[05:22:59] Whee https://github.com/wikimedia/less.php/releases/tag/v1.8.0 [05:23:27] Any Packagist admins around? [06:01:16] MaxSem: I can do it in a bit [13:25:16] TimStarling: i found a remex bug w/ PHP 7.0 [13:25:32] https://gerrit.wikimedia.org/r/467470 [16:45:30] For T194052, I need to make a (very small) change to base-convert, which is under vendor/wikimedia, and which also lives on packagist and github. Do I start with Gerrit code review via the vendor git repository and then someone updates packagist and github afterward, or how does that work? [16:45:33] T194052: base-convert no longer works under PHP 7.2 because dividend and divisor are no longer truncated to integer when bcmath is installed and gmp is not - https://phabricator.wikimedia.org/T194052 [16:55:07] bpirkle: submit the patch to https://gerrit.wikimedia.org/r/plugins/gitiles/base-convert/, and then once it's merged we can tag a new release and update it in mw/vendor [16:55:29] legoktm: will do, thank you [17:38:09] bd808: re https://phabricator.wikimedia.org/T180278#4669337 do you have a list of repos that were the special PHP apps you worked on? [17:39:41] legoktm: wikimania scholarships and grant review are the two that I helped put into prod. I can add repo links to that ticket [17:40:47] there is a 3rd shared library/framework part of that too that is where most of the 3rd party things are pulled in [22:52:15] MWException from line 64 of /srv/mediawiki/php-1.32.0-wmf.24/includes/MWNamespace.php: MWNamespace::getAssociated does not make any sense for given namespace -8673 [22:52:27] Same repeated 10,000 times with different negative ids [22:52:33] Not quite sure what it make of it. [22:52:36] fatal exceptions [22:53:42] Krinkle, can you live hack it to add the trace? [22:53:51] fatals have traces [22:53:56] I'm looking at it [22:54:10] it's SpecialContributions :) [22:54:22] I guess maybe some form of https://phabricator.wikimedia.org/T150324 [22:54:29] hm [22:54:46] namespace=-8673%29%20AS%20xytg%20WHERE%201438%3D1438%20OR%20%28SELECT%20%28CASE%20WHEN% [22:54:56] I feel like there was a time when not all such errors/warnings/stuff came with a trace? [22:54:57] clearly someone messing about trying to sql inject (and failing) [22:55:37] every fatal and exception has had a trace for 4+ years, every php error (notice/warning/"error") has a trace since 2 years. [22:55:59] must be the notice/warning/"error" I was thinking of then [22:56:02] ok [22:56:20] As of a month, we log the php errors as logstash severity "error" instead of "info", which for some made them appear for the first time due to filters on dashboards. [22:56:58] Also, some dashboards were/are querying type:hhvm instead of type:mediawiki channel:error, which means it's just plain stderr with no trace or any mw context. [22:57:18] As of last week, we no longer send them to hhvm when they go to mw/error [23:05:14] updated the task [23:05:26] seems to be triggered by 'associated' namepaces [23:05:38] regular namespace=-1234 is handled gracefully