way to get it to not do that, without a with, that would be cool. [05:59] At least sometimes. [05:59] AFAICT using WITH as a way to declare what is essentially a variable is good style SQL wise. [05:59] lifeless: I don't know of a way — I think it's one of those cases where it just wasn't the extra optimizer time. [06:00] Yes, it's fine. [06:00] lifeless: Perhaps good style, but unconventional. [06:00] Well no, that's exactly what it's for. [06:00] Right. But it's not very much. [06:00] I've rarely seen it used. [06:00] And it's ugly in Storm. [06:00] It's relatively new. [06:00] Right. New => unconventional. [06:00] wgrant: so, help make it better in storm? [06:01] wgrant: unconventional in our code base perhaps. [06:01] I've read enough horror stories in that MP. [06:01] The main "good" reason to use it is to remove such duplication. The main practical reason is as an optimization barrier. [06:01] anyhow, when storm is three times the size of direct queries, I really start to wonder why we use it. [06:01] How does it work in storm? [06:01] jtv: see the use of with_ in https://code.launchpad.net/~lifeless/launchpad/bug-793809/+merge/63635 [06:02] Ah, great to have that as a keyword isn't it? :) [06:03] yes (no) [06:03] wgrant: so, I [06:03] 've changed the browser/ indenting as its much of a muchness [06:03] (though its still fugly) [06:03] Not a muchness! [06:03] Heh. [06:03] the bug.py indent is really harder to read [06:04] Oh? [06:04] def foo............ [06:04] fwejijw [06:04] """..... [06:04] dunno about you but that plays havoc with my eyes [06:05] It's not ideal. [06:05] But not much is. [06:06] I can raise a discussion on list about this if you like, but this is the sort of thing I really hate about coding standards. [06:06] We've had that discussion. [06:06] We have [06:06] jtv: I don't think the discussion is over [06:06] I hate it too, but having the consistency and stability is probably more valuable than making it prettier in that case. [06:06] Can the bikeshed not change colour again? [06:07] Exactly. [06:07] StevenK: I've been consistently arguing one thing. [06:07] lifeless: And you keep bringing it up because the bikeshed is a different colour, it seems. [06:08] StevenK: no, I keep bringing it up because I have neither been convinced that the status quo is a good place to be, nor that what I am arguing for is a bad place to be. [06:08] IOW, the bikeshed needs a new coat of paint [06:08] or to not be a bikeshed [06:09] The great thing is, if we change it now, the people who argued the other way get their say again. [06:09] hah [06:10] That can't be bad, right? [06:10] interestingly PEP8 says we are wrong. [06:10] not for function definitions [06:10] but for function calls [06:11] jtv: healthy discussion is fine. [06:11] lifeless: In the words of the Debian Policy maintainers, why must we make hundreds (if not thousands) packages (read as: lines of code) insta-buggy? [06:11] StevenK: you seem you have your troll hat on. Can you take it off please? [06:11] If healthy discussion is fine, by all means let's have more of it. Now excuse me while I get some work done. [06:12] so I was discussing this with my reviewer; the rest of you just piled in :) [06:12] lifeless: Do I have to? :-( [06:12] lifeless: Because you didn't want to listen to your reviewer, which you deny. Seriously, it's pretty clear to me that you're the one with the troll hat on today. [06:16] right, breath taken. [06:16] jtv: there is a huge difference between dialogue and instruction. [06:17] jtv: if our reviewers are having a dialogue, then its up to the coder to listen and choose; making it an absolute must-be-X stops it being a dialogue and makes it instructions. [06:17] lifeless: In my experience, reviewers can do either. [06:18] lifeless: Must-be-X is usually of the form "That is horrid! *needs fixing*" [06:19] I have been wondering whether removing that from reviewers entirely would help folk understand the cultural distinction [06:20] I'm going to go off to the side and finish this specific discussion with wgrant, we can chew the fat about big picture stuff another time [06:27] lifeless: Hm, so anybody can r-c bless truly r-c stuff now? [06:27] yes === almaisan-away is now known as al-maisan [06:28] Great. [06:28] Makes sense. [06:28] more centralised authority undone [06:28] Yup. [06:28] let folk -shock, horror- make decisions. [06:28] Except for formatting :P [06:28] I have a list. [06:50] Eep. [06:51] wgrant: ? [06:51] An OOPS editing a bug. [06:51] Trying to find it. [06:53] I can't find CE anywhere... [06:53] Ah, wampee's rsync config has changed. [06:54] AssertionError: Attribute bug_subscribers not in object > [06:54] <_mup_> Bug #97266: Suggestions menu next to "Choose..." doesn't do anything