[00:26:11] AaronSchulz: ack, left 2 questions. [00:26:17] RE: setMulti, makes sense [00:26:32] I actually had a vague sense that in the lower classes we always end up using the Multi variant on the wire [00:26:52] I guess not. would like to better understand the cost overhead and where it sits, but fine for sure, just curious/learning. [03:16:48] https://twitter.com/vrandezo/status/1380341525054287872 [03:20:53] HaeB: He, nice [03:23:15] Looks like the comparison is about https://en.wikipedia.org/wiki/Dragon [03:23:34] based on seeing Ninedragonwallpic1.jpg in the timeline and GlobalUsage [03:24:32] hm.. no, that one is quite a bit larger [03:25:28] ah en.m. [03:26:03] 407 Kb transferred on load (no scroll,lazy loading) [03:54:19] Krinkle: replied [04:03:15] AaronSchulz: ok, so rather than onhost being a GET-only route, we'd have wancache only use it on get, and not use it on SET [04:03:30] that makes sense indeed, I suppose not limited to tombstones per se [04:03:34] but either way is fine [04:03:43] might be simpelr due to get-with-set to not special case it there [04:03:54] no harm anyway even if prod does store it (which it won't) [04:10:24] AaronSchulz: Ok, LGTM. let's land this :) [06:27:47] Krinkle: did you mean to change https://gerrit.wikimedia.org/r/c/mediawiki/core/+/672514/ to +1 ? [09:57:24] phedenskog: 5 second session windows, but presumably CRuX will record and take into account all the windows? [09:58:19] and treat of of them as a separate pageview [14:06:29] gilles: hmm I wonder what the starting point would be of the second window (or whichever the banner appears in) would be [14:06:38] Might it depends on user interaction [16:41:01] Krinkle: ah, because it's your commit ;) [16:41:34] * AaronSchulz +2s [17:26:12] yeah... [17:48:54] Krinkle: https://gerrit.wikimedia.org/r/c/mediawiki/core/+/643385/4 is conflicted [17:49:48] yeha, I'm rebasin' it [17:49:58] https://gerrit.wikimedia.org/r/c/mediawiki/core/+/673331/ [17:50:00] is ready tho [17:52:22] ci taking forever on https://gerrit.wikimedia.org/r/c/mediawiki/core/+/672514/12 [17:54:30] https://integration.wikimedia.org/zuul/?filter=gate [17:54:31] soon :) [17:54:42] (that link doesn't actually work, I still haven't fixed that yet.) [18:21:02] Krinkle: lmdb for php_dba sounds cool; http://www.lmdb.tech/bench/microbench/ [18:23:16] interesting stats, bookmarked [18:23:31] does it beat array/opcache for read-only though? [18:24:05] but perhaps something for the local server stash to consider or as better default than cdb for some things [18:24:33] leveldb powers chrome's localstorage afaik [18:26:12] https://github.com/mpremus/lmdb-php exists as a wrapper too (newer than the dba one) [18:29:17] ah, that requires a c extension. [18:29:38] if the algo is so good, then a pure php implementation might still beat some of the others (e.g. cdb or sqlite) [18:30:11] unles it depends on certain direct low level concepts that might be difficult to approximate otherwise [18:32:24] rebased and amended https://gerrit.wikimedia.org/r/c/mediawiki/core/+/643385/ [18:32:31] what do you think re 1.36? [18:33:07] (it's only been 24h :P) [18:34:24] Krinkle: just doing "sudo-apt get install lmdb-utils php-dba" does the trick [18:35:19] Krinkle: that is still failing ci [18:36:16] different failure this time [18:36:21] I forgot to update the optoin name there [18:36:35] It passed right before I did the find/replace [18:36:40] locally :D [18:38:08] alright, now it passes locally, amednemtn up [20:40:41] Amir1: https://gerrit.wikimedia.org/r/c/mediawiki/extensions/Wikibase/+/676106 should be good to go [20:40:57] there was a lot of test failure spam before, sorry [21:33:40] ori: Thanks. I take a look after the coffee kicks in [23:05:31] Krinkle: https://gerrit.wikimedia.org/r/c/mediawiki/core/+/643385 should change the 1.37 notes [23:11:07] AaronSchulz: well, I did it this way intentionally [23:11:18] I was asking earlier there and here about 1.36 [23:11:20] it's only just been cut [23:11:35] seems perhaps worthwhile renaming there rather than diverging and having two mutli-dc related breaking changes [23:18:33] ok [23:55:21] Krinkle: https://gerrit.wikimedia.org/r/c/operations/mediawiki-config/+/677418/3 should be merged now [23:55:33] (not that it is merged) [23:57:17] yeah, I'll roll that out monday or tuesday morning