[00:00:11] pizzzacat: you were making the new branch the same as master, for now. if you set the initial SHA-1 to the commit *before* your first feature patch, that is more correct [00:00:28] huh: hi [00:00:45] oh awight the last part of that makes it make sense [00:01:18] awight: I know the "X" button exists, BTW [00:01:51] huh: I think I swore at the fullscreen banner myself, when I noticed that I couldn't dismiss it with the usual or clicking outside the dialog. [00:02:07] but it won't come back, now that you've closed it :) [00:02:09] huh: but is it annoying /enough/ ? :P [00:02:18] ;) [00:02:29] moar pink! Call in Jimmy photoze! [00:02:52] There are reasons we don't get to design banners. [00:04:23] (CR) Ejegg: [C: 2] typo-O [extensions/DonationInterface] - https://gerrit.wikimedia.org/r/175591 (owner: Awight) [00:05:57] No colors for grrrit? [00:06:18] (Merged) jenkins-bot: typo-O [extensions/DonationInterface] - https://gerrit.wikimedia.org/r/175591 (owner: Awight) [00:07:27] huh: Just two. :) [00:11:23] awight: Do you need anything else to go out in the payments deploy I'm about to do? [00:11:33] ejegg and pizzzacat too. [00:11:40] not I [00:11:47] K4-713: no [00:12:03] Good. This should be simple. [00:12:05] ... [00:12:11] I have no idea why I typed that just now. [00:12:15] :) [00:12:21] * K4-713 would like more trouble [00:15:01] I'm good [00:15:28] ejegg: what about yr WP styles? [00:15:54] awight: Oh, we could do the button stuff [00:16:02] GC though, I think [00:16:17] The RTL bit will require talking to GC [00:16:28] ejegg: oh! [00:16:28] (PS1) Katie Horn: Merge branch 'master' into deploy [extensions/DonationInterface] (deployment) - https://gerrit.wikimedia.org/r/175612 [00:16:33] hehe too late [00:16:46] you guys are rad [00:16:57] i'm heading out - see you mañana! [00:17:06] later skater [00:17:06] atgo: Later! [00:17:13] Yeah, for RTL we need to tell them to source a different stylesheet for specific langs [00:17:14] ...you win this time, awight. [00:17:16] so long atgo [00:17:19] K4-713: that was creepy [00:17:28] * awight dons tin foil [00:17:31] what K4-713, you're not going to jinx him? [00:17:34] btw, i owe you a coke. [00:17:35] (CR) Katie Horn: [C: 2] Merge branch 'master' into deploy [extensions/DonationInterface] (deployment) - https://gerrit.wikimedia.org/r/175612 (owner: Katie Horn) [00:18:13] If I keep this up, everybody's going to notice I can read minds. [00:18:17] hahaha [00:18:37] Just please don't tell me what happens in tomorrow's episode. [00:18:55] back in a bit, gonna relocate [00:19:21] ewulczyn: just confirming that you saw this handoff: https://wikimedia.mingle.thoughtworks.com/projects/online_fundraiser/cards/2175 [00:19:48] ewulczyn: There was a 100% test under the new client-side banner code, today, 2300-2400 UTC. [00:20:04] If there's a Faraday cage around my desk when I get to work tomorrow, I'll understand. [00:20:47] ewulczyn: We wanna know that the impressions are at the same proportion of total pageviews as in the last few days... [00:23:14] fyi the 100% US test came down about 20 mins ago PPena AndyRussG awight K4-713 [00:23:50] Cool. :) [00:24:05] meganhernandez: Ah OK... Uh that was due to...? (sorry if I didn't see relevant backscroll...) [00:24:31] AndyRussG: good to test 100% US with buckets A, B, C & D tomorrow morning? I was aiming for around 6 o 6:30am pacific [00:24:50] AndyRussG: it was the full hour [00:25:00] awight: Ah OK thanks [00:25:03] meganhernandez: sure! I'll be here [00:25:07] time flies when you're having fun [00:25:31] A fix for the previewing banners on wiki issue is almost done, should be deployed this eve [00:25:43] meganhernandez: ^^ [00:26:07] oh right, yeah we’d need that before we can test those new banners [00:26:53] meganhernandez ok thanks [00:27:06] (PS2) AndyRussG: Fix previewing banner on-wiki [extensions/CentralNotice] - https://gerrit.wikimedia.org/r/175606 [00:27:09] Woah. New phabricator spam. [00:27:18] hah. I tricked you. [00:27:36] CURSES! [00:27:40] Foiled again. [00:27:42] (CR) AndyRussG: "Tests now pass..." [extensions/CentralNotice] - https://gerrit.wikimedia.org/r/175606 (owner: AndyRussG) [00:28:02] Okay, deploying payments things. [00:28:21] (CR) Awight: [C: 2] Fix previewing banner on-wiki [extensions/CentralNotice] - https://gerrit.wikimedia.org/r/175606 (owner: AndyRussG) [00:28:34] :) ^ [00:28:41] AndyRussG: what do you think about https://gerrit.wikimedia.org/r/#/c/175597/ ? [00:28:50] awight: just about 2 rebase... [00:28:54] k [00:30:28] awight: that patch won't bork produciton configs, I guess? It's actually not set to false anywhere? [00:30:40] (PS2) AndyRussG: default to client banner choice [extensions/CentralNotice] - https://gerrit.wikimedia.org/r/175597 (owner: Awight) [00:31:11] AndyRussG: it's set to true, unconditionally, now [00:31:18] (Merged) jenkins-bot: Fix previewing banner on-wiki [extensions/CentralNotice] - https://gerrit.wikimedia.org/r/175606 (owner: AndyRussG) [00:31:29] ooo hooo so now it'll be set to true twice [00:31:38] :) [00:31:39] Double positive makes a negative? [00:31:55] Tests pass... https://gerrit.wikimedia.org/r/#/c/175597/ [00:31:55] !log updated payments to 3e3cda8f07af9f7f7 [00:31:57] double positive causes suspicion [00:31:59] Logged the message, Master [00:32:05] K I'll merge [00:32:24] nice to see that the qunit actually would have caught the bug :) [00:32:31] If I only ran it correctly ugggh [00:32:44] so close! [00:33:52] K4-713: is the WR1 thing going to be admitted to the operating room?... [00:34:08] I just need to poke it with sticks for a while. [00:34:17] ...until I get time to fix it for real. [00:34:27] ok [00:34:29] And by "fix it" I mean use the WX files. [00:34:41] awight: funny I don't see the bannerController.lib tests running on Jenkins: https://integration.wikimedia.org/ci/job/mwext-CentralNotice-qunit/248/consoleFull [00:34:49] ...and then gently nudge the old code out to sea. [00:35:08] Still let's merge 4 now [00:35:11] AndyRussG: ouch, 00:30:48 Warning: No route to fetch banner choice data configured. [Called from CNBannerChoiceDataResourceLoaderModule::getChoices in /srv/ssd/jenkins-slave/workspace/mwext-CentralNotice-qunit/src/extensions/CentralNotice/includes/CNBannerChoiceDataResourceLoaderModule.php at line 60] in /srv/ssd/jenkins-slave/workspace/mwext-CentralNotice-qunit/src/includes/debug/MWDebug.php on line 300 [00:35:32] awight: yeah saw that. That's also due to default config [00:35:34] AndyRussG: but I do see the tests running: 00:30:53 ext.centralNotice.bannerController.lib - allocations test cases...OK [00:35:57] Unfortunately, we don't get a line per scenario... but they're there. [00:36:25] ccogdill: obt payment methods are back up. :) [00:36:32] That's... ideal, bpay... [00:36:40] woo! [00:36:41] awight: ah yes ur right... (derp again) [00:36:48] I wasn’t expecting it so soon, thanks k4-713 [00:37:09] We should have independent confirmation, and then have the-wub re-enable the buttons on donate. [00:37:34] (CR) AndyRussG: [C: 2] "\o/" [extensions/CentralNotice] - https://gerrit.wikimedia.org/r/175597 (owner: Awight) [00:38:05] (Merged) jenkins-bot: default to client banner choice [extensions/CentralNotice] - https://gerrit.wikimedia.org/r/175597 (owner: Awight) [00:38:11] ccogdill: Wait, actually, that's obt and rtbt. [00:38:45] ccogdill: bpay, sofort, ideal... enets? [00:39:13] yandex? webmoney? [00:39:38] Probably! [00:39:40] awight: see any issues ahead of deploy? [00:39:44] woop! [00:39:54] ccogdill: It was a silly thing. [00:40:01] But it was in there pretty good. [00:40:07] glad to hear it, silly is easier [00:40:14] it’s not on the form yet - any idea on an ETA [00:40:31] AndyRussG: looks good, but I'm not looking :) [00:40:31] I mean, it's fixed in production now. [00:40:37] ccogdill: ^^ [00:40:53] ohh that’s a the_wub thing huh k4-713 [00:40:57] Yep! [00:41:14] got it, thanks! [00:41:16] I'll send out an email about it in a minute. [00:41:23] Just thought you'd like to know early. [00:41:55] oh excellent, saves me from sending one :) [00:42:11] awight: O_O [00:42:18] * AndyRussG looks and looks [00:49:08] K4-713: cute photo, whenever you need one... [00:49:17] * K4-713 swivels [00:53:04] meganhernandez are those the final numbers on the A/b spreadsheet? [00:55:30] (PS1) Awight: Merge master to wmf_deploy [extensions/CentralNotice] (wmf_deploy) - https://gerrit.wikimedia.org/r/175621 [00:55:42] (CR) Awight: [C: 2] Merge master to wmf_deploy [extensions/CentralNotice] (wmf_deploy) - https://gerrit.wikimedia.org/r/175621 (owner: Awight) [00:56:15] (Merged) jenkins-bot: Merge master to wmf_deploy [extensions/CentralNotice] (wmf_deploy) - https://gerrit.wikimedia.org/r/175621 (owner: Awight) [01:00:26] waiting for Jenkins merge... [01:01:35] K4-713: oooof, I broke all that? [01:01:42] Who knew "return" could be so important [01:01:48] Looks like some of it was broken a bit before. [01:01:54] :p [01:03:12] K4-713: oh, that Complete Discography was of Minor Threat. Don't know how that affects your decision to cp. [01:05:03] awight: yeah Jenkins has been a bit pokey tonite [01:07:14] awight: thanks for preparing the patches, also smoke tested locally [01:07:14] AndyRussG: I noticed it runs both test and gate+submit jobs, even after I've CR+2'd. [01:07:26] It should cancel the "test" jobs IMO [01:07:26] Hmmm [01:07:37] otherwise, it's wasting 50% of the cpu! [01:07:38] Maybe it tests the merged version? [01:08:16] ...what the heck, fredge. [01:11:38] K4-713: ? [01:11:50] Are you not getting the failmail? [01:12:01] Not yet... [01:12:03] Oh, for... [01:12:05] What fanmail? [01:12:12] "Data too long for column" [01:12:19] Fail mail. [01:12:20] :p [01:12:33] Well, that's dumb. [01:12:36] i'm not a big fan of fail [01:12:47] Apparently, I love it. [01:13:21] This is what I get for testing weird payment methods with... outrageously long names. [01:13:28] Hmmm [01:13:35] awight: anything I can do BTW? [01:13:36] who'da thunk it would be auf deutsch? [01:13:41] Right? [01:15:01] The weirder thing is that some weirdo put it in our system in long form. [01:15:34] !log disabling fredge consumer [01:15:39] Logged the message, Master [01:18:21] I think I'm just going to... increase these field sizes from varchar 16 to 32. [01:18:43] go for it! [01:22:59] awight: is deploy? [01:23:55] Oh I see ops [01:30:10] AndyRussG: ejegg: yep, I stepped on a mine. Trying to fix my shtitstorm at the moment [01:30:54] awight: erp it happens, don't worry, it'll be fine [01:32:08] lemme know if I can do anything [01:33:08] AGH. [01:33:17] awight: there are no deploys scheduled right after this one, so provided the sites are open and u have time, there's no rush [01:33:38] GC apparently sent us an order ID of "GCTEST 141121003". [01:33:49] ...you may notice that most of that is not numeric. [01:34:17] I think it's a new format for expressing irrational numbers [01:35:13] Now I have to deploy things to get a clean run. [01:36:06] AndyRussG: ok thanks for checking [01:36:42] base 36/ [01:36:44] ? [01:39:23] * K4-713 facedesks [01:39:36] We were doing so well. [01:39:41] Kind of. [01:40:39] K4-713: it'll be fine... [01:41:54] I should clarify. *You* guys have been doing awesomely. [01:43:02] K4-713: you're doing great! don't facedesk, that must hurt on a stand-up desk... also pls lemme know if I can help with anything... [01:43:27] * K4-713 blinks slowly at an uncommitted hairball on local [01:43:36] AndyRussG: deployed the CentralNotice patches... [01:43:44] This complicates things. [01:43:51] awight: thanks much!! [01:43:52] * awight borrows a desk for face-ing [01:44:13] Yeah I can test banners [01:45:13] http://en.wikipedia.org/?banner=B14_1030_enUS_dsk_gwy_gc&force=1&debug=true [01:45:39] Still gotta leave the debug=true in there until the cache oodles out [01:47:48] Cache fully oodled now! [01:48:36] I'm gonna reply to James A. on centralnotice-admin list, fiddle with Mingle... [01:49:10] awight: thanks again for so much deploying effort! [01:51:31] (PS1) Katie Horn: Downgrading the severity of something that GC just did to us [wikimedia/fundraising/crm] - https://gerrit.wikimedia.org/r/175634 [01:52:14] omg [01:52:45] Who wants to play "find the actual two lines of real change in the whitespace mess?" [01:53:02] BLARGH [01:53:15] why... [01:53:35] well anyway my Gerrit diff view defaults to ignore whitespace :) [01:53:48] Oh. Well, never mind, then. [01:54:02] nice try ;) [01:54:19] Actually... you know what? [01:54:24] (CR) Awight: [C: 2] "Designed to be tampered." [wikimedia/fundraising/crm] - https://gerrit.wikimedia.org/r/175634 (owner: Katie Horn) [01:54:26] (Merged) jenkins-bot: Downgrading the severity of something that GC just did to us [wikimedia/fundraising/crm] - https://gerrit.wikimedia.org/r/175634 (owner: Katie Horn) [01:54:30] uh oh [01:54:39] I just decided that I did that for a really good reason. [01:54:46] Just kidding. [01:54:49] Ehm. [01:54:50] :p [01:54:59] I'm through with this day... [01:55:05] U need more supports? [01:55:19] And: I can't believe you're using a standing desk. Masochist. [01:55:24] Please stop that. [01:55:26] (PS1) Katie Horn: Revert "Downgrading the severity of something that GC just did to us" [wikimedia/fundraising/crm] - https://gerrit.wikimedia.org/r/175637 [01:55:34] wat [01:55:59] (CR) Awight: [C: 2] "/me keeps eye on the shells..." [wikimedia/fundraising/crm] - https://gerrit.wikimedia.org/r/175637 (owner: Katie Horn) [01:56:01] (Merged) jenkins-bot: Revert "Downgrading the severity of something that GC just did to us" [wikimedia/fundraising/crm] - https://gerrit.wikimedia.org/r/175637 (owner: Katie Horn) [01:56:12] awight: That number gets used in an exec statement. [01:56:22] oh my god [01:57:23] awight: I need to think about this for a second. [01:57:33] I should be able to bail out on that whole transaction at that point. [01:57:43] K4-713: can u update the regex to match the stupid new nonnumeric thing? [01:57:43] But it's bigger than that. [01:57:59] awight: I just want to skip everything that is unusual at this point. [01:58:20] Also: Strictly numeric is safe. [01:58:25] fair enough. What are the circumstances in which GC tries to kill us? [01:58:41] Unknown. [01:58:48] This is the first time they have done this. [01:59:08] we should send those transactions back, along with a bill for our time. [01:59:12] ...so, if I had to guess, it's got 100% of everything to do with holding the door open for our wr1 implementation. [01:59:28] I am getting really tired of being right about this kind of thing. [02:00:21] I'm pretty sure they said that they're only generating these for us, and a very small number of others. [02:02:04] why add a digit... when you can totally change the format and start using random punctuation to get more bits! [02:02:17] Also, spaces where there are usually none. [02:02:23] no. [02:02:29] Oh yes. [02:03:42] * awight checks pizzzacat's drawer. [02:03:44] Only one left [02:03:48] * awight bares teeth [02:03:56] what are you looking for awight? [02:04:04] beerz? [02:04:11] awight: If you beat me unconscious with it first, that's okay. :) [02:04:14] pizzzacat: yes hidden celebratory supplies [02:04:23] K4-713: it's on. [02:04:29] whatever happened to K4's tequila stash??? [02:04:36] *alleged [02:04:40] I had to give it back. [02:04:44] * awight checks fridge before getting all Lord of the Flies on it [02:04:45] * K4-713 looks sad [02:04:51] awwwww [02:07:11] awight: Confirmed, btw. These test transactions do not appear in the WX files. [02:07:19] I had to look. [02:14:28] The really unacceptable thing about this iTunes monopoly is that it can't decode FLACs. [02:22:27] meganhernandez: fyi, AndyRussG fixed the banner= issue, as of a few minutes ago. [02:23:10] meganhernandez: and awight courageously deployed :) [02:25:12] huh? what about the aweseome possum? [02:26:19] AndyRussG: He quit. [02:26:56] that’s fantastic, i see banners! [02:27:01] ok will prep banners for the morning test [02:27:21] AndyRussG: re that banners to logged-in people, do you think it might have been cached RL data? Or is anonymous filtering also performed in the client code? [02:27:51] meganhernandez: cool! Pls give a shout if there are any issues... [02:28:02] awight: huh? [02:28:23] did something bad happen to logged-in people? [02:29:34] AndyRussG: oh sorry, I CC'd you to a Phabricator thing, still learning how to use it. https://phabricator.wikimedia.org/T75812 [02:30:43] awight: ohh, I'm still learning I should check my phabricator spambox [02:30:50] whoooo [02:34:06] awight: I can confirm that's happenning, booo :( [02:34:20] AndyRussG: really? I couldn't reproduce it locally. What did you do? [02:34:24] Log in? [02:34:33] hmm [02:34:41] https://en.wikipedia.org/w/index.php?title=Main_Page&country=FR&reset=1 [02:35:18] AndyRussG: I don't get one [02:35:31] awight: and you're logged in? cleared cookies? [02:35:36] ... which slightly increases my belief in the theory I gave above [02:35:40] clearing now... [02:36:07] awight: check your browser's call to Special:BannerLoader to be sure [02:36:14] AndyRussG: nope I can't reproduce [02:36:28] Are you sure you aren't clearing your logged-in cookie? :) [02:36:32] I'm getting it if I log in with an incognito / private browsing session [02:36:44] ejegg: weird [02:36:45] seriously. how are you logged in w/o a cookie [02:37:03] awight: clear cookies, then log in? [02:37:22] Yeah, cookies persist within the window in the private browsing session in FF [02:39:33] weirdly, I don't get the banner. But I do see a bunch of junk in the RL response. [02:40:02] I think the RL data response is cached, and does not fragment across logged-in [02:40:29] well that would really suck [02:40:34] pardon my language [02:41:46] awight: are you getting the banners if you log out? Does it vary for you on logged in/logged out? [02:42:05] RL must vary on user because it sends user data in to the client [02:42:47] AndyRussG: yeah I see banners if logged out [02:43:04] And then you log in and they go away like they're supposed? [02:43:08] AndyRussG: so you think the RL data module class might be getting the wrong logged-in-ness? [02:43:18] AndyRussG: yeah. however, I see them in the RL data response [02:43:38] awight: that's reall weird. [02:43:50] No, we're not filtering for logged-in-ness on the client since that's supposedly done on the server [02:43:51] What do you get for https://bits.wikimedia.org/en.wikipedia.org/load.php?debug=false&lang=en&modules=ext.centralNotice.bannerChoiceData%2CbannerController%7Cext.centralNotice.bannerController.lib%7Cext.centralauth.centralautologin.clearcookie%7Cext.eventLogging%7Cext.uls.init%2Cinterface%2Cpreferences%2Cwebfonts%7Cext.visualEditor.track%2Cve%7Cext.visualEditor.viewPageTarget.init%7Cjquery.accessKeyLabel%2CbyteLength%2Cclient%2Ccookie%2CmwExten [02:44:13] So something more wierd must be going on for you not to see them, can u debug? might be another issue to track down... [02:44:26] I was pretty sure I checked that we were getting the user correctly [02:44:31] How did we expect that to work... [02:45:11] If my theory is correct, the only fix is to send both anon and logged-in banners in the RL data response, and the columns showAnon and showUser... [02:45:15] then filter in the client [02:45:49] Yea that would also suck [02:45:59] though not sooo terribly I suppose [02:46:04] grrr [02:47:16] AndyRussG: just realized there is a smaller but terrifying chance that the RL response will get cached in response to a logged-in user [02:47:24] in which case, many many people will not get banners [02:47:26] urp [02:47:33] I'm having a problem pushing my day's patch to gerrit [02:47:43] :( [02:47:43] AndyRussG: what are your work hours tonight :) ? [02:47:56] Let's summon Krinkle............ [02:48:29] AndyRussG: we're talking here about turning the client-side switch off again [02:48:30] Krinkle: boo! Have we done something terrible that doesn't properly distinguish logged-ins w/ anons? [02:49:06] awight: gotta do watcha gotta do [02:50:02] awight: also pinging Roan on -dev [02:50:25] AndyRussG: I'm happy to wait until we get some RL heads to think about this [02:50:30] anyone know what this means? ! [remote rejected] HEAD -> refs/publish/master/BigEnglishModules (change 173632 closed) [02:50:39] but I don't see how the RL request could possibly distinguish between logged-in and anon [02:50:56] or, working the other way (pushing without using gerrit) ! [remote rejected] HEAD -> refs/for/master (change 173632 closed) [02:51:04] pizzzacat: that it's been merged or abandoned already. [02:51:20] the patch I'm trying to push? [02:51:21] pizzzacat: in this case, merged :) [02:51:29] yeah [02:51:38] how is that possible, since I haven't pushed it yet? *scratches head* [02:51:51] pizzzacat: https://gerrit.wikimedia.org/r/#/c/173632 ? [02:52:04] That's the one it's complaining about. is that the head? [02:52:08] it's after this [02:52:26] ok. then you have an amended or rebased version of ^^ preceding the patch you're pushing. [02:52:32] rebase your new work [02:52:35] so I sometimes get this: [02:52:51] well I won't paste [02:52:54] AndyRussG: If I'm right about the non-fragmentation thing, we definitely have to turn the client choice off until fixed [02:52:56] basically a few commits [02:53:03] and it asks if I want to push them [02:53:14] awight: yeah, roan is looking now [02:53:18] AndyRussG: ok great [02:53:25] and I told this to someone and they said I could just push them and it ignores the ones I've already done [02:53:27] meganhernandez is here too, we're trying to triage [02:54:59] what am I supposed to be rebasing awight?] [02:55:40] this is what I was trying to avoid earlier. :-/ [02:56:07] seems like you guys are busy so I can just deal with it in the morning but it sucks having spent all day and not pushing it. :( [03:06:35] AndyRussG: fyi I've opened a deployment window, I'm turning off client-side choice until we can fix. It's better this way... [03:06:47] Yep. [03:06:56] * K4-713 looks stern [03:07:00] :) [03:07:42] awight: Yeah right way 2 go [03:08:02] I'll put in a comment on the phabricator report [03:09:15] We can differentiate anon/logged-in on the client, and maybe if it's decided that RL could be patched to distinguish, then we could switch back to filtering on the server [03:09:16] Bah [03:09:39] eh I think it's a minor change--there isn't much additional payload, if that's what you're thinking about [03:09:44] awight: as before, lemme know if I can do anything [03:09:58] Yeah, it's not terrible. But also too much to do in 5 minutes or less [03:10:15] I'll work on the patches for fixing and we can see if we turn it on again tomorrrow? [03:10:20] definitely! [03:10:31] K :) thanks BTW [03:12:06] AndyRussG: ok the switch is off... [03:12:17] I'll wait around for a few, then going home. [03:12:37] awight: K thank U... I'll check and ping when it's all cached out [03:16:33] awight: before u go---any news on the reason why you weren't getting the banners? (if not we can investigate tomorrow) [03:17:43] awight: K enwiki is back to Special:BannerRandom, thanks once again [03:18:57] meganhernandez: as per above, we ran into a problem distinguishing between logged-in and anonymous users with the new system, and logged-in users were suddenly getting fundraising banners. We've turned the new system off and will see if we can fix it tomorrow. [03:19:34] AndyRussG: Hi [03:19:38] ok thank you for the update AndyRussG [03:19:46] awight: [03:19:49] What's up? [03:20:05] i’ll still run a test in the morning, but i don’t think you need to monitor since the new central notice changes are not out [03:20:18] meganhernandez: OK sounds good... we'll keep everyone updated... [03:20:29] thank you! [03:20:46] np! likewise, especially thanks 4 being patient... :) [03:20:51] Hi Krinkle, uhh, we realized that we were mistaken in thinking the RL module could distinguish logged-in vs anon status [03:20:52] AndyRussG: fwiw I won't be around tomorrow, but ejegg|away can do the config deploy to turn it back on. [03:20:59] modules are shared by wiki(dbname), skin and user language (e.g. french user with monobook on enwiki) [03:21:10] awight: yeah! ejegg|away will do great! [03:21:18] modules are otherwise stateless. [03:21:28] that confirms our fear. [03:21:33] ok thx, gotta run! [03:21:39] awight: bye, take care! :) [03:22:21] AndyRussG: How big is the banner load if you were to do that on the client side as well? Like all the other factors you're doing client side now as well (cause you're sending them all to the client right and selecting there instead of server-side) [03:22:23] Krinkle: yeah we just learned about the magic by which a user ID for logged-ins is sent with the page HTML [03:22:57] Krinkle: yeah we can do it, it's just a bit of refactoring, here there and everywhere [03:23:35] plenty of scripts are logged-in-users-only (e.g. the load() call only occurrs for logged in users, this can be done server side from an OutputPage-accessible hook using User::isAnon() and calling addModules conditionally) [03:23:38] So we turned off the feature, which otherwise was performing quite nicely, and will see if we can get it up and running again tomorrow... [03:23:52] but the source code of a module itself, once loaded, can't vary on user. [03:24:23] AndyRussG: OK. [03:24:43] Krinkle: hmmm, wait a minute... [03:24:52] AndyRussG: If the size of the banner pool is significantly larger when selecting both anons and lius, let me know. There may be a few tricks we can try. [03:25:07] But.. tomorrow. [03:25:08] Uhh yeah no it's not really that big [03:25:44] Krinkle: based on what u said, it occurs to me then that we could add _different_ modules for logged-in vs anon users [03:25:49] https://github.com/wikimedia/mediawiki-extensions-CentralNotice/blob/master/includes/CNBannerChoiceDataResourceLoaderModule.php#L37 [03:25:54] I should've seen that as a red flag [03:26:11] Eh np :) yeah that's where we borked [03:26:19] it using wgUser is strange though [03:26:29] as if that's somehow different from context->getUser; it isn't [03:26:48] Well.. I didn't see user params coming in on the RL URLs [03:27:19] Indeed. That's done for legacy user scripts modules. [03:27:36] Hmmm [03:27:40] But you don't wanna do that, since you'd throw away your client-cache almost completly, and explode server-side cache. [03:27:59] right [03:27:59] as it uses username [03:28:04] yeah [03:28:32]