[00:46:02] <tgr>	 legoktm: where does @default for seccomp come from?
[00:46:12] <tgr>	 I get a warning about it not being recognized
[00:55:14] <tgr>	 the manpage suggests using the defaults is just --seccomp and --seccomp=foo,bar automatically adds the defaults
[01:09:10] <legoktm>	 tgr: https://firejail.wordpress.com/features-3/man-firejail/ lists @default, what firejail version are you using?
[01:09:15] <legoktm>	 (it works for me)
[01:09:30] <tgr>	 0.9.44.6
[01:09:44] <tgr>	 the one apt installs on vagrant
[01:10:03] <legoktm>	 I'm on 0.9.52
[01:11:44] <legoktm>	 https://github.com/netblue30/firejail/commit/34ee8e03f58c4c51c3aa29f553e06570d0654db2 I think
[01:12:20] <legoktm>	 so 0.9.50
[01:12:33] <legoktm>	 stretch-backports has that, but are you still on jessie?
[01:12:41] <legoktm>	 you could also build it yourself, it has no dependencies
[01:14:38] <legoktm>	 hmm but we're using --seccomp=@default in production, and that has 0.9.44
[01:15:05] <tgr>	 I don't think it breaks anything, it just gets ignored
[01:15:26] <legoktm>	 that's not really great either...
[01:15:54] <legoktm>	 I'll ask moritz to backport 0.9.50+
[01:16:02] <tgr>	 the manpage suggests the defaults always get applied on .44 anyway, unless you use seccomp.keep or something like that
[01:16:23] <tgr>	 I'm trying to find an easy way to test seccomp, but not much luck so far
[01:16:57] <legoktm>	 lilypond uses the execve syscall, I tested with that
[01:17:15] <tgr>	 yeah, execve is easy to test
[01:21:18] <legoktm>	 I gotta go afk now, I'll make sure to follow up on the @default usage
[01:21:38] <legoktm>	 thanks for taking a look at the patches
[01:29:12] <tgr>	 (I'm using vagrant, which is on jessie, yes. Not for long though.)
[02:20:29] <legoktm>	 tgr: thanks :)
[12:32:36] <RoanKattouw>	 DanielK_WMDE_: addshore: Yay re MCR patch being merged. I reviewed part of it yesterday but didn't get around to submitting my comments. I just submitted them and they're mostly about incorrect doc comments
[12:32:49] <addshore>	 RoanKattouw: awesome!
[12:33:00] <addshore>	 I can file some followups today
[12:33:07] <addshore>	 I think Daniel is mostly off not until the new year
[12:33:23] <RoanKattouw>	 OK cool
[12:33:37] <RoanKattouw>	 Also did you want to do some lag experiments in beta?
[12:34:12] <addshore>	 so I dont think it is really needed, the 250 -> 500ms lag there always is between the master and slave seem enough to reproduce things needed
[12:34:21] <addshore>	 I also setup a master and replica on my local setup now :)
[12:35:12] <addshore>	 there were a couple of issues with the patch I fixed last night after testing on beta, the one remaining issue is fixed with https://gerrit.wikimedia.org/r/#/c/399795/ and https://gerrit.wikimedia.org/r/#/c/399764
[12:36:15] <RoanKattouw>	 OK cool
[12:38:06] <addshore>	 Once that lands on beta I'll do a bunch more testing on all the sites
[12:39:20] <RoanKattouw>	 I just +2ed both
[12:51:30] <addshore>	 whee
[12:51:59] <addshore>	 and scap has just finished running! :D
[12:52:07] <addshore>	 *goes to test it again*
[12:54:52] <addshore>	 Yup, that issue is resolved, woo!
[12:54:57] <addshore>	 this patch might stick this time :D
[13:00:42] <addshore>	 meh, found another bug!
[13:01:22] <addshore>	 very similar
[13:04:59] <addshore>	 https://phabricator.wikimedia.org/T183548 
[14:28:46] <addshore>	 RoanKattouw: https://gerrit.wikimedia.org/r/#/c/399812/ is similar to the other one you just merged
[15:34:31] <halfak>	 o/ DanielK_WMDE_ 
[15:34:39] <halfak>	 Could I ask a quick Q about multi-content revisions?
[15:35:07] <halfak>	 I'm thinking about creating a page that has an arbitrary number of "contents" with their own independent histories. 
[15:35:28] <halfak>	 Does that fit?  Or is there a fixed number of "contents" per page?
[15:50:21] <RoanKattouw>	 halfak: I think DanielK_WMDE_ might be off for the holidays already but addshore might still be around
[15:50:42] <halfak>	 Thanks RoanKattouw :) 
[15:51:14] <halfak>	 RoanKattouw, some time after the holidays, I'd really love to have you walk me through flow's 0.5 implementation of suppression. 
[15:51:22] <halfak>	 I want to learn from your experience. :D 
[15:52:26] <RoanKattouw>	 It's a 0.5 implementation of versioning in general and it's a can of worms that's not great to open
[15:52:35] <RoanKattouw>	 And yeah I'd be happy to talk to you about it some time
[15:53:19] <RoanKattouw>	 Good (short) reading in the meantime if you haven't already seen it: https://www.mediawiki.org/wiki/Everything_is_a_wiki_page
[15:53:24] <halfak>	 great!  Yeah, I just want to examine the can.  We can leave it closed :D 
[16:19:59] <addshore>	 halfak: that doesnt sound like it fiyts
[16:20:01] <addshore>	 *fits
[16:20:35] <addshore>	 For MCR there will be a single history / collection of revisions for ALL slots of a page
[16:20:47] <addshore>	 but each revision has multiple slots
[16:20:59] <halfak>	 Sure.  Each slot has it's own history though, right?
[16:21:14] <addshore>	 also arbitrary number of "contents" also doesn't rally sound like it fits well
[16:21:58] <halfak>	 Sounds like "slots" is the right term
[16:22:12] <addshore>	 halfak: I mean, yes, kind of, but when your looking through mediawiki, you have the page history, and in there each slot of a revision can remain the same between different page revisions
[16:22:15] <halfak>	 Arbitrary as in, some pages will have 5.  Others will have 2.  
[16:22:29] <addshore>	 halfak: so the slots of designed to have names
[16:22:40] <halfak>	 Yeah.  they
[16:22:42] <halfak>	 'd be named
[16:23:32] <halfak>	 E.g. revision/1234 would have slots "article quality", "draft quality", or "topic"
[16:23:36] <addshore>	 I mean, anything is possible right ;)
[16:23:47] <halfak>	 addshore, just trying to figure out the data model of MCR
[16:23:58] <addshore>	 halfak: yes, that seems possible, and probably in the list of use cases it's designed for
[16:24:01] <halfak>	 I've get to see a diagram of the shape. 
[16:24:04] <halfak>	 *yet
[16:24:12] <halfak>	 addshore, interesting :) 
[16:24:14] <addshore>	 *goes to search for a diagram*
[16:24:18] <halfak>	 Thanks :) 
[16:25:04] <addshore>	 mhhm, cant find a diagram right now!
[16:26:16] <halfak>	 Ok no worries.  I can dig later.  Let me not totally side-track you :) 
[16:26:20] <halfak>	 Thanks for your help :D 
[16:31:39] <halfak>	 relevant to what I'm thinking: https://www.mediawiki.org/wiki/JADE/MCR_example/Revision/1234
[16:32:00] <halfak>	 I've used sub-page transclusions to mimic multi-content revisions. 
[16:32:28] <anomie>	 halfak: The high-level idea of MCR is that things you would now have separate pages for won't have to have separate titles in MCR. And the same for stuff that's weirdly embedded in wikitext to avoid having a separate page. For example, a template and its /doc subpage and its TemplateData, or an article and its metadata like which categories it's in.
[16:33:18] <halfak>	 anomie, in my case, there would be no main page that has peripheral content.  Does that make sense in MCR?
[16:33:49] <anomie>	 halfak: But it probably doesn't make a whole lot of sense to separate one small structured data Content blob across several slots, particularly when the set of slots isn't well-defined.
[16:34:39] <halfak>	 anomie, slots isn't well-defined?
[16:34:55] <anomie>	 By "isn't well defined", you have wp10 and draft_quality on there now, but I've seen you add and remove similar things from ORES as needs change.
[16:35:05] <halfak>	 anomie, never removed
[16:35:17] <halfak>	 Added as needs arise.  Then they continue forever
[16:35:33] <anomie>	 Ok, s/remove/stop using/
[16:35:41] <halfak>	 never stop using
[16:35:47] <halfak>	 :/
[16:35:52] <halfak>	 Not sure what you've been seeing
[16:36:38] <anomie>	 Maybe I'm mistaken then. But you could, if you decide one of your topics isn't useful anymore.
[16:36:54] <anomie>	 (I'm sure "topics" is the wrong word)
[16:37:09] <halfak>	 So the different sub-pages in this case implement different schemas.  they're JSON.  So I'd want to validate them with different JSON schemas
[16:37:46] <halfak>	 Each "namespace" (in this case "Revision") would have a set of slots coupled with JSON schemas.
[16:38:16] <anomie>	 In your example they're both just trivial strings.
[16:38:55] <halfak>	 anomie, they aren't just trivial strings.  The third is an array of strings.  They will be based on a limited set of potential values. 
[16:39:09] <bd808>	 anomie: do you want to meet today?
[16:39:09] <halfak>	 Also, i'm not sure why that is relevant. 
[16:39:33] <halfak>	 FWIW, the JSON schemas *will* be relatively simple. 
[16:39:39] <halfak>	 In most cases.  
[16:41:01] <anomie>	 bd808: Ugh, my reminder didn't go off. If it's ok with you, let's skip it this week.
[16:41:18] <bd808>	 anomie: okey doke. Have a good break!
[16:45:23] <halfak>	 Another example: https://www.mediawiki.org/wiki/JADE/MCR_example/Edit/1234
[16:46:10] <addshore>	 RoanKattouw: Another one https://gerrit.wikimedia.org/r/399838
[17:47:34] <RoanKattouw>	 addshore: Ahm... another one? https://en.wikipedia.beta.wmflabs.org/wiki/Sandbox
[17:47:43] <RoanKattouw>	 [Wj1FJApEE4AAAEx7hZEAAAAF] /wiki/Sandbox Wikibase\DataModel\Services\Lookup\EntityLookupException from line 44 of /srv/mediawiki/php-master/extensions/Wikibase/lib/includes/Store/RevisionBasedEntityLookup.php: Failed to unserialize the content object.
[17:47:53] <addshore>	 interesting, I havn't seen that one before
[17:48:06] <addshore>	 at a guess, because it is the sandbox, it has some very old revisions in
[17:50:09] <addshore>	 hmm, wait, that has Wikibase in the stacktrace
[17:50:23] <RoanKattouw>	 Yes, also the exception is a Wikibase exception
[17:53:03] <addshore>	 RoanKattouw: shall I file a task or do you want to? :)
[17:54:25] <addshore>	 RoanKattouw: right now I think the best course of action is leave it in master and revert the changes in the branch before the train at the start of jan
[17:54:27] <addshore>	 no_justification: ^^
[17:54:45] <no_justification>	 Agreed
[17:54:48] <addshore>	 IMO having this time on beta is very valuable
[17:55:16] <addshore>	 It would be nice to be able to wrap it all in a switch, but due to how tangled it all is that s a bit tricky.
[17:55:50] <addshore>	 I'm around all over the christmas period and am going to keep jabbing violently at beta, and will be around to do the revert for the first train too
[18:03:56] <addshore>	 I'll keep a list of patches to revert (currently standing at 8) and applying cleanly
[18:06:42] <addshore>	 Infact, I'll keep that @ https://wikitech.wikimedia.org/wiki/User:Addshore/MCR_Revert :)
[18:23:42] <RoanKattouw>	 > mw.config.get( 'wgContentLanguage' )
[18:23:44] <RoanKattouw>	 'deployment'
[18:23:49] <RoanKattouw>	 Dear beta labs, WTF
[18:23:53] <RoanKattouw>	 Sincerely, Roan
[18:24:27] <RoanKattouw>	 addshore: You file the task, I'm in a task-filing stack three levels deep in VisualEditor right now
[18:27:06] <bd808>	 RoanKattouw: what is "beta labs"? ;)
[18:27:23] <RoanKattouw>	 betatoolclusterforgelabs
[18:27:48] <subbu>	 RoanKattouw, s/labs/cluster/
[18:27:55] <bd808>	 oh, you mean the deployment-prep Cloud VPS project :)
[18:28:27] * bd808 is snarky today
[18:28:44] * subbu maybe shouldn't have made the drive by comment after just paging into the channel .. but was trying to be helpful :)
[18:28:46] <addshore>	 RoanKattouw: will do
[18:40:09] <legoktm>	 addshore: is feature flag not an option?
[18:40:49] <addshore>	 legoktm: I mean, in theory yes, it could be rather messy and tangle 
[18:40:51] <addshore>	 *tangled
[18:41:28] <legoktm>	 If its temporary I think its okay... can't be messier than $wgDisableAuthManager
[18:41:45] <addshore>	 could load alternate Revision classes based on the flag, and then wrap all the other logic changes in the flag
[18:41:52] <addshore>	 I'm happy to try and write up a patch for it :)
[18:42:08] <addshore>	 Just fixing up https://gerrit.wikimedia.org/r/#/c/399856/ now
[18:44:12] <RoanKattouw>	 subbu: While you're here
[18:44:17] <RoanKattouw>	 I found a fun bug
[18:44:25] <RoanKattouw>	 $ echo '<p><span typeof="mw:Transclusion" data-mw="{&quot;parts&quot;:[{&quot;template&quot;:{&quot;target&quot;:{&quot;href&quot;:&quot;Template:Echo&quot;,&quot;wt&quot;:&quot;Echo&quot;},&quot;params&quot;:{&quot;1&quot;:{&quot;wt&quot;:&quot;<table class=\&quot;foo\&quot;>&quot;}}}}]}"></span></p>' | bin/parse.js --html2wt --prefix mediawikiwiki
[18:44:26] <RoanKattouw>	 {{Echo|<table class="foo">}}
[18:44:34] <RoanKattouw>	 Note the missing |1=
[18:44:51] <legoktm>	 addshore: are you planning to be working next week?
[18:44:56] <addshore>	 yup
[18:45:24] <RoanKattouw>	 If you remove the <brackets>, the 1= comes back
[18:48:12] <addshore>	 RoanKattouw: finally filed it as https://phabricator.wikimedia.org/T183583
[18:48:50] <addshore>	 right, feature flag time :D
[18:48:57] <addshore>	 DanielK_WMDE_: ^^ FYI
[18:49:17] <RoanKattouw>	 subbu: (Filed as T183584)
[18:49:18] <stashbot>	 T183584: Parsoid incorrectly serializes (html2wt) templates with named parameters containing equals signs (=) and angle brackets (<>) - https://phabricator.wikimedia.org/T183584
[18:49:50] <subbu>	 RoanKattouw, thanks for the bug report .. we'll take a look.
[18:51:16] <RoanKattouw>	 Thanks. It probably doesn't happen very often.
[18:51:20] <RoanKattouw>	 The bug, I mean
[18:52:27] <addshore>	 legoktm: default with MCR off i guess? worth removing the REL notes or leave them there for now?
[18:52:46] <legoktm>	 Default with on
[18:55:47] <subbu>	 RoanKattouw, right, if it had, we would have heard about it by now :) we have code to detect the presence of '=' in tpl args and use the N=arg format .. but i think that code probably over-optimistically assumes that html-tags will be parsed atomically in transclusions (although that is probably a reasonable / better way to parse transclusions)
[18:56:02] <anomie>	 We've replaced "labs labs labs" with "ha ha you forgot the new name" :/
[18:56:34] <addshore>	 legoktm: okay! :)
[18:57:56] <RoanKattouw>	 addshore: So, change of plan, you're not going to prepare a series of reverts for wmf.15 but instead you're introducing a feature flag?
[18:58:05] <RoanKattouw>	 Do I understand that correctly?
[18:58:43] <addshore>	 well, i have the reverts, but I'm going to see if I can do the feature flag this evening
[18:59:15] <addshore>	 If I can and it isn't horrible, then we can run with that :)
[19:01:27] <addshore>	 I just emailed wikitech-L with an update
[19:01:32] <addshore>	 now going to have pudding
[19:02:44] <anomie>	 addshore: FYI, you broke a bunch of tests when $wgCommentTableSchemaMigrationStage is not MIGRATION_OLD again. You can probably look for where I had to add that in https://gerrit.wikimedia.org/r/#/c/380669/.
[19:03:20] <legoktm>	 anomie: can we have jenkins run the tests with all three migration modes?
[19:03:42] <addshore>	 legoktm: that would be something similar to what we now have for the contentmodel switch
[19:04:12] <anomie>	 legoktm: Four: MIGRATION_OLD, MIGRATION_WRITE_BOTH, MIGRATION_WRITE_NEW, and MIGRATION_NEW. I don't know how we'd go about that, but I'm sure it's theoretically possible.
[19:04:22] <addshore>	 anomie: it should just be a case of forcing the tests to run with a specific global using setMwGlobals
[19:05:17] <addshore>	 legoktm: anomie for contentmodel in revision we made https://github.com/wikimedia/mediawiki/blob/master/tests/phpunit/includes/RevisionDbTestBase.php and https://github.com/wikimedia/mediawiki/blob/master/tests/phpunit/includes/RevisionNoContentHandlerDbTest.php for example
[19:05:25] <anomie>	 Let me clarify: "I don't know how we'd go about that without a bunch of crazy hacks in MW's unit testing"
[19:05:38] <legoktm>	 we can do what we do in PreprocessorTest, let me find it
[19:06:53] <legoktm>	 anomie: https://phabricator.wikimedia.org/source/mediawiki/browse/master/tests/phpunit/includes/parser/PreprocessorTest.php;63c30080f550da8ddaad1b9522560f5279e2dda4$44 and instead of the class name, it would be the different global setting, and each test case would do $this->setMwGlobals
[19:07:21] <anomie>	 addshore: I don't think we'd want to do that for the tests of everything that touches the comment table, though. And soon the actor table too.
[19:08:06] <addshore>	 anomie: well, in theroy we only need to update the tests when we update DefaultSettings.php to change the value
[19:08:34] <addshore>	 If users change config settings to break tests there isn't really anything we can do
[19:09:46] <anomie>	 addshore: We can write the tests to use setMwGlobals to set the needed values, or make what the tests check adapt, or avoid checking the things that vary if they're not part of what's actually being tested.
[19:10:13] <anomie>	 legoktm: Same response: I don't think we'd want to do that for the tests of everything that touches the comment table, though. And soon the actor table too.
[19:13:25] <legoktm>	 hm right
[19:36:31] <legoktm>	 Krinkle: ideally I'd like to get the @covers validator patch merged, then when I run libraryupgrader against everything for the MediaWiki-CodeSniffer upgrade, we'll spot any other failing extensions
[21:40:46] <anomie>	 DanielK_WMDE: Regarding mwjames's comments on https://gerrit.wikimedia.org/r/#/c/399445/, I wouldn't put it past SemanticMediaWiki to be doing "CREATE TEMPORARY TABLE" calls on a DB_REPLICA handle...
[21:41:02] <Reedy>	 lol
[22:05:14] <DanielK_WMDE>	 hehe. well. *that* would break.
[22:05:27] <DanielK_WMDE>	 but not for the reason he quoted
[22:06:15] <DanielK_WMDE>	 anomie: I have been wondering whether we should make the wrapping optional, and only enabled it in unit tests per default.
[22:07:04] <DanielK_WMDE>	 it's a breaking change, in that it merans getConnection doesn't return a Database instance in all cases. I fixed a few extensions, but I suspect there may be more that type-hint against Database (or DatabaseBase) instead of IDatabase.
[22:07:04] <anomie>	 DanielK_WMDE: Considering the Beta Cluster bug, I'd say keep the default the same everywhere.
[22:07:16] <DanielK_WMDE>	 hmhm...
[22:07:53] <DanielK_WMDE>	 well, the wrap logic is irrelevant to beta, since it uses real replicas. the wrapper is only for when  you get a master connection when asking for a replica
[22:43:02] <bd808>	 somebody should tell mwjames that temp tables created on mariadb/mysql hosts under replication mess everything up
[22:43:55] <bd808>	 per jaime in https://phabricator.wikimedia.org/T179628#3788031