[00:39:20] flood warning https://scontent.fsyd3-1.fna.fbcdn.net/v/t39.30808-6/291831135_411853470984406_1896963075167070283_n.jpg?stp=dst-jpg_p180x540&_nc_cat=1&ccb=1-7&_nc_sid=2c4854&_nc_ohc=4VOk0GerspcAX-QAIfv&_nc_ht=scontent.fsyd3-1.fna&oh=00_AT-EJzer_BD-wOw7oP4OZ2nLIn7GzyaOIOzU8pbDx-X8Rw&oe=62C87743 [00:40:07] what business does a waterway called "wollomi brook" have in flooding such an area [00:40:16] a river, fine, but a brook? [00:57:24] In Dutch a brook is how we refer to pants/trousers [01:03:09] hmm, it's a germanic word, "corresponding in form to Middle Dutch broek" [01:04:43] German bruch (masculine and neuter), moor, marsh, bog, fen [01:04:55] I think it's Dutch that has some explaining to do there, not English [01:09:19] Wiktionary has both senses of broek (marsh and trousers) and says they were also the same in middle dutch, however, marsh is male whereas trousers is female [01:09:36] so that's how you can tell which one is meant [01:12:19] oh, and Old English had brōc from the same source from which we get breeches [01:17:41] OED says trousers was bréc/bræc in Old English, already distinct from bróc (water) at that time [01:23:01] we do have a few geographical names in the Netherlands that end in -broek. I believe in our case those historically referred to low moist lands, e.g. swamp-like. e.g. https://en.wikipedia.org/wiki/Babyloni%C3%ABnbroek and https://en.wikipedia.org/wiki/Bornerbroek. But the most common use and connotation with broek is pants for us, e.g. everyday it's put your broek on, or put your onderbroeken in the laundry :) [01:27:34] last year I was explaining to platform team how our nearest river is pretty tame and small, but now it's overflowed its banks halfway up the valley https://scontent.fsyd3-1.fna.fbcdn.net/v/t39.30808-6/291958020_3101382940174942_4224387813612468367_n.jpg?_nc_cat=1&ccb=1-7&_nc_sid=730e14&_nc_ohc=lSdhpCOwmqgAX_yJe_V&_nc_ht=scontent.fsyd3-1.fna&oh=00_AT_SbOXeQtw-92boxd_GLtp4frlC59vadQIw6nYmqiqpLQ&oe=62C92EB6 [01:35:13] the height is 9m and rising [01:45:51] * Krinkle tries to surpress Netherlandish impulses for building dikes and storm flood barriers [01:48:52] I imagine in coming years this or alternative approaches will need a lot of attention [01:56:17] rainfall is not like a storm surge though, the water's got to go somewhere https://www.abc.net.au/news/2020-09-12/report-slams-waragamba-dam-raised-wall-plans-indigenous-heritage/12656878 [01:58:15] ah, I see. this is coming from inlands (trying) to go out to the ocean [01:58:26] but there's a city in the way [01:58:55] the outskirts of a city [01:59:32] the water either has to flow along the floodplain or be stored inland [02:00:49] ack [02:03:44] I suppose rerouting only increases the scale of the problem as it's presumably not all raining down into a nice confined area, but from a large area concentrated and drawn to this one outflow. [02:03:44] none of this flooding is really a surprise, there were models of course [02:05:31] and it's always been this way, there was an account from someone in Lismore, which is a country town which was devastated early this year by flooding [02:06:23] he was asked if he will leave, he said his father's property was destroyed by flood in the 70s, and his grandfather's in the 50s, so he'll rebuild just like they did [02:07:43] people with that generational wisdom understand the costs of living near a river [02:09:59] on the other hand, they are littering [02:10:28] they build their stuff and then leave it strewn all up and down the coast on the beaches once every 50 years or so [02:13:11] right.. [02:14:17] to confirm, this country town is down from the dam, not the one treatened with being flooded more heavily as result of would-be raising of the dam, right? [02:16:33] I think Lismore's river is uncontrolled [02:16:51] I suppose there's at least some element of choice there. I dont know to what extent people who live there have the oppertunity/resources to move elsewhere if they wanted to. [02:16:51] just a natural river [02:17:20] NL is a lot more space constrained I imagine. It's not that we wanted to preserve the towns near the ocean per-se. [02:20:00] the state of NSW is a bit bigger than France, population density of 10/km² compared to NL 423/km² [02:25:30] coastline length is comparable though, straight line coastline length is about 1100km vs 340km [02:30:08] GDP per linear dimension about $0.41 vs $3.10 [02:34:25] billion USD per km [03:12:27] also, white people in Australia didn't settle the river valleys as a last resort, they settled the river valleys first, for farming [03:13:46] the FB photo I shared is Yarramalong, where it is still that way, only the valleys are cleared and settled, the ridges are untouched [04:41:13] TimStarling: how much if any will you be backporting around the getParser fix for Scribunto? [04:41:35] Looks like the core parts of it are safe to backport, at least the new method on Parser and ParserFactory? [04:42:25] I suggested a few temporary workarounds for miraheze and said that my main patch would not be for backport [04:42:42] if someone wants to backport it, that's fine, but I wasn't planning on doing that [04:44:53] okay [05:38:37] thanks for reviews again, my list is getting so short [05:39:10] I have a +1 from vgutierrez to deploy https://gerrit.wikimedia.org/r/c/operations/puppet/+/801621 so I might do that now [10:18:03] Krinkle: right, it seems that the Title::getContentModel() call in initializeArticle() is what triggers the first local DB connection + query now [10:18:11] I think there is another small optimization opportunity here [10:18:46] WikiPage::pageData always triggers a DB query, but populates LinkCache with the results, so if we call getContentModel() on the WikiPage instance in initializeArticle() rather than on the Title, we may be able to avoid one query to the page table [10:19:17] as it simply moves the WikiPage init earlier, but also seeds LinkCache with the result row data, so any future access on that Title can leverage data from LinkCache [15:32:24] mszabo: T206498 [15:32:25] T206498: MediaWiki needlessly queries page data twice on page views - https://phabricator.wikimedia.org/T206498 [16:05:56] ha! [16:14:18] https://gerrit.wikimedia.org/r/c/mediawiki/core/+/811333 [16:14:46] I haven't been able to truly verify this yet because we have a badly written BeforeInitialize hook handler that likewise triggers an early Title init that I'll have to do something about first [16:28:05] the caller fname in your log from yesterday does seem to line up with what Title::getContentModel() + cold LinkCache would produce [21:29:26] mszabo: php7.4 bench: https://phabricator.wikimedia.org/T311386#8052799 [21:30:02] looks like some of your wikis are 7.3.33 currently [21:31:49] Krinkle: yeah, the plan is to go to 8.0 after the 1.37 rollout is complete [21:31:57] for now everything is on 7.3.33 still [21:32:08] I see you are on LuaSandbox 4 already [21:32:16] according to our Special:Version we are on 3.x stll [21:32:18] not sure why.. [21:32:25] hmmm [21:32:31] https://gerrit.wikimedia.org/r/plugins/gitiles/mediawiki/php/luasandbox/+/refs/heads/master/package.xml [21:33:08] Seems 4.0 only drops PHP 5 support so it should be fine [21:34:15] and "Remove memory leaks in data_conversion.c" [21:35:43] that seems generally useful [21:42:49] I'll also want to play with preload a bit, I see the overarching attempt to eagerly add mostly everything is postponed for now [21:42:57] so I'd only do something limited and work my way up [21:43:44] ref T240775 https://stitcher.io/blog/php-preload-benchmarks [21:43:44] T240775: RFC: Support PHP 7.4 preload - https://phabricator.wikimedia.org/T240775 [21:43:50] Tim was doing some work in that area [21:44:00] It's tricky with multiversion the way we currently have it [21:45:05] plus as I see it broke logging in the local dev env somehow [21:45:45] one thing I'm kinda hoping for is that preload could mitigate or maybe even eliminate the performance hit from having to autoload the 300+ HookRunner interfaces every request [21:49:07] That's one way indeed. [21:49:17] Another is to swap out the runtime implementation with one that doesn't implement the interfaces [21:49:41] and leave the current one for Phan/CI/IDEs only. ref T274041 [21:49:42] T274041: Reduce performance impact of HookRunner.php loading 500+ interfaces - https://phabricator.wikimedia.org/T274041 [21:50:13] I'm looking to do that for load.php at least since for local development, especially with debug=2 which I'd likek to make the default, will make more requests to localhost that go through PHP [21:50:23] which currently all get a 200ms hit or so initialising PHP much of which is spent in HookRunner [21:50:36] and preload isn't really an option in local dev I think unless it continously restarts every few seconds. [21:51:04] I mean, we'll want to be able to test it locally of course, but not as a default setting for the local dev. [21:51:32] yeah, it may have to be opt-in for local dev and then only on beta [21:52:14] the PECL memcached library finally also got a proper 3.2.0 release thanks to Remi [21:53:09] yeah saw that, nice to see all the work Remi is doing. [21:53:35] PHP extension necromancing :)