[00:00] rubylicious [00:00] mbarnett: Which was the old version of libc-bin? [00:00] That's the only vaguely interesting one, and it looks boring :( [00:01] https://pastebin.canonical.com/44062/ [00:01] lifeless: That's what I would have done. [00:01] mbarnett: Hmmm, that's a bit more interesting. [00:01] less rubylicious [00:02] Yet still very boring. [00:03] mbarnett: Do you know how to apply pygdb to codebrowse? [00:04] i do not [00:04] mwhudson: Could you assist with interrogating codebrowse2? [00:05] mbarnett: We tried to do it yesterday, but the process accidentally got killed. [00:05] loggerhead history doesn't seem to have naything interesting [00:07] wgrant: yeah, but let me grab some lunch so i can concentrate a bit [00:07] unfortunately, we are about to go losaless. [00:07] mbarnett: No spm? :( [00:07] spm: is pretending that flying half way around the world is tiring. [00:07] Heh. [00:11] ah [00:15] well, i do actually need to go assault my face with some dinner. Do you want me to leave codebrowse 2 in its degenerate state and try and hop back later? Or shall i restart it to avoid the potential (at least a bit) of another outage? [00:16] Personally I'd leave it. [00:16] They seem to die in tandem now. [00:17] yeah [00:17] okay [00:20] lifeless: Your closing-bugs-from-changelogs.txt failure is really confusing. [00:22] Ah, I see now. [00:23] wgrant: yes, thus asking for help [00:24] lifeless: I have a quick fix, but I am looking at ways to actually fix the test properly. [00:25] Meh, quick fix it is. [00:26] lifeless: http://pastebin.ubuntu.com/573729/ [00:26] lifeless: It looks for a packageupload with a changesfile in the upload_distroseries. [00:26] cool [00:26] So if we set the upload_distroseries to something else, the test can continue to illegally create multiple packageuploads. [00:26] I'll just wrap this arc and then apply and test [00:29] * wgrant foods. [00:40] wgrant: Thanks, I believe I have successfully comprehended the user's corrupted merge directive email now [00:57] maxb: I can give you more traceback or the email text if you need them. [00:59] The user pasted the email text in the question, and on close inspection it is indeed malformed, so it's ok [01:00] Great, thanks for dealing with that. [01:00] Its an interesting feature, which I keep thinking I might use [01:01] Apart from gpg-signing mails without involving thunderbird is something I've not figured out yet [01:01] Has anybody had an ec2 run Windmill-fail this week? [01:01] maxb: Doesn't bzr do that for you? [01:03] maxb: gpg | mail -s ... ? === MTecknology is now known as billmeye === billmeye is now known as MTecknolgoy === MTecknolgoy is now known as MTecknology [01:11] wgrant: My evil plan to fix my branch from yesterday using + didn't work. :-( [01:13] StevenK: :( [01:13] My evil plan to disable WindmillLayer is getting stronger. [01:15] wgrant: So I suspect the only option is to stop using structured and escape the arguments before interpolating [01:17] StevenK: Or interpolate the recipe manually, before you go into structured. [01:18] Not a bad idea, just struggling how to implement that. [01:19] And trying not to context switch [01:19] Switching between branches of devel and db-devel really sucks [01:28] oh, I meant gpg-signing in a way integrated with bzr send, that doesn't involve saving the merge directive off to a file, manually signing, and sending [01:30] maxb: ask in #bzr, or read the docs ;) [01:32] I read the source, and ended up trying to hack a plugin :-) [01:32] maxb: -why-?! [01:38] wgrant: was there a sqlite change perhaps? [01:38] lifeless: That was my guess. [01:38] But it doesn't look like it. [01:38] But the list I obtained may well not be exhaustive. [01:42] Why? Well, I have to confess, because it felt like something that ought to be possible :-) [01:42] maxb: I mean, it is possible for most mailers I know of, withiut plugins [01:43] maxb: what mailer do you use? [01:57] Project db-devel build #403: STILL FAILING in 5 hr 35 min: https://hudson.wedontsleep.org/job/db-devel/403/ [02:00] do we still use personlocation for anything? [02:00] No. [02:00] sinzui: ^ [02:00] NFI where those queries come from, either. [02:01] File "/home/robertc/launchpad/lp-branches/working/lib/lp/registry/model/person.py", line 708, in time_zone [02:01] if self.location is None: [02:01] File "/home/robertc/launchpad/lp-branches/working/lib/lp/services/propertycache.py", line 116, in __get__ [02:01] value = self.populate(instance) [02:01] time localisation [02:02] Oh, that's on PersonLocation? :/ [02:02] I always assumed that was just the lat/lon. [02:02] apparently not [02:03] guess how many queries all_distro_archive_ids creates [02:03] That could be a bug. [02:03] Four? Five? [02:03] primary, partner, debug.. maybe only three. [02:03] actually two [02:03] :( [02:04] from this call to getCurrentSourceReleases [02:04] hmm [02:04] time ot land tihs and get the branch changing that landed [02:04] Ahh [02:08] https://code.launchpad.net/~lifeless/launchpad/bug-724033/+merge/51676 [02:08] Has anybody looked at the Windmill SNAFU in depth? [02:09] I am wondering if the xvfb-run issue is in fact a symptom of the killing, not the hang. [02:09] xvfb-run issue? [02:10] lifeless: See the end of https://lpbuildbot.canonical.com/builders/lucid_lp/builds/669/steps/shell_6/logs/stdio [02:10] We get an X error. [02:11] I've been presuming that was seen at the start of the hang. [02:11] But maybe it's because of the mass murder after the hang is detected. [02:13] I don't see an X error [02:13] oh [02:15] There is the occasional thread leak on ec2. [02:15] yes, I think the X process went before the tes [02:15] t [02:15] And locally it just completely fails. [02:24] StevenK: mumble? [02:25] thumper: I was about to have some lunch, but sure. [02:25] StevenK: it'll be quickish [02:32] wgrant: thanks, thats off to ec2 now [02:32] lifeless: Great. [02:34] who is meant to be OCR today [02:34] WILLIAM! [02:35] wgrant: I can haz review please [02:36] Oh, so I am. === wgrant changed the topic of #launchpad-dev to: Performance Tuesday | https://dev.launchpad.net/ | firefighting: - | On call reviewer: wgrant | https://code.launchpad.net/launchpad-project/+activereviews [02:37] lifeless: Ah, I see you are realising the precaching resultset this time :P [02:38] ftw [02:38] on the rusty scale, that interface is problematic [02:38] Yes [02:51] https://code.launchpad.net/~thumper/launchpad/fix-mantis-warnings-bug-218384/+merge/51679 anyone? [02:51] https://code.launchpad.net/~wgrant/launchpad/squash-cw-protocolerrors/+merge/51680, also anyone? [02:51] thumper: Looking. [02:52] wgrant: thansk [02:52] wgrant: I'll look at yours [02:53] Thanks. [02:56] wgrant: do we need the transaction.commit(), or is a store flush enough? [02:56] thumper: checkwatches checks that no transaction is active. [02:56] So it has to be a commit. [02:56] ok [02:57] grr [02:57] 0 tests run in 4:11:34.318468, 0 failures, 0 errors [02:57] wgrant: has your fix landed yet ? [02:57] lifeless: merge devel. [02:58] wgrant: shouldn't the ec2 image start with devel? [02:58] wgrant: or is it client side? [02:58] thumper: Yes, but ec2test-remote.py is copied from the local branch. [02:58] wgrant: ok [02:59] wgrant: I had a windmill fail in ec2 [02:59] lifeless: Could you mentor https://code.launchpad.net/~thumper/launchpad/fix-mantis-warnings-bug-218384/+merge/51679? [02:59] lifeless: When? [03:00] today [03:00] Forward pls. [03:00] I want all the data I can get. [03:00] sent [03:00] How do I actually get a lazr-js branched landed to trunk (after the mp step)? [03:01] grr [03:01] wtf: [03:01] print test_tales("branch/fmt:bzr-link", branch=branch) [03:01] Differences (ndiff with -expected +actual): [03:01] - lp://dev/fooix [03:01] ? ^ [03:01] + lp://dev/~eric/fooix/bar [03:01] ? +++++++++++ +++++++++ ^^^ ++++++ ++++ [03:03] lifeless: Dev focus setting fail? [03:05] * thumper agrees with wgrant [03:05] huwshimi: commit to trunk [03:05] huwshimi: I don't think it is PQM controleld [03:06] thumper: Thanks. [03:08] thumper: Thanks. [03:14] thumper: could you eyeball bug 702524? [03:14] <_mup_> Bug #702524: Merged branches keep 'Development' status when their merge proposal is marked as 'Merged' < https://launchpad.net/bugs/702524 > [03:14] lifeless: yep [03:14] lifeless: although I think I know the answer [03:14] thumper: can you mentor wgrants review of https://code.launchpad.net/~lifeless/launchpad/bug-724033/+merge/51676 ? [03:15] wgrant: Can haz review? https://code.launchpad.net/~stevenk/launchpad/db-recipe-description-really-this-time/+merge/51682 ? [03:16] lifeless: Can you mentor my review of StevenK's? [03:16] wgrant: :-( [03:16] wgrant: I know where you live ... [03:17] :( [03:18] wgrant: yes [03:18] just chewing through 3 hours mail ;) [03:18] Thanks. [03:19] wgrant: What was your plan with adding recipe first? [03:20] lifeless: yes, I'll look at wgrant's review [03:21] StevenK: You could put a %%s in the template, format the other args into it using structured(), then format the escaped recipe in using %. [03:21] *or* you could fix structuredstring to work nicely with other structuredstrings. [03:23] Does anybody have a valid objection to disabling WindmillLayer until someone works out what is wrong? [03:24] * StevenK tries to remember how to do type checking in Python [03:24] I have never managed to get it to run reliably on any local machine, it is entirely broken on Natty, it leaks threads sometimes in random tests, and sometimes hangs entirely. [03:25] StevenK: isinstance? [03:28] wgrant: Revision 12469 can not be deployed: needstesting [03:29] looks like its half-qaed - were you looking at that ? [03:29] lifeless: One was a test change. [03:29] The other needs a LOSA or somebody to decide that we don't care. [03:30] is there a library of matchers for use with storm anywhere? [03:31] there are some in the lp tree [03:31] like StormRecorder + HasQueryCount [03:31] lifeless: Do you have any hints for debugging thread garbage? [03:31] I'm more interesting in query/result set stuff currently [03:32] https://code.launchpad.net/~james-w/launchpad-work-items-tracker/storm/+merge/51684 [03:32] (includes unit testing of lplib-using code) [03:32] I trhink the storm folk have started using matchers [03:32] so storm specific stuff perhaps land there? [03:37] nothing jumps out from the filenames [03:39] GRRRR [03:39] stupid dist utils [03:41] * thumper stabs it in the face [03:45] oops [03:46] oh well, hopefully it won't break the build [03:46] lifeless: Hm? [03:47] wgrant: I lp-landed rather than ec2 landed [03:47] Heh. [03:47] lifeless: do you know anything about setup.py, tests and the additional_tests method? [03:47] thumper: a -little- [03:47] thumper: lazr-js ? [03:47] lazr.enum [03:47] ah [03:47] trying to run: ./setup.py test [03:47] should that cause the additional_tests method to be checked? [03:48] because it isn't [03:48] I have a pdb break point in there [03:48] and it isn't breaking [03:48] http://mail.python.org/pipermail/python-checkins/2006-March/050452.html [03:48] I'm trying to get it to run the actual tests [03:49] so additional_tests is for non unittest tests [03:49] what non unittest tests do you have? [03:49] it is to load doctests [03:49] ah, what it means is 'tests that loadTestsFromName does not find' [03:50] so, I personally /hate/ setup.py's test support [03:50] I'm starting to too [03:50] I would always just use a test_suite method [03:51] lifeless: O hai. Can haz review of wgrant's review of my MP? [03:52] I did? [03:52] lifeless: I can't see your scribble on https://code.launchpad.net/~stevenk/launchpad/db-recipe-description-really-this-time/+merge/51682 ? [03:53] done [03:53] must have glitched [03:53] lifeless: Thanks, tossing it at ec2 [03:54] lifeless: Do you know how to get details on threads that are leaked by tests? [03:54] wgrant: yes [03:54] I've stuck a pdb in there, but I can't get much useful from the Thread. [03:55] ok [03:56] so depending on how it was constructed [03:56] subclasses have a replaced run() (IIRC), typical uses supplies a callable to the thread which its run() will call [04:00] wgrant: Can you see any clues on http://pastebin.ubuntu.com/573778/ ? [04:00] wgrant: you can also attach gdb [04:00] wgrant: and get pretty good info from thread apply all pybt [04:00] lifeless: Thanks. [04:02] I thought we'd changed thread leaks to be non fail though [04:04] They are a non-fail. [04:04] But Windmill sometimes leaks them, and Windmill sometimes fails. [04:04] I am hoping that this is not a coincidence. [04:09] wtf- librarian.txt fail in my /branches optimisation branch [04:10] lifeless: I saw one of those a week ago. [04:10] What is the error? [04:10] A 200 UploadFailed? [04:10] yeah [04:11] passes locally [04:11] Yeah. [04:11] And on ec2 >9/10. [04:16] wgrant: Can haz clue? [04:16] StevenK: recipe_build_details is probably raising an AttributeError or similar. [04:17] Hrm. A better traceback would be awesome. [04:18] TypeError: not all arguments converted during string formatting [04:18] :-( [04:19] Ah ha, I get it. [04:19] \o/ another one bites the dust [04:22] wgrant: so, bug 688130 - what needed losa answering? [04:22] <_mup_> Bug #688130: Statistics clean-ups and extra tests

< https://launchpad.net/bugs/688130 > [04:22] lifeless: It needs the rosetta pofilestats script to be run. [04:27] I think my branch:+index patch is reasonably successful :) [04:28] https://code.qastaging.launchpad.net/~software-store-developers/software-center/trunk/+index rendered first hit on qastaging. Which is awesome, if I do say so myself. [04:30] Project devel build #488: STILL FAILING in 5 hr 29 min: https://hudson.wedontsleep.org/job/devel/488/ [04:32] thumper: Actually it appears that we do use PQM for lazr-js [04:33] Jenkins, you make me sad. [04:33] checkwatches.txt passed, but a Windmill test failed? Baaaaah [04:34] thumper: Does that mean landing with lp-land or something? === MTecknology is now known as pfSensory [04:36] huwshimi: pqm-submit [04:36] wgrant: Thanks === pfSensory is now known as MTecknology [04:37] bzr pqm-submit -m "[r=foo] bar" --public-location=bzr+ssh://bazaar.launchpad.net/~path/to/branch --submit-branch=bzr+ssh://bazaar.launchpad.net/~path/to/trunk [04:39] No fair actually landing with '[r=foo] bar' [04:39] wgrant: Thanks, I was just trying to figure all that out :) [04:40] StevenK: Oh, cmon [04:42] lifeless: bug-724033 seems to have an empty commit message. [04:48] wgrant: Sucess! [04:49] StevenK: Exccellent. [04:49] wgrant: see https://pqm.launchpad.net/ [04:50] lifeless: Ah, great. [04:52] StevenK, wgrant: DistroSeries.parent_series seems to have shifted in meaning a bit… should we keep the old "previous series in same distro" meaning or get rid of it? [04:52] jtv: It hasn't shifted. [04:53] Any derivation stuff that is using it now is not altering its meaning; it's just using it until we introduce something to replace it. [04:53] Do we expect the replacement to work in the same way? [04:54] Awwwww. The user on prod with deleted recipes doesn't exist on qastaging [04:54] That is not known. [04:55] wgrant: okay, do we expect to have DistroSeriesDifferences between e.g. maverick and natty? [04:56] jtv: I don't believe so. [04:56] Not in the initial implementation, at least. [04:56] That isn't the point [04:58] Publishing details [04:58] * Built by deleted recipe for Jelmer Vernooij. [04:58] \o/ [05:00] Yay. [05:00] * wgrant is QAing the greatest OOPS fix of all time. [05:01] wgrant: Oh, the checkwatches stuff? [05:02] StevenK: No, the loggerhead thing. [05:02] Unfortunately it is slightly non-trivial, because both qastaging and staging are borked. [05:11] Sigh. Can't hear myself think. [05:13] Some ungrateful person with a leaf blower is outside. Doesn't he know that the BoM forecast high winds for tonight, so it's POINTLESS [05:13] Heh. [05:14] Hmm, this is slightly upsetting. [05:15] wgrant: I fervently hope that remark is not connected to SteveK's. [05:15] No, codebrowse. [05:15] Ah well, talk of upsetting. [05:16] On the bright side, "parallel a-f" has landed on qastaging. [05:16] And mawson. [05:16] And it works. [05:17] You tested it? [05:17] Yes. [05:17] Great! Thanks. I'll mark it qa-ok then. [05:17] I already did. [05:17] Didn't I? [05:17] Oh, looks like you did. :) Pages don't refresh very well here. [05:21] GAH! He's back! [05:22] I have somewhere that guy can store the leaf blower when he is finished using it ... [05:29] D: [05:29] Codebrowse, you suck. [05:33] lifeless: So, you know how that loggerhead fix was meant to stop loggerhead from OOPSing when haproxy checked it? [05:35] StevenK: do SPRs ever get deleted? I'm asking because SPPH.supersededby isn't indexed. [05:39] StevenK: do SPRs ever get deleted? I'm asking because SPPH.supersededby isn't indexed. [05:41] jtv: Not at the moment. [05:41] :( [06:10] No IS for a week and a half makes me very sad. [06:14] wgrant: yes? [06:15] lifeless: Well, my testing showed that it in fact introduced a new OOPS. It does, but in a much rarer case than the one it fixes. [06:15] Bug #726985 [06:15] <_mup_> Bug #726985: codebrowse OOPSes with GeneratorExit when connection closed early < https://launchpad.net/bugs/726985 > [06:15] I think we should still deploy. [06:15] Also, feel like waking elmo? [06:15] wgrant: is that if you abort from a client ? [06:15] guava has had a heart attack again. [06:15] *sigh* [06:15] is s gsa around ? [06:15] I pinged pjdc/bradm a few minutes back. [06:15] No response. [06:16] give them 10 [06:16] then dig up the number and call [06:16] I'm mid-dinner, sorry [06:16] k [06:16] lifeless: I believe the client has to be in the DC, bypassing haproxy. [06:18] bouncing [06:18] elmo: tiaz might already be. [06:18] (see #is) [06:39] StevenK: do you have a clear idea of what DSDJ will look like, database-wise? [06:43] jtv: One row [06:44] StevenK: one row per object? Ground-breaking. [06:44] Maybe I'm just not aware of the actual design questions you're facing. :) [06:44] jtv: A DSD already has an interface and a schema ... [06:45] Oh, the job itself. Er. Not sure. :-) [06:45] Yes, the part I'm quite interested in right now is whether, for the initial population of DSD, I can create DSDJ rows directly in SQL. [06:46] As tends to happen, a join structure has formed itself in my mind's eye quite early on and it takes ages to figure out any reasons why it might be wrong. :)/ [06:46] I was also wondering about supersededby… When is a SPPH "superseded" exactly? [06:47] I imagine I can ignore ones that have that column set, but might that also be a sufficient condition? [06:49] In other words, if I selected from SPPH on sourcepackagename and distroseries where superseded is null, would that give me (more or less) just the SPPHs I wanted or would there be a lot of dead wood in there? [06:49] Selecting on supersededby is not going to give you anything useful. [06:50] Ah [06:50] Conversely, will I be able to ignore ones where supersededby is not null? [06:50] You should filter on status, if anything. [06:50] (And also, why is supersededby not indexed? Do SPRs never get deleted?) [06:50] SPRs do not get deleted at the moment. [06:51] What are you trying to do, exactly? [06:51] And why? [06:51] Trying to figure out how to get the "latest" SPPH for a given source package. [06:51] jtv: due to the bug about not having a sourcepackage object in the db [06:51] Source package, source package name, or distribution source package? [06:51] jtv: this is actually hugely expensive [06:52] getCurrentSourceReleases does it [06:52] wgrant: source package [06:52] jtv: Are you sure? [06:52] lifeless: ah, thanks [06:52] well [06:52] it gets what most people will think of as a source package :) [06:52] which is not a SourcePackage or DistroSeriesSourcePackage [06:52] If you have a SourcePackage, you should be able to call .currentrelease. [06:52] wgrant: I was until you asked that. :) A given sourcepackagename for a given distroseries… I *think* that's what I want! [06:53] Right, that's an actual SourcePackage. [06:53] except its versioness [06:53] which is mondo confusing modelling [06:53] lifeless: I imagine what most people think of as a source package is what we call a distributionsourcepackage… I've never heard of DistroSeriesSourcePackage. [06:54] Most people would think of an SPR as a source package. [06:54] jtv: no, most people think of source packages as 'source package release' [06:54] Ah [06:54] Indeed. [06:54] jtv: there is a 'most recent source package release' [06:55] I'd like to really overhaul the soyuz model [06:55] but not while we're in the hole [06:55] lifeless: How? [06:55] So should I be looking for the most recent SPR instead of the most recent SPPH when looking for distroseries differences? [06:56] lifeless: for someone who's postponing stuff until we're not in the hole, you sound suspiciously eager to get back into it. :) [06:56] More power to you, but… sounds like a tall order. [06:56] SELECT id FROM spph WHERE spph.distroseries = yourfavouritedistroseries AND spph.pocket = 0 AND status IN (1, 2) ORDER BY datecreated DESC LIMIT 1 [06:56] Oh, and spph.archive = yourfavouritearchive [06:57] wgrant: I don't have a specific schema handy. [06:57] wgrant: would the pocket and archive matter? [06:57] What are your vague thoughts, then? [06:57] wgrant: I would start by gathering all the actual use cases the distro *want* - partly what we do today, partly their wishlist. [06:58] and design to directly match that as a fast schema, with efficient solutions across the board. [06:58] and finally look at how to migrate. [06:58] StevenK: I can list some problems [06:59] jtv: The archive is essential. The pocket is less essential, but you probably still want to restrict it. [06:59] StevenK: we infer data on every operation; we have 1:M relationships that our business rules want to be 1:1; our schema supports invalid states [07:00] jtv: Since we don't know what it means to have a DSD to a non-Release pocket. [07:00] wgrant: oh do I get all the PPAs if I don't restrict on archive? Also, another thing that I think matters is the package. :) [07:00] jtv: Um, yeah, package would be nice too. [07:01] So that's a join through SPR I guess. [07:01] Right. [07:01] StevenK: our schema should naturally constraint the states to be valid; data learnt should not be reinvestigated on every operation [07:01] StevenK: we should be able to do things like '100 most recent daily builds' in 3-4ms. [07:02] StevenK: without 'denormalisation' [07:02] wgrant: if I neglected to restrict by archive, would that mean that I would get all the PPAs as well? [07:02]