[01:29:10] is there any legitimate reason subpages should show up on maintenance reports? [13:23:40] c: Probably. What "maintenance reports" are you talking about? [13:24:32] anomie: subpages are inherently part of the main article and shouldnt be double categorized. not a one size fits all situation [13:28:46] c: If you're talking about the main article namespace on Wikipedia, there generally aren't subpages there in the first place (exceptions: rmwiki and tenwiki). The same seems to mostly hold true for other Wikimedia content wikis too, although there are a number of Wikiversities where subpages are enabled in the main namespace. [13:29:02] anomie: no i am not [13:29:15] c: Then you should probably specify what you are talking about, specifically. [13:29:31] "main article" sounds like you're talking about Wikipedia. [13:29:53] anomie: i am speaking in general for any random wiki. subpages should not show up on uncategorized pages for example [13:30:44] c: I doubt you can sensibly speak "in general" since whether including subpages is appropriate depends very heavily on how the specific wiki is using subpages and the purpose of the specific maintenance report. [13:31:43] anomie: in this case subpages are included/embedded in the article [13:34:12] c: In that specific case, where a wiki is transcluding article-space subpages into the parent page and looking at the uncategorized pages report, then yes, you probably wouldn't want subpages to show up. Change those variables and the answer may change too. You might be best served by an extension to add a version of uncategorized pages specific to your wiki's use case than trying to make a blanket statement about all maintenance pages on all [13:34:12] wikis. [13:34:59] what variables? and i dont think such a specific extension exists [13:35:42] <_joe_> anomie: how can I check if luasandbox is using the library or shelling out to the lua interpreter instead? [13:36:29] <_joe_> also, is it possible that there is some resource contention when running a lot of luasandbox instances in php-fpm? like some shared sempahore [13:37:18] _joe_: The LuaSandbox PHP extension never shells out. Scribunto shells out depending on the value of $wgScribuntoDefaultEngine. [13:37:30] <_joe_> anomie: oh ok thanks [13:37:39] <_joe_> sorry, I even knew that :) [13:38:15] <_joe_> I'm trying to understand the reason for a mysterious pattern in php-fpm behaviour and the slow log traces were rich in luasandbox functions [13:42:19] _joe_: There's one process-global rwlock in luasandbox, IIRC. It controls access to the data structure used for allocated timers. Look for timer_hash_rwlock in timer.c. [13:42:48] <_joe_> but it's per-process [13:43:05] <_joe_> so it would possibly affect multiple threads, but fpm doesn't do that [13:43:21] <_joe_> I fear the issue is deeper within the bowels of php-fpm :/ [13:43:37] <_joe_> things like opcache are shared across processes [13:47:28] I don't think luasandbox uses opcache, at least not explicitly. I suppose it's possible PHP might do something under the hood. [15:26:34] Krinkle: Your statement in https://gerrit.wikimedia.org/r/c/mediawiki/core/+/531704 that we use @internal instead of @private does not seem to be true. We seem to have a mixture of both used for the purpose.