[00:00] bah [00:00] I selected a component of main [00:00] status new [00:00] 5.53 seconds [00:00] main is tiny. [00:00] Doesn't count. [00:00] Project devel build #518: FAILURE in 4 hr 48 min: https://hudson.wedontsleep.org/job/devel/518/ [00:00] ok [00:00] universe works [00:03] and its the - guess what - spph join thats the problem [00:03] sinzui: if you are still online, can you tell me how to construct a PrettyOverlay but with a different BOUNDING_TEMPLATE. I'm not sure how to override the default one [00:04] oh. [00:04] That just went over my head. I think I got a nose bleed thinking about it [00:04] sorry :-) [00:05] i want to replace the close (x) with a pair of (tick) (cross) buttons [00:05] since the user needs to be able to poke some widgets on the overlay and then either accept or cancel [00:06] i could get the close div after construction and insert a new node [00:06] that would do it [00:07] wgrant: whats silly is that we get back no packages [00:07] getBuildRecords is a mess :( [00:08] lifeless: ... oh? [00:09] wgrant: see actual time=0.011..0.011 rows=0 loops=1276160) [00:09] perhaps thats just misleading [00:09] we do get 76 bugs === Ursinha-afk is now known as Ursinha [00:10] Ursinha: Hi. [00:11] hi [00:11] it's carnaval, so this might be an automated response :P [00:11] haha, OK. [00:12] wgrant, what's going on? [00:12] wallyworld: I'll take a look [00:12] thumper: ta [00:12] Ursinha: I was just wondering what sends the daily error reports these days. launchpad-errors-yesterday in lp:oops-tools seems to be out of date. [00:13] * Ursinha checks lpqateam's crontab [00:13] Ursinha: is matsubara_ around ? [00:14] lifeless, unlikely, 9pm of a carnaval's tuesday night [00:14] heh, ok :) [00:15] wgrant, so, it's a script called report (heh) [00:16] generated automatically by oops-tools [00:16] wgrant, why? [00:16] lifeless, do you need something specific? [00:17] Ursinha: I have a oops tool merge proposal, hadn't heard anything in a few days [00:17] Ursinha: I want to change a prefix. [00:17] Was hoping to change it there too. [00:17] wgrant, hm, prefixes now are changed using an admin django page [00:17] Ah, handy. [00:17] prefixes are loaded dynamically from oops-tools database [00:18] I'm sure there's a wiki page explaining that [00:18] * Ursinha looks [00:18] there is [00:18] matsubara mailed it to me :) [00:18] wgrant, https://dev.launchpad.net/Foundations/QA/OopsToolsSetup [00:20] Ursinha: Thanks. [00:20] Who has access to add a new prefix to a report? [00:24] ok, got that down to 800ms. [00:24] nice [00:24] Nice. [00:24] What did you change? [00:25] Hm, build breakage. [00:25] Same one us hudson. [00:25] Intriguing. [00:25] Because that's a spurious failure. [00:25] Must be time-based. [00:25] https://bugs.launchpad.net/launchpad/+bug/731679/comments/3 [00:25] <_mup_> Bug #731679: distribution:+bugs timeout searching for bugs by component < https://launchpad.net/bugs/731679 > [00:25] I've seen this happen before, but dismissed it as coincidence :/ [00:26] time to dig into storm, I just wish it was more pleasant [00:27] lifeless: Can't you easily word that as a subselect? [00:28] wgrant: you'd think so [00:28] Does it break the plan? [00:31] sinzui: still around? [00:31] I am [00:31] I'm just encoding that video :) [00:33] thumper: I was just about to have dinner. I will return in about an hour. [00:33] sinzui: ack [00:55] wtf [00:56] ??? [00:56] This test failure. [00:56] It is real. [00:56] But completely unrelated. [00:56] 12555 must have changed test order somehow. [00:57] 'bin/test -m lp.services.job.tests.test_runner -1cvvt TestTwistedJobRunner.test_timeout' reproduces the failure even before then. Dropping the -m works fine. So it needs some test module loaded before it will work. [00:58] Hmm, ampoule can't import _pythonpath. [01:07] mtaylor: does mysql 5.1 support 'with foo as (...) select ... where my.id in (select thing from foo)' ? [01:31] Gnrgh. [01:32] ? [01:33] This test. A subprocess fails if lp.testing imports anything from canonical.testing.layers. [01:33] Must be a circular import, I guess, but Twisted is good at swallowing details. [01:34] eugh [02:08] \o/ 176 / 0 LanguageSet:CollectionResource:#languages [02:08] 58 / 155 BugTask:+index [02:11] lifeless: I have never seen 'with foo as (...)' - but I could of course have just missed it [02:12] mtaylor: thanks, I will write this as a postgresql only tested feature then :) [02:13] lifeless: my pleasure! [02:17] Excellent, the top 50 exceptions now encompass all with multiple occurrences. [02:27] https://bugs.launchpad.net/storm/+bug/731739 [02:27] <_mup_> Bug #731739: with foo as (select...) SELECT ... not supported < https://launchpad.net/bugs/731739 > [02:29] now to figure out how to do a snapshot of storm [02:30] You've done it already? [02:30] yes [02:30] sufficient for our needs anyhow [02:32] https://code.launchpad.net/~lifeless/storm/with/+merge/52630 [02:47] lifeless: You linked to the Postgres 9.0 docs, it's supported earlier? === StevenK changed the topic of #launchpad-dev to: Performance Tuesday! | https://dev.launchpad.net/ | firefighting: - | On call reviewer: StevenK | https://code.launchpad.net/launchpad-project/+activereviews [02:50] StevenK: It's new in 8.4. [02:51] That's probably okay, then. [02:53] StevenK: its supported wherever lp is supported [02:53] I seem to be getting lp.services.job.tests.test_runner.TestTwistedJobRunner.test_timeout failing every time on EC2 [02:53] wgrant is working on that [02:53] thumper: Yes, I'm looking at that. [02:53] others getting this? [02:53] wgrant: awesome, thanks [02:53] About to give up and just work around it, though. [02:53] Jenkins is to [02:53] Ampoule is almost more evil than I am willing to deal with. [02:53] wgrant: want to talk through it? [02:54] thumper: That test breaks with an unknown subprocess failure if lib/lp/testing/__init__.py imports anything from canonical.testing.layers (r12555 introduces one such import). [02:55] thumper: The test also fails when run on its own with -m lp.services.job.tests.test_runner, even before r12555. [02:55] hmm... [02:55] But with a slightly different failure. [02:55] it was failing intermittantly [02:55] It always has, yes. [02:55] and there is a bug for that [02:55] There is. [02:55] Bug #505913 [02:55] <_mup_> Bug #505913: TestTwistedJobRunner.test_timeout fails intermittently

< https://launchpad.net/bugs/505913 > [02:56] how are you working around it? [02:56] Removing the import. [02:57] Running with -m causes ampoule to fail to import _pythonpath. If I stop it from attempting to use _pythonpath, we get the same error that r12555 introduces. [02:57] Any useful diagnostic information is swalled by Twisted, though. [02:57] +ow [02:57] hmm... [02:58] It also still happens if I stop the tests from running in a subprocess. === lifeless changed the topic of #launchpad-dev to: https://dev.launchpad.net/ | firefighting: - | On call reviewer: StevenK | https://code.launchpad.net/launchpad-project/+activereviews [03:00] thumper: it seems one is not allowed to modify a CollectionField attribute via the web service? i get a "You tried to modify a collection attribute" error [03:00] That's right. [03:00] You might want a List instead. [03:01] eg. IBug['tags'] [03:01] wgrant: thanks. will have a look [03:02] wallyworld: Modifying CollectionFields is hard... what does it mean to remove an object from a resultset? [03:02] wallyworld: hmm [03:02] wallyworld: I'm sure leonardr suggested to me that we'd need add and remove methods to post to. [03:02] wallyworld: or perhaps an update? [03:02] wallyworld: you're one of the first attempting this [03:02] wgrant: i thought a collectionfield attribute were like lists [03:03] wallyworld: so you get to blaze the way [03:03] thumper: i'm replacing the entire field's value so i may as well change it to a List [03:04] wallyworld: Note that it changes the representation significantly; it becomes inline instead of a link. [03:04] This has performance implications. [03:05] wgrant: i'm not familiar with the terminology in this context - inline vs link [03:05] wallyworld: I think you want an update method for the API [03:06] thumper: so leave it as CollectionField and use an update method? if List fields are mutable, given the size of the list is small (<20), does it make sense just to go with that for this case? [03:07] Is this for SourcePackageRecipe.distroseries? [03:08] wgrant: it is for recipe build series [03:08] Right. [03:08] List is probably not appropriate there. [03:08] I would have (add|remove)Series named ops, I think. [03:09] hmm. seems like overkill given the size of the collection and the fact that we really just replace one version with another [03:09] by version i should say value [03:10] add/remove is typically more suited to larger collection sizes [03:10] I think updateSeries should be sufficient [03:10] Store.find(LibraryFileAlias).set(content=None) [03:10] TypeError: unbound method find() must be called with Store instance as first argument (got SQLObjectMeta instance instead) [03:10] * StevenK grumbles more [03:10] StevenK: Store.of(something).find [03:10] StevenK: IStore(LibraryFileAlias) [03:10] Or that. [03:11] wgrant: I've reverted to getCandidates returns ids [03:12] wallyworld: Hmm, a collection can have custom named operations. [03:12] Possibly not collection fields, though... hmmm. [03:13] wgrant: i'm grepping the code as we speak to figure out how to do it :-) [03:13] i'll do a named_post [03:13] SPR.updateseries() or something [03:14] wallyworld: Right, that's easy enough, but from looking at lazr.restful it seems like it might be possible to do something on the collection itself. [03:14] But updateSeries() is probably simpler. [03:15] i'll have a poke around. i'm generally against fine grained collection updates [03:15] especially when the semantics of this is a complete replacement [03:15] of the collection's values [03:16] Right. [03:16] where it would make sense for example is adding/removing a single person from a team [03:18] wgrant: not only done, done and working. [03:18] lifeless: Excellent! [03:23] I can has reviewer? https://code.launchpad.net/~lifeless/launchpad/bug-731679/+merge/52631 [03:23] * StevenK looks [03:24] No reviewer for you. Come back when you have a diff [03:24] :-P [03:24] don't blame me [03:25] lifeless: Does it require the storm changes, or are those seperate? [03:25] requires em [03:25] I've a snapshot ready to commit to the downloadcache [03:26] StevenK: it has a diff [03:27] Yes, I'm reading it [03:27] thanks! [03:27] lifeless: orig_store, and then you call orig_store.with_(with_clause) when you'd already set store = store.with_(with_clause)? [03:27] That version number looks like a lie. [03:27] StevenK: different with_clause [03:28] And yes, I'd prefer that the storm changes get approved and landed first [03:28] StevenK: note the for loop [03:28] StevenK: not going to happen [03:28] StevenK: storm is basically stalled. [03:28] StevenK: I've 0 intention of letting us stall just because its stalled. [03:28] Orsum [03:30] feel free to review my storm changes though [03:30] thumper: https://code.launchpad.net/~lifeless/launchpad/bug-731679/+merge/52631 when you have a tick [03:31] but its an odd coding style there [03:31] lifeless: I glanced at them, but I'm not comfortable reviewing [03:31] * thumper has two ticks [03:31] thumper: thanks [03:32] lifeless: what was the concern addressed on irc? [03:32] thumper: which concern? [03:32] I still think that version number is wrong. [03:32] lifeless: the one that StevenK says in the code review [03:32] thumper: I'd prefer the changes to Storm landed [03:32] [14:28] < lifeless> StevenK: storm is basically stalled. [03:33] Ergo ... [03:33] I don't care if they land, but that 0.18.0.99 is deceptive. [03:33] wgrant: pick a version, I'll roll with it [03:33] wgrant: note that it does string algebra [03:34] wgrant: so 0.18.0.0r14 will be stupidly painful to make [03:34] lifeless: We normally use a suffix like -lp-r1234 [03:34] wgrant: -how- [03:34] lifeless: How are you building it? [03:34] wgrant: like I say, storms setup.py / __init__ do algebra on a constant [03:34] wgrant: setup.py sdist [03:36] Store.of(LibraryFileAlias).find().set(content=None) [03:36] All our previous custom Storms have had a good suffix. [03:36] That doesn't work either, am I doing it wrong? [03:36] StevenK: find(LibraryFileALias) [03:37] lifeless: What about 'setup.py egg_info -b-lp-r1234 sdist'? [03:37] wgrant: that worked, I'll use it [03:40] wgrant: Still ends up with AttributeError: 'NoneType' object has no attribute 'find' [03:40] StevenK: IStore(LibraryFileAlias)... [03:41] Ah, didn't bother to look further than the find() [03:41] StevenK: Store.of() only applies to instances that have either come from a store or been added to a store. [03:41] Store.of(someobject) [03:41] IStore(someclass) [03:42] Ahh [03:42] StevenK: this was the only thing wrong with your example I think [03:42] StevenK: also, when something doesn't work, show how :) [03:42] thumper: thanks [03:43] IStore(LibraryFileAlias).find(LibraryFileAlias).set(content=None) results in: [03:43] CompileError: Don't know how to compile type Reference of [03:44] StevenK: don't use reference columns [03:44] contentID [03:44] Huh, that should work. [03:45] * wgrant blames the SQLObject layer. [03:45] Trying contentID [03:45] reference columns are basically broken in storm and upstream are not receptive [03:45] wgrant: no, its not sqlobject layer [03:48] * wgrant gives up. [03:48] wgrant: Oh? [03:49] The ampoule thing. [03:49] * wgrant is testfixing. [03:50] Rargh, how can setting content == NULL for every LFA *still* have SPRs being processed! [03:51] clearly something doesn't do what you expect [03:51] I'm tempted to blindly ignore 'has no DSC' and continue [03:55] wgrant: is 681770 a dupe? [03:55] lifeless: Already done. [03:56] cool [04:00] wgrant: Any thoughts? [04:00] StevenK: Can you push up your latest changes? [04:03] wgrant: Done. [04:04] StevenK: https://code.launchpad.net/~wgrant/launchpad/testfix/+merge/52632 [04:04] Looking at yours. [04:05] Same [04:06] thumper: https://code.launchpad.net/~wgrant/launchpad/testfix/+merge/52632 [04:14] * StevenK nails jtv to the Internet [04:14] lifeless: Thanks. [04:14] It probably won't help, though. [04:15] StevenK: no, but at least you got some fun out of it [04:15] wth is this exposed TemporaryStorageManager:CollectionResource:#temporary_blobs [04:17] Possibly only so apport can upload to it. [04:17] I can't think why an API client would want to read the collection. [04:18] StevenK: Is it only test_hourly_script failing for you? [04:18] wgrant: Yes. [04:18] StevenK: Are you killing the LFAs there? [04:18] Ah, yes. [04:20] wgrant: and yet, they are [04:21] lifeless: Grep logs and strangle them? [04:24] StevenK: I wonder if LaunchpadScriptLayer might not use the right DB. [04:31] wgrant: Revision 12550 can not be deployed: needstesting [04:31] Preload BinaryPackageNames before listing successful copies' displaynames. [04:32] lifeless: Fixed. [04:35] wgrant: I wonder if using IMasterStore would make any difference. I suspect not. [04:36] sigh [04:36] overengineering [04:36] https://code.launchpad.net/~gmb/launchpad/processapportblobjob-api-bug-513191/+merge/20209 [04:36] StevenK: I doubt it. [04:37] lifeless: Ah, of course. That doesn't need a listable collection, though. [04:37] StevenK: Check the name of the DB that each uses. [04:37] wgrant: precisely. [04:38] wgrant: How can I do that? [04:38] StevenK: wgrant: tests that use IMasterStore need to use it consistently or they will break [04:38] you can't use IStore in any helper :) [04:38] ditto ISlaveStore [04:39] lifeless: This one runs a script. [04:39] wgrant: I also particularly love the 'last 50' statement which is a total lie. [04:39] lifeless: Ah, it's not actually a security hole, because the blobs don't have launchpad.View. [04:39] So the list is empty. [04:40] wgrant: but are filtered in the appserver [04:40] lifeless: Sure, so it's slow. But my main concern is that we aren't leaking them. [04:40] wgrant: sure, thats a concern too [04:40] just noting [04:40] this is basically utter crack [04:41] What happens if you remove the collection_default_content? [04:41] wgrant: look at the end of lib/canonical/launchpad/database/temporaryblobstorage.py [04:41] Or does export_webservice_collection require it? [04:41] wgrant: needs it AFAIK [04:41] I've a faster query [04:41] lifeless: Don't use it. [04:41] Return an empty list instead. [04:41] wgrant: wow, r12557 is >:( [04:42] mwhudson: YES. [04:42] did you annotate for great justice? [04:42] Which bit? [04:42] who added the flamingly incorrect comment [04:42] The circular import one? [04:43] That was added in the problematic revision. [04:43] ah ok [04:43] anyway, i have a bbq to get to, better than being frustrated by software [04:43] wgrant: you've checked you don't see the blobs [04:43] So, I spent a couple of hours trying to work out what was up, but I only ended up finding out that buildout, Twisted and Ampoule all need to DIE. [04:43] wgrant: right ? [04:43] lifeless: Yes. [04:43] wgrant: but - how does the 'is blob X processed' work then ? [04:44] lifeless: lazr.restful always checks for launchpad.View. [04:44] Regardless of what the security adapters say. [04:44] Anyone can view a blob, if they know its UUID. [04:44] wgrant: but does't htat also check for View ? [04:45] interface="canonical.launchpad.interfaces.temporaryblobstorage.ITemporaryBlobStorage" /> [04:45] So, no. [04:46] glue the dots together gfor me [04:46] lazr.restful always checks for launchpad.View. But the security declarations can make anything publicly available. [04:46] lazr.restful checks for launchpad.View before it shows an object in a collection representation, that is. [04:47] For returning a single entry resoruce it will respect the security proxies. [04:47] ah [04:47] so collections are spethial [04:47] Yeth. [04:49] * lifeless considers lp-landing this [04:49] * wgrant defers getBuildRecords until tomorrow, and instead fixes a couple of quick checkwatches OOPSes, since even checkwatches is likely to incite less violence than getBuildRecords and test_timeout. [04:49] lifeless: What have you changed? [04:50] wgrant: lp:~lifeless/launchpad/bug-731736 [04:50] Project devel build #519: STILL FAILING in 4 hr 49 min: https://hudson.wedontsleep.org/job/devel/519/ [04:51] https://code.launchpad.net/~lifeless/launchpad/bug-731736/+merge/52634 [04:52] lifeless: As long as nothing else uses that property, and all the blob webservice tests pass. doit. [04:52] wgrant: I've changed one test which psases and the default collection method which there are no other references to outside of the webservice definition and implementation [04:52] Right. [04:52] Just checked the diff in LH [04:52] and there are no other hits for the webservice collection name either [04:53] You have my codestar. [04:53] heh [04:53] I was just self reviewing :) - thanks though [04:54] That works too. [04:54] sounds like bsg (codestar) [04:54] That was the feel I was going for. [04:55] so top timoeuts 1, 2, 3, all have patches pending deploy [04:55] And 1 and 3 are actually fixed. [04:55] 2 maybe not. [04:56] time to look at 4 again [04:56] bah [04:56] no oops in the report [05:02] wgrant: Grabbing the DB name in garbo-hourly is launchpad-main-master [05:02] hah [05:02] they missed a status [05:02] field.status: [u'NEW', u'INCOMPLETE_WITH_RESPONSE', u'INCOMPLETE_WITHOUT_RESPONSE', u'INVALID', u'WONTFIX', u'EXPIRED', u'CONFIRMED', u'TRIAGED', u'INPROGRESS', u'FIXCOMMITTED', u'FIXRELEASED'] [05:02] StevenK: The postgres DB name is the interesting one. That's the store name. [05:03] lifeless: Just opinion missing? [05:03] wgrant: Did you miss my question of how to actually get it, then? [05:03] wgrant: yeah [05:03] wgrant: this is a /bugs/ search. [05:03] https://bugs.launchpad.net/launchpad/+bug/716774 [05:03] <_mup_> Bug #716774: Timeout on MaloneApplication:+bugs < https://launchpad.net/bugs/716774 > [05:03] StevenK: Well, you could check the postgres log. [05:04] Rather than adding debug prints everywhere. [05:04] /var/log/postgresql/postgresql-8.4-main.log is fairly useless [05:04] Ah, you don't have statement logging on? [05:05] Doubtful [05:05] log_statement='all' in postgresql.conf [05:05] It's very handy. [05:10] wgrant: If that logs into the main log, I can't see it [05:11] /var/log/postgresql/postgresql-8.4-main.log [05:16] wgrant: Doesn't seem to work for me. :-( [05:17] StevenK: You put it at the bottom of /etc/postgresql/8.4/main/postgresql.conf, and it doesn't have any other log_statement lines? [05:17] Rargh, there's two! [05:18] Heh. [05:21] Now to distill the log [05:22] Looks to use the correct DB name [05:22] Both of them? [05:23] wgrant: They all use launchpad_ftest_23325 [05:23] StevenK: sorry was at guitar lesson, thanks lifeless for approving [05:23] :( [05:23] Can you see the LFA UPDATES? [05:24] Yes [05:24] and you commit ? [05:24] Yes [05:24] There is a transaction.commit() [05:24] But can you see it in the log? [05:24] Some layers might not have the transaction manager configured properly. [05:24] But it still fails on LaunchpadZopelessLayer :/ [05:24] [2011-03-09 16:19:28 EST] launchpad_main@launchpad_ftest_23325 LOG: statement: UPDATE LibraryFileAlias SET content=NULL [05:24] [2011-03-09 16:19:28 EST] launchpad_main@launchpad_ftest_23325 LOG: statement: COMMIT [05:24] That seems pretty definitive. [05:25] The next bits are all garbo_hourly, so that's the script running [05:26] Oh. [05:26] Ha ha ha. [05:26] Ha ha ha ha ha. [05:26] I bet those SPRs have *no files at all*. [05:26] Fuck sampledata, seriously. [05:27]