[01:03:15] <anomie>	 Krinkle: Try using a more recent version of HHVM? 3.6.6 is pretty old.
[01:05:18] <Krinkle>	 anomie: https://docs.travis-ci.com/user/languages/php#HHVM-versions
[01:05:21] <Krinkle>	 3.12?
[01:05:35] <Krinkle>	 seems they provide them, but 'hhvm' defaults to 3.6
[01:05:48] <anomie>	 3.12 is what we use on the cluster
[01:07:42] <Krinkle>	 anomie: release notes say 3.6.5+
[01:07:51] <Krinkle>	 we should probably u pdate that
[01:09:54] <Krinkle>	 anomie:  https://gerrit.wikimedia.org/r/339096
[01:11:05] <Krinkle>	 Seems since 4e6810e4a2c1d821d8d108c7974ac16917561764 the travis test also fails PHP 7 due to an extra 'stack' property
[01:11:11] <Krinkle>	 testExceptionErrors
[01:23:04] <anomie>	 Krinkle: Do you have a bug for the PHP 7 error?
[01:24:41] <Krinkle>	 anomie: Not yet
[01:31:32] <anomie>	 Krinkle: Well, I think https://gerrit.wikimedia.org/r/339099 should fix it.
[01:36:25] <Krinkle>	 anomie: Cool. One other question in case you know.. Sometimes deleteDefaultMessages/deletEqualMessages fails recently about UserGroupMembership::insert()
[01:36:26] <Krinkle>	 Caught exception UnexpectedValueException: UserGroupMembership::insert() needs a positive user ID. Did you forget to add your User object to the database before calling addGroup()?
[01:36:46] <Krinkle>	 Looks like addGroup after newSystemUser/steal no longer works. Requires some kind of change?
[02:36:20] <tgr>	 Krinkle: since adding ug_expiry, it's not possible to add groups to a User that's not in the database, maybe that's related
[02:37:19] <Krinkle>	 tgr: Isn't it a fairly common pattern for maintenance scripts to use a system user and at run-time ensure it is a bot? 
[02:37:27] <Krinkle>	 Presumably that code isn't all broken now?
[02:37:47] <Krinkle>	 I suppose either this script does it differently or wasn't found in migration
[02:38:37] <tgr>	 I don't think it is
[02:39:18] <Krinkle>	 Hm.. newSystemUser has create=>true and will insert into the database if not already
[02:39:27] <tgr>	 tests that just use "new User" instead of "$this->get[Mutable]TestUser" are broken though
[02:39:34] <Krinkle>	 using createNew()
[02:40:17] <Krinkle>	 So how can it not exist?
[02:40:41] <tgr>	 newSystemUser should never return a user that does not exist
[02:41:05] <Krinkle>	 Exactly
[02:41:20] <Krinkle>	 https://github.com/wikimedia/mediawiki/blob/master/maintenance/deleteEqualMessages.php#L165-L173
[02:42:01] <tgr>	 could be some sort of cache invalidation issue, I have seen that elsewhere
[02:43:18] <Krinkle>	 I'll file a task next time I get it
[02:43:26] <Krinkle>	 Lost the trace and wiki I got it on last time
[02:45:51] <tgr>	 Krinkle: T142134 seemed to be related to WAN cache invalidation problems in tests, maybe it's the same underlying issue
[02:45:52] <stashbot>	 T142134: BounceHandler UnSubscribeUserTest::testUnSubscribeUser test failing - https://phabricator.wikimedia.org/T142134
[06:56:34] <tgr>	 anomie: sorry, forgot to ask about SoS - if you have anything extra, could you add it to the etherpad and mail Bernd if it requires an explanation?