[06:35:54] <tgr>	 all this time I thought we use __METHOD__ all over the place as a default function argument because that somehow magically takes the name of the calling method, but that doesn't seem to be the case
[06:36:29] <tgr>	 why are we doing it then? as a hint of what value you are supposed to provide?
[06:42:13] <legoktm>	 Yeah, that's what I thought it was used as (a hint)
[08:55:56] <legoktm>	 Reedy: https://integration.wikimedia.org/ci/job/phpunit-coverage-publish/434/console PasswordBlacklist has too many tests.
[09:02:35] <legoktm>	 Reedy: https://gerrit.wikimedia.org/r/410723
[09:10:45] <hashar>	 legoktm: or just test a few from the giant file ? :]
[09:11:20] <legoktm>	 yeah probably, I'll let Reedy decide what he wants :)
[09:11:38] <hashar>	 oh while you ar earound
[09:11:50] <hashar>	 thanks A TON for the migrate of the Jenkins job that compiles the php  extensions
[09:13:18] <legoktm>	 no problem :)
[09:13:26] <legoktm>	 though I didn't finish yet, there's still hhvm and php55 left
[09:14:51] <hashar>	 for php55 you can use the jessie image then you would need to inject a specific source list . dockerfiles/php55/Dockerfile.template has the details
[09:15:02] <hashar>	 but maybe that image can just be extended
[09:16:14] <legoktm>	 I think it can be extended
[10:05:19] <addshore>	 stephanebisson: not yet!
[10:05:29] <addshore>	 Not had much time to spend looking at it, but perhaps today!
[14:33:33] <stephanebisson>	 addshore: I don't know if it tells you anything: "Revision::getContent: Cannot get content: Failed to load data blob from tt:375056: Failed to load blob from address tt:375056"
[14:35:24] <addshore>	 well, it defintly gives me something to look at!
[14:35:46] <addshore>	 So stephanebisson it apparently cant access row with id 375056 in the text table
[14:36:09] <addshore>	 which wiki was that from?
[14:37:05] <stephanebisson>	 addshore: testwiki
[14:37:14] <addshore>	 mysql:research@analytics-store.eqiad.wmnet [testwiki]> select * from text where old_id = 375056;
[14:37:14] <addshore>	 Empty set (0.01 sec)
[14:37:34] <addshore>	 so the text entry doesnt exist
[14:39:06] <addshore>	 And I can't find anything that refers to text id 375056 in the archive or revision table O_o
[14:39:07] <stephanebisson>	 addshore: This is happening during a page move. It is 100% reproducible in prod but 0% in beta or local (where there is no possible replica lag)
[14:39:45] <addshore>	 just on regular page move? or only via this flow optin?
[14:40:24] <addshore>	 I have a master replica setup locally and was hoping to reproduce it yesterday but dont have parsoid setup so I can't make flow let me opt in apparently
[14:40:47] <stephanebisson>	 addshore: via Flow opt-out but it also shows up from time to time during other page moves. They should start showing up in logstash in channel T184670.
[14:40:48] <stashbot>	 T184670: [wmf.16-regression]  Fatal exception of type "Flow\Exception\InvalidDataException"  for opting out from "Structured Discussions on user talk" - https://phabricator.wikimedia.org/T184670
[14:41:01] <stephanebisson>	 right :(
[14:41:56] <stephanebisson>	 addshore: it's not too hard to setup with mw-vagrant but from scratch it would be a pain
[14:42:40] <addshore>	 So, i already fixed one issue very similar to this for the flow stuff, again due to master / slave / stuff not quite happening right https://phabricator.wikimedia.org/T184691
[14:43:59] <addshore>	 I'm pretty sure we also fixed an issue with moving pages, but I can't find the ticket, so I might be imagining that
[14:45:08] <addshore>	 stephanebisson: I'll have a look now I have this stack trace and see what I dig up
[14:45:59] <stephanebisson>	 addshore: thanks. let me know if you have any leads
[14:48:00] <addshore>	 seems a bit odd that it is refering to a new text id though, when this is a page move, so there wouldnt be a new id, unless the id is for the text of a redirect being left behind (which is new content)
[14:49:29] <stephanebisson>	 addshore: well, it does newNullRevision() on an existing revision and then fail to get content from it, ultimately all that is rolled back at the end
[14:50:11] <addshore>	 so newNullRevision uses the test_id of the previous revision, so it should exist :/
[14:50:21] <stephanebisson>	 but $newRev->getContent() should try to read from master in this case, I can't tell if it does by looking at the code
[14:50:29] <stephanebisson>	 right
[14:50:34] <addshore>	 *text_id
[14:51:10] <addshore>	 well, even if it doesnt read from the master, the parent revision creation happened in another request anyway? so it would be there?
[14:51:33] <stephanebisson>	 in the context of the Flow opt-out, there are several operations, edits + moves
[14:51:39] <addshore>	 okay
[14:51:43] <addshore>	 *goes to look at that again*
[14:51:54] <stephanebisson>	 so the parent rev may have been *just* created and not replicated
[14:53:09] <addshore>	 aah yes, removeArchiveTemplateFromWikitextTalkpage then addCurrentTemplate and finally movePage
[15:00:53] <addshore>	 DanielK_WMDE: hows your week off going?
[15:06:37] <addshore>	 stephanebisson: got a meeting now but will continue looking after, I added one of the stacktraces to the ticket
[16:19:09] <addshore>	  stephanebisson so, I managed to run the code path locally, and I don't get the issue, even with a master replica setup
[16:20:40] <stephanebisson>	 addshore: interesting. can you force some lag in your setup?
[16:21:07] <addshore>	 stephanebisson: probably :)
[16:21:34] <addshore>	 but even the 200-1000ms lag there always is has been enough to catch all of the other other replication related issues
[16:21:47] <stephanebisson>	 addshore: also, make sure you have a wikitext talk page before playing with the Flow beta feature