[06:23:48] (PS1) Liangent: Make LanguageFallbackChainFactory work with $wgDisabledVariants [extensions/Wikibase] - https://gerrit.wikimedia.org/r/76060 [06:24:48] Lydia_WMDE: how can I request an emergency review & backport? [06:25:01] liangent: hey [06:25:03] wasup? [06:25:18] https://gerrit.wikimedia.org/r/#/c/76060/ [06:25:35] zhwiki is broken :/ [06:25:40] (CR) jenkins-bot: [V: -1] Make LanguageFallbackChainFactory work with $wgDisabledVariants [extensions/Wikibase] - https://gerrit.wikimedia.org/r/76060 (owner: Liangent) [06:25:45] how is it broken? [06:27:11] Lydia_WMDE: all property fetches fail [06:27:12] {{#property [06:27:38] ok [06:27:52] i will push people to take a look asap [06:29:58] (PS2) Liangent: Make LanguageFallbackChainFactory work with $wgDisabledVariants [extensions/Wikibase] - https://gerrit.wikimedia.org/r/76060 [06:31:31] (CR) jenkins-bot: [V: -1] Make LanguageFallbackChainFactory work with $wgDisabledVariants [extensions/Wikibase] - https://gerrit.wikimedia.org/r/76060 (owner: Liangent) [06:34:16] (PS1) Liangent: WIKIBASE MASTER FAILS? [extensions/Wikibase] - https://gerrit.wikimedia.org/r/76062 [06:35:49] (CR) jenkins-bot: [V: -1] WIKIBASE MASTER FAILS? [extensions/Wikibase] - https://gerrit.wikimedia.org/r/76062 (owner: Liangent) [06:36:43] (CR) Liangent: "The jenkins failure is from somewhere else." [extensions/Wikibase] - https://gerrit.wikimedia.org/r/76060 (owner: Liangent) [06:46:13] liangent: Heh, yeah it's been failing for a while now. [07:02:13] (CR) Aude: "normally, a missing message means the message key gets displays. (ick, but does not take the site down)" [extensions/Wikibase] - https://gerrit.wikimedia.org/r/75997 (owner: Aude) [08:02:40] morning legoktm :> [08:44:51] hey DanielK_WMDE [08:45:37] wikibase currently is failing tests, and here's one way to fix it, but I'd like you to take a look: https://gerrit.wikimedia.org/r/#/c/76022/ [08:59:18] tzafrir: congrats on your change being merged and deployed! [09:11:26] (CR) Daniel Kinzler: [C: 2] Removed failing test for illegal time value [extensions/Wikibase] - https://gerrit.wikimedia.org/r/76022 (owner: Denny Vrandecic) [09:12:13] Denny_WMDE: the test is in the right place where it is, it's not testing logic from another repo. [09:12:48] (Merged) jenkins-bot: Removed failing test for illegal time value [extensions/Wikibase] - https://gerrit.wikimedia.org/r/76022 (owner: Denny Vrandecic) [09:12:58] it tests the constraints for the wikibase time type, which is indeed a wikibase thing. [09:13:18] (PS3) Tobias Gritschacher: Fix: Edit summaries have a double space [extensions/Wikibase] - https://gerrit.wikimedia.org/r/75869 [09:13:22] (PS2) Jeroen De Dauw: Fix multiple design issues in this component [DNM] [extensions/WikibaseDatabase] - https://gerrit.wikimedia.org/r/75873 [09:13:23] (CR) jenkins-bot: [V: -1] Fix: Edit summaries have a double space [extensions/Wikibase] - https://gerrit.wikimedia.org/r/75869 (owner: Tobias Gritschacher) [09:13:24] (PS3) Aude: Make LanguageFallbackChainFactory work with $wgDisabledVariants [extensions/Wikibase] - https://gerrit.wikimedia.org/r/76060 (owner: Liangent) [09:13:32] Denny_WMDE: hin hindsight though, i should habe been more defensive, and use try/catch ionstead of just a comment [09:13:35] (CR) jenkins-bot: [V: -1] Fix multiple design issues in this component [DNM] [extensions/WikibaseDatabase] - https://gerrit.wikimedia.org/r/75873 (owner: Jeroen De Dauw) [09:14:32] DanielK_WMDE: I didn't remove the comment due to the validator remark [09:14:48] does this mean we need to do something about the validator too, or can I remove the test and comment? [09:15:27] (CR) Denny Vrandecic: [C: 2] LanguageFallbackChain: Avoid one more unneeded Language construction [extensions/Wikibase] - https://gerrit.wikimedia.org/r/73413 (owner: Liangent) [09:15:31] (CR) jenkins-bot: [V: -1] LanguageFallbackChain: Avoid one more unneeded Language construction [extensions/Wikibase] - https://gerrit.wikimedia.org/r/73413 (owner: Liangent) [09:16:19] liangent: you there? [09:16:20] (PS14) Addshore: Special page for entities without description [extensions/Wikibase] - https://gerrit.wikimedia.org/r/67634 (owner: Bene) [09:17:56] (PS4) Tobias Gritschacher: Fix: Edit summaries have a double space [extensions/Wikibase] - https://gerrit.wikimedia.org/r/75869 [09:18:44] (PS2) Tobias Gritschacher: WIKIBASE MASTER FAILS? [extensions/Wikibase] - https://gerrit.wikimedia.org/r/76062 (owner: Liangent) [09:19:18] aude: there? [09:21:09] (CR) Tobias Gritschacher: "not anymore" [extensions/Wikibase] - https://gerrit.wikimedia.org/r/76062 (owner: Liangent) [09:23:51] [travis-ci] wikimedia/mediawiki-extensions-Wikibase#478 (master - cd77b2a : vrandezo): The build was broken. [09:23:52] [travis-ci] Change view : https://github.com/wikimedia/mediawiki-extensions-Wikibase/compare/ce1a5774441a...cd77b2acbe98 [09:23:52] [travis-ci] Build details : http://travis-ci.org/wikimedia/mediawiki-extensions-Wikibase/builds/9508864 [09:25:17] hi pragunbhutani [09:25:56] ah sweet travis … there we have one of these failing tests [09:25:58] let's figure it out [09:26:12] aude: I haven't been able to chat with jon yet so I was wondering if there are other things that I could work on in the meanwhile? [09:26:27] mobile skin/view html? [09:26:36] yes, I need his help with that [09:26:54] so anyone knows why https://travis-ci.org/wikimedia/mediawiki-extensions-Wikibase/jobs/9508870 fails? [09:26:57] mock up the html (with the entity data hardcoded) [09:26:58] (PS1) Tobias Gritschacher: Add information about summary's length limit to API documentation [extensions/Wikibase] - https://gerrit.wikimedia.org/r/76066 [09:27:06] I think we work on completely opposite time zones [09:27:09] so we figure out how to implement the layout, the css, etc. [09:27:12] pragunbhutani: yeah [09:27:36] hoo: aude DanielK_WMDE JeroenDeDauw Tobi_WMDE addshore ? any idea why https://travis-ci.org/wikimedia/mediawiki-extensions-Wikibase/jobs/9508870 fails? i think that's the "randomly failing test" i mentioned during daily [09:27:38] Denny_WMDE: talking about fixing failing tests: https://gerrit.wikimedia.org/r/#/c/75852/ [09:27:43] hmm, okay I'll do that :) [09:27:52] (PS2) Tobias Gritschacher: Add information about summary's length limit to API documentation [extensions/Wikibase] - https://gerrit.wikimedia.org/r/76066 [09:28:00] Denny_WMDE: That was the one I tried to hunt down some days ago... no luck [09:28:06] pragunbhutani: now, i'm not great at css and think you do better at this than i would :) [09:28:20] we'll require tests for the ClaimHtmlGenerator as well wouldn't we? [09:28:26] oh I'm comfortable with css :) [09:28:28] i think that is the hardest part [09:28:35] yes, we need tests [09:28:38] Denny_WMDE: that jenkins job is refusing to load for me [09:28:48] okay so are we going to defer that for later? [09:28:57] i made a patch that would allow injecting the template as a parameter [09:29:05] I'm trying to make a mental roadmap of the project [09:29:10] pragunbhutani: sure [09:29:13] ah, the wf-template? [09:29:13] 1) Wikibase\Test\StoreTest::testRebuild with data set #0 (Wikibase\SqlStore) [09:29:13] DBQueryError: A database error has occurred. Did you forget to run maintenance/update.php after upgrading? See: https://www.mediawiki.org/wiki/Manual:Upgrading#Run_the_update_script [09:29:13] Query: INSERT INTO `unittest_wb_items_per_site` (ips_item_id,ips_site_id,ips_site_page) VALUES ('10','enwiki','spam') [09:29:13] Function: Wikibase\SiteLinkTable::saveLinksOfItem [09:29:13] Error: 1062 Duplicate entry 'enwiki-spam' for key 'wb_ips_item_site_page' (localhost) [09:29:21] JeroenDeDauw: ^ [09:29:40] it's important to work on both, but coding the layout is more important for now [09:29:43] aude: okay so we already have the bit that's required for claim html generation [09:29:48] a mobile skin comes next [09:29:50] hoo: yes, i thought so [09:29:54] pragunbhutani: kind of yes [09:29:57] followed by tests for [09:30:05] JeroenDeDauw: is that the on that kept failing in travis the last few days? [09:30:10] ClaimHtmlGenerator/Both [09:30:20] i can see a mobile view then calling getHtmlForClaim or whatever and providing it a template as an argument [09:30:28] the template could be different than the default view [09:30:37] Denny_WMDE: yeah, that is the random one [09:30:49] then mobile skin do $mobileView->render() or something (in the end) [09:31:00] JeroenDeDauw: thanks for confirming [09:31:35] so, can we all try to hunt it down? [09:31:40] Denny_WMDE: Might be because the thing that is supposed to keep labels etc. unique is disabled on MySQL for unit tests... presumably because of http://bugs.mysql.com/bug.php?id=10327 [09:31:45] hoo: can you share your finding so far? [09:32:11] When Jenkins is put in front of Travis, only some commits go through, and an interferrence pattern appears on Travis: https://travis-ci.org/wikimedia/mediawiki-extensions-Wikibase/builds [09:32:17] We can thus conclude commits are waves [09:32:27] aude: yeah I'm not entirely sure about how it is that I'll be using the ClaimHtmlGenerator in the mobile part as yet [09:32:43] if we'll make a skin, or a template I'm not sure [09:32:55] that's what I need to ask Jon [09:32:58] pragunbhutani: what do you think? [09:33:05] hoo: hmmm [09:33:33] see, Jon did tell me that there is now a way where you can extend an existing skin to create one of your own [09:33:35] our template system does have some issues [09:33:43] so I think that extending SkinMinerva might be an option [09:33:44] it's super simple, maybe too simple [09:33:52] pragunbhutani: sure [09:34:14] Abraham_WMDE: Denny_WMDE: where is the lydia? [09:34:40] Denny_WMDE: no idea... could it be that travis is using the same mysql instance for running multiple test suits? [09:34:45] pragunbhutani: it would just need to have different layouts, for example depending on content type [09:34:47] that would be a recipe for disaster... [09:34:48] of a page [09:34:51] aude: SkinMobile already renders the entire thing (in that broken fashion) so I don't think I will be working on that [09:34:53] JeroenDeDauw: she works today for TC [09:35:00] pragunbhutani: ok [09:35:06] JeroenDeDauw: do you know ---^^ [09:35:07] aude: did you see the layout that Pau Giner suggested? [09:35:25] DanielK_WMDE: It's not, no [09:35:37] Wait, I'll add that to the rest of the mockups and then share a link [09:35:44] ok [09:35:53] DanielK_WMDE: Please let me know after reviewing the latest changes. I'll clarify this dependency problem with you so that future changes don't get messed up like these have. [09:36:07] I set myself up a similar set about a week ago (for these test failures) and these happen all the time [09:36:25] no matter whether the DB is newly created or has run the suite 5+ times [09:37:02] DanielK_WMDE: Denny_WMDE: TravisCI creates an entire new VM and then freshly sets up MW for each build job [09:39:02] JeroenDeDauw: then i have no clue how wb_ips_item_site_page could get a duplicate [09:39:22] neither do I... [09:39:38] nileshc: i just send you an email with a bunch of get/gerrit hints. [09:39:55] let me know if still have questions. [09:39:55] DanielK_WMDE: Yup, I just saw it. Reading now. [09:39:59] ok [09:40:15] DanielK_WMDE: I even grepped Wikibase for spam and disabled all tests containing the string... no luck still... but some tests in DataModel use the string either (I disabled some of them...) [09:40:28] * aude wonders why php5? [09:40:32] 5.5 [09:40:50] is it consistently php5.5 failing? and not other versions? [09:41:20] https://travis-ci.org/wikimedia/mediawiki-extensions-Wikibase/jobs/9470476 is php5.4 [09:41:35] DanielK_WMDE: By empty item you mean an item that doesn't exist in the suggester right? This is referred to in the code by Anonymous item...We have the AnonymousEntitySuggesterServlet and the ExistingEntitySuggesterServlet. [09:41:40] yea, that'S odd... the mysql version is the same, right? [09:41:41] aude: My VM has PHP 5.5 [09:42:16] nileshc: i mean an item that has no properties yet (because it was just now created) [09:42:16] DanielK_WMDE: assume so [09:42:38] nileshc: ...so there's nothing to base suggestions on [09:42:42] DanielK_WMDE: Ah. Yes Myrrix handles those cases itself. [09:43:09] aude: http://www.mediawiki.org/wiki/User:Pragunbhutani/GSoC_2013_Implementation_Approaches#UI_Mockup_v3 [09:43:16] pragunbhutani: thanks [09:43:31] nileshc: so much the better :) [09:43:32] aude: sure thing :) [09:43:34] DanielK_WMDE: I think it may give random suggestions...I'll test that out and add/refactor accordingly. [09:44:25] pragunbhutani: i think it's nice though might not have all the editing enabled right away [09:44:32] (unlikely) [09:44:37] I know [09:44:39] aude: no, as far as i understand it is not consistently php 5.5 [09:44:42] sometimes it is [09:44:46] sometimes it is 5.4 [09:44:49] right JeroenDeDauw ? [09:44:49] Denny_WMDE: i see [09:44:53] so we have two alternate UIs for now [09:44:57] one with actions and one without [09:45:02] I think it's nice this way [09:45:18] since we can simply keep adding stuff to the 'with actions' UI as they come along [09:45:18] My VM on which it fails consistently has PHP 5.5.0 and mysql Ver 15.1 Distrib 5.5.31-MariaDB, for Linux (x86_64) using readline 5.1 [09:45:23] Denny_WMDE: ^ [09:45:49] so for you it is always 5.5 hmm… ok… let's see if JeroenDeDauw remembers about travis if this is the case too [09:45:53] pragunbhutani: so when i click on one of the statements, then what does it look like? [09:46:23] and it would have the menu / search box at the top, right? [09:46:28] aude: he didn't say anything about that but the way I envision it, I think it should slide and expand [09:46:46] yes, definitely [09:46:51] aude, hoo: looking at this one, it is 5.3 https://travis-ci.org/wikimedia/mediawiki-extensions-Wikibase/builds/9388528 [09:46:54] same test [09:46:56] this is just the 'content area' if we may call it that [09:47:00] 5.5 passed that run [09:47:00] pragunbhutani: ok [09:47:06] 5.3 did not [09:47:12] weird [09:47:23] interesting [09:47:23] for me that test failed in all cases [09:47:58] for text, it could just expand horizontally [09:48:06] pragunbhutani: so, i think we should include the new mockups in our weekly status update [09:48:13] and for images etc, both horizontally and vertically [09:48:14] and maybe another mail? [09:48:19] aude: i'm fine with merging liangent's variant fix, just gave it a +1. i don't feel confident giving it a +2 without anyone else looking at it though. [09:48:28] DanielK_WMDE: i was looking at it [09:48:34] ok [09:48:42] aude: sure, I'm going to add clarifications to the doubts you had as well [09:48:47] pragunbhutani: ok [09:48:56] hoo: "that" test? failed how? [09:49:28] DanielK_WMDE: Well, it failed because of the database error (unique key) [09:49:32] how do i find this test? [09:49:49] it's one of the storage tests AFAIR [09:49:49] (stupid question, sorry) [09:49:56] (Abandoned) Jeroen De Dauw: WIKIBASE MASTER FAILS? [extensions/Wikibase] - https://gerrit.wikimedia.org/r/76062 (owner: Liangent) [09:50:06] repo/tests/phpunit/includes/store/StoreTest.php:55 [09:50:08] Denny_WMDE: ^ [09:50:24] the constraint thing is happening in different tests [09:50:24] Denny_WMDE: you hit squiggely-o and type "StoreTest" [09:50:33] label and description uniqueness in sqlite [09:50:39] hoo: thanks [09:50:41] that test runs in sqlite only [09:50:45] hoo: Denny_WMDE: the weird thing is, I cannot find where the test data even comes from [09:50:53] enwiki and spam [09:50:55] but not a problem in 5.3 apparently (on jenkins) [09:50:55] Denny_WMDE: it is not just 5.5. Also 5.4 [09:50:56] Tobi_WMDE: Me to... [09:51:07] aude: but the failures are all in mysql as far as i have seen [09:51:12] Denny_WMDE: no [09:51:15] Tobi_WMDE: the api set aliases test uses this combination, but only for labels [09:51:24] there is something in changeHandlerTest but I think we skip this test [09:51:26] JeroenDeDauw: yeah, thanks for confirming. even 5.3 at times [09:51:27] aude: no? [09:51:36] If you want to investigate, take a look at WikibaseDataModel... [09:51:38] aude: no? [09:51:42] Tobi_WMDE: ^ [09:51:47] JeroenDeDauw: is an array of EntityIds sufficent for badges? [09:51:53] we were looking at it last friday [09:51:56] me an hoo [09:52:00] Denny_WMDE: also, sometimes for client tests and sometimes for repo tests [09:52:03] There are some tests still with @group Wikibase and these also use "spam" as test string [09:52:08] Tobi_WMDE, JeroenDeDauw: i fixed several things there, we should no longer skip changehandlertest. if there are still issues, i want to know about them. [09:52:08] the are supposed to be pointing to items now [09:52:10] aude: but on https://travis-ci.org/wikimedia/mediawiki-extensions-Wikibase/builds/9388528 we see it is mysql [09:52:23] one sec... [09:52:31] aude: all failures i have seen were in mysql [09:52:33] Tobi_WMDE: awesome. where do you see that from? [09:52:47] DanielK_WMDE: AFAIR these are running fine on my (travis like) env., so it should be ok to run them on travis as well [09:52:50] https://travis-ci.org/wikimedia/mediawiki-extensions-Wikibase/jobs/9241404 [09:53:16] i think we removed the automatic loading of tests inr epo [09:53:23] aude: That was an older failure I've already resolved [09:53:25] and that one is not occurring now, right? [09:53:31] hoo: ok, by resetting the entities [09:53:35] aude: I've put that back in place [09:53:35] in some place [09:53:41] hoo: ok [09:53:51] Denny_WMDE: e.g. https://travis-ci.org/wikimedia/mediawiki-extensions-Wikibase/builds/9367374 [09:53:52] I think it needed 4 different fixes until everything passed... [09:53:53] aude: no, it seems to be a different test and error now [09:54:04] ok, though similar [09:54:09] Tobi_WMDE: yaay! [09:54:11] :Ü [09:54:11] aude: 7 days ago. i *think* i fixed that last week. [09:54:12] sounds similar [09:54:21] DanielK_WMDE: ok :) [09:54:37] just sounds like same symptoms [09:54:41] something not resetting [09:54:46] lazowik: I geuss so. I was initally expected this array to contain strings, so this is now endeding up being more complex. No objections though [09:54:54] not cleaning up [09:55:06] DanielK_WMDE: I see that I confused the "empty item" case with something else. Since this is an anonymous item, without *any* data, the recommendation engine won't be able to do anything. So we need to suggest top N most popular properties for that right? [09:55:12] aude: not really... [09:55:21] yeah, so, it all seems to happen randomly, on any php version, on both client and repo, and only in mysql, not in sqlite as far as we can tell [09:55:22] nileshc: yes. [09:55:36] DanielK_WMDE: not convinced [09:55:40] but at least it is always the same failure [09:56:19] Denny_WMDE: should I prevent from duplicate badges? [09:56:27] aude: the thing is that we explicitly check the uniqueness constraint before trying to save. that check passes. then we get a failure on the unique key. that's *very* strange. [09:56:30] lazowik: yes [09:56:35] error on add? [09:56:43] lazowik: no [09:56:45] DanielK_WMDE: ok [09:56:50] just a no-op [09:56:51] maybe tests run parallel [09:56:57] api? [09:56:59] lazowik: just a no-op [09:57:00] and something is interfereing? [09:57:06] aude: that was my suspicion, but jeroen sais no... [09:57:13] lazowik: or rather, be consistent to what the labels do [09:57:14] i don't mean across builds [09:57:23] hm... [09:57:24] lazowik: i mean aliases [09:57:24] ok [09:57:28] but within a single build, maybe somehow it runs parallel processes? [09:57:32] sure :) as always [09:57:37] no idea [09:57:38] lazowik: :) [09:57:46] can anyone reproduce? [09:57:57] aude: Phpunit isn't parallel [09:57:59] I can... [09:58:10] hoo: i know and it shouldn't be [09:58:18] but who knows what travis does [09:58:29] I still suspect the missing uniqueness checks... [09:58:35] google turns up paratest [09:58:37] ok [09:58:46] but i doubt that travis uses that [09:58:47] DanielK_WMDE: It doesn't [09:58:51] gree [09:58:53] agree [09:58:57] just an idea :) [09:59:07] I use Fedora's PHPunit and get the same failure... even all the time [09:59:14] hoo: ok [09:59:25] DanielK_WMDE: Ah. So here are the 3 cases that I need to do: "suggest properties for an empty item", "suggest qualifiers for a given property", "suggest properties to use in source references"...the first 2 need the same top N properties. The last one will be served by the usual property suggester. A special case of the last one where it's a "new" or "empty" source ref, we'll again suggest top N, similar to the "empty item" case. [09:59:31] [tests@localhost ~]$ phpunit --version [09:59:32] PHPUnit 3.7.22 by Sebastian Bergmann. [10:00:32] nileshc: the java based REST service for suggesting a property takes a list of properties as its input, right? [10:00:54] nileshc: and now you added an alternative service that takes an item ID, and automatically figures out the properties that are already there, and uses that? [10:01:20] aude: I'm adding this to the weeky summary [10:01:26] pragunbhutani: great [10:01:27] does this go under discussion or development? [10:01:35] hoo: thats at least something if it's not completely random for you [10:01:38] i'd say discussion [10:01:42] okay [10:01:50] lydia or addshore can move it if they like [10:02:03] indeed :) [10:02:17] DanielK_WMDE: Yes. The latter is exactly similar to the "A bought X so B might also buy X" style recommendation, where the data is already present. [10:02:34] ok, so let's see [10:02:42] this things calls store->rebuild [10:02:44] and that's all [10:02:45] right? [10:03:06] and it does not return from the rebuild, but throwas a DBQueryError [10:03:39] hmmmm [10:03:55] nileshc: but that will onyl work if the data in the suggester is kept up to date in real time. which is pretty much out of the question. i'm writing a longish comment con the change now. [10:05:28] DanielK_WMDE: The data in the suggester *could* be kept up to date in real time if every time someone adds/removes/updates an entity, a function in the suggester API is called. [10:06:15] DanielK_WMDE: I mentioned about this in an email. That'd mean that the suggester won't lag behind for 1.5-2 weeks until the next wiki dump. Thoughts? [10:06:48] nileshc: possible, in theory, but too costly/brittle for now. performing an edit must not rely on external services to return in time. So the update needs to be asynchronous. We'd a) need a mechanism for that and b) have to hope that the asynchronous update would happen fast enough [10:07:07] i frankyl don't see the benefit. we can easily implement this functionality in the API on the php side of things [10:08:02] are failing tests fixed? :D [10:08:06] or not ? :< [10:08:12] DanielK_WMDE: Implement which functionality in the PHP API? (By suggester API I meant the PHP API itself that'll call the web service to add/update the entity) [10:08:24] addshore: the timevalue thing is fixed [10:08:25] addshore: not the randomly failing ones [10:08:30] :/ [10:08:32] addshore: the one on jenkins is [10:08:45] addshore: so jenkins should be happy again. travis isn't, there's still thy mysterious key constraint failure [10:08:46] addshore: jenkins runs sqlite, right? [10:08:53] yes [10:09:03] Denny_WMDE: but i run all my local tests with mysql. never seen this issue# [10:09:07] that's why we probably don't see the random one popping up there [10:09:12] weird [10:09:16] and i'm on php 5.4.9 [10:09:17] (CR) Addshore: [C: -1] "The pages themselves look good but I cant see that you have addressed aude's comments. (see below)" [extensions/Wikibase] - https://gerrit.wikimedia.org/r/67634 (owner: Bene) [10:10:00] DanielK_WMDE: BTW, on the Myrrix side every web service is asynchronous. [10:10:00] (CR) Addshore: [C: 2] Normalize item ids before trying to remove doubles [extensions/Wikibase] - https://gerrit.wikimedia.org/r/75533 (owner: Hoo man) [10:10:30] Denny_WMDE: I'm on php 5.4.7 with mysql and I never saw this issue locally [10:10:33] so how does the store get filled before the rebuild test is being called? [10:10:42] :-/ [10:11:07] JeroenDeDauw: aude: did you ever get the errors locally? [10:11:14] Tobi_WMDE: Which test group are you running? [10:11:20] so far it seems only hoo and travis who can get them [10:11:22] (Merged) jenkins-bot: Normalize item ids before trying to remove doubles [extensions/Wikibase] - https://gerrit.wikimedia.org/r/75533 (owner: Hoo man) [10:11:22] nileshc: still, callign out to any external service, and waiting for it to return, should not be done within a web request if it can be avoided. [10:11:25] Denny_WMDE: no [10:11:26] hoo: wikibase [10:11:33] which exact test is it? [10:11:43] DanielK_WMDE: In any case updating/adding something through the web service will be instantaneous (since it's async) and the actual update will be be in miliseconds too. [10:11:48] DanielK_WMDE: Yes, that is true... [10:11:49] nileshc: live updates are nice, and we can add them, but we should not rely on them at the moment. we can implement the required functionality easily on the php side [10:12:14] addshore: Wikibase/Test/StoreTest::rebuldTest [10:12:18] rebuild [10:12:28] addshore: https://travis-ci.org/wikimedia/mediawiki-extensions-Wikibase/jobs/9508870 [10:12:29] Anyone want to finish the review? :) https://gerrit.wikimedia.org/r/#/c/74127/ [10:12:34] nileshc: http is never async. it'S a synchronous protocol per definition. the response may be pretty quick because the update is just queued and not processed right away, but still: [10:12:38] DanielK_WMDE: "we can implement the required functionality easily on the php side" - live updates to the Myrrix model functionality? in pure PHP? I don't get you... [10:12:54] nileshc: what if the servlet engine is down? you'd be freezing *all* edits witing for a tcp timeout [10:13:21] nileshc: live updates are not required functionality. [10:14:10] DanielK_WMDE: Yes...so shall I keep the existing-item property recommender servlet or remove it for now? [10:14:59] nileshc: remove it. what i mean is: we want to say "suggest properties to add to item q123". the REST service implements "suggest more properties given these properties". So, we implement "suggest properties to add to item q123" by loading the properties already in q123, and then passing that list to the REST service [10:15:29] that way, the myrrix data can be a week old, and it would still work well enough [10:15:37] Denny_WMDE: here [10:17:39] DanielK_WMDE: Oh! I see, that's cool. But this has two problems: (i) the data based on which that Myrrix is recommending is old - so technically it'll give some results, but they may not necessarily be true. (ii) It's a tad bit slower because it'll involve a DB call to fetch the properties (but this might not really count) [10:18:07] DanielK_WMDE: We can also easily have something like an incremental delta file that is added to or subtracted over 24 hours and a cron job is set up to update Myrrix once a day based on that data? [10:18:26] DanielK_WMDE: Just ideas. [10:18:53] liangent: can you put me and a few people as reviewers on the changesets that you want to get reviewed? [10:19:14] liangent: I am a bit confused about the state [10:19:47] Denny_WMDE: what state? [10:21:04] (CR) Addshore: [C: 2] Make ReferencedEntityFinder use Entity::getAllSnaks [extensions/Wikibase] - https://gerrit.wikimedia.org/r/72962 (owner: Daniel Kinzler) [10:21:05] nileshc: good idea, but let's get the baseline working first :) [10:21:47] liangent: i'm looking at the patch you want backported [10:22:01] (also looking at failing random tests, unrelated) [10:22:12] it's true that it's an extra db query. but it's to a database we are already connected to. and we will probably be able to optimize this to not retrieve the entire entity, but just the list of properties. [10:22:26] (Merged) jenkins-bot: Make ReferencedEntityFinder use Entity::getAllSnaks [extensions/Wikibase] - https://gerrit.wikimedia.org/r/72962 (owner: Daniel Kinzler) [10:22:31] actually, we already can, based on the pagelinsk table. though that's a bit hackish. [10:22:47] [travis-ci] wikimedia/mediawiki-extensions-Wikibase#479 (master - e72a331 : jenkins-bot): The build is still failing. [10:22:47] [travis-ci] Change view : https://github.com/wikimedia/mediawiki-extensions-Wikibase/compare/cd77b2acbe98...e72a331c4f11 [10:22:47] [travis-ci] Build details : http://travis-ci.org/wikimedia/mediawiki-extensions-Wikibase/builds/9510594 [10:25:19] DanielK_WMDE: Oh, I see...alright. So as of now to get the properties bit "complete", I need to do those 3 things you mentioned in your email, and I need to remove the existing entity suggester. [10:25:54] DanielK_WMDE: About suggesting the top N items from the DB, wouldn't it be better to do it in the PHP API? It's database, soo... [10:26:42] ok, i think i figured it out [10:26:47] with the failing error [10:26:57] i will write this after i come back, need to run to the office [10:27:05] DanielK_WMDE: I will write the simple map-reduce scripts for sorting the properties by their popularity, but about doing the select...order by.. ? [10:27:08] and then fix it for now [10:27:15] yay! [10:27:34] so hoo, aude, DanielK_WMDE, Tobi_WMDE, addshore, JeroenDeDauw - thanks for your help, no need to continue on this for now i guess. [10:27:39] be back in an hour or so [10:27:39] ok! [10:28:08] ok [10:28:14] (CR) Liangent: "Disabled variants should behave as if they don't exist as supported variants, thus not listed in getVariants() and hasVariant(). However (" [extensions/Wikibase] - https://gerrit.wikimedia.org/r/76060 (owner: Liangent) [10:28:28] aude: ^ [10:28:33] liangent: ok [10:31:27] [travis-ci] wikimedia/mediawiki-extensions-Wikibase#480 (master - acf3c92 : jenkins-bot): The build was fixed. [10:31:27] [travis-ci] Change view : https://github.com/wikimedia/mediawiki-extensions-Wikibase/compare/e72a331c4f11...acf3c929b6a0 [10:31:27] [travis-ci] Build details : http://travis-ci.org/wikimedia/mediawiki-extensions-Wikibase/builds/9510948 [10:32:54] (CR) Addshore: [C: 2] Add information about summary's length limit to API documentation [extensions/Wikibase] - https://gerrit.wikimedia.org/r/76066 (owner: Tobias Gritschacher) [10:34:29] (Merged) jenkins-bot: Add information about summary's length limit to API documentation [extensions/Wikibase] - https://gerrit.wikimedia.org/r/76066 (owner: Tobias Gritschacher) [10:34:34] nileshc: hm... since calculating the top N data sets is simmilar to updating the suggester, it's probably best to keep that in one place. it'S a simple map of key -> order list of value/score pairs. I don't care much where it's stored [10:35:28] nileshc: mediawiki alreads has an (optional) dependency on a fast file based k/v store (let me dig up the name) [10:35:56] DanielK_WMDE: (i) MySQL, (ii) Tomcat in-memory, much like the list of properties that is already stored in a modified Fast HashMap [10:36:11] (PS1) Jeroen De Dauw: Match changes in WikibaseDatabase component [extensions/WikibaseQueryEngine] - https://gerrit.wikimedia.org/r/76071 [10:36:22] ok..I'll be back in a few..too hungry. :P [10:36:22] (CR) jenkins-bot: [V: -1] Match changes in WikibaseDatabase component [extensions/WikibaseQueryEngine] - https://gerrit.wikimedia.org/r/76071 (owner: Jeroen De Dauw) [10:37:59] (CR) Addshore: [C: 2] Fix: Edit summaries have a double space [extensions/Wikibase] - https://gerrit.wikimedia.org/r/75869 (owner: Tobias Gritschacher) [10:38:24] (PS3) Jeroen De Dauw: Fix multiple design issues in this component [DNM] [extensions/WikibaseDatabase] - https://gerrit.wikimedia.org/r/75873 [10:38:57] (PS4) Jeroen De Dauw: Fix multiple design issues in this component [DNM] [extensions/WikibaseDatabase] - https://gerrit.wikimedia.org/r/75873 [10:39:14] FUCK! [10:39:15]