[11:46:14] hi everyone, my name is John and I am the sysop of Commerce Network Nightly Build wiki [11:46:37] I have a problem in my wiki [15:33:41] ChristianWikia: ping [15:33:43] Inez_: ping [17:15:00] 20% talk time [17:15:15] Oh hey [17:15:22] I'm doing a second 20% this week... ish [17:15:25] After I deploy AFTv5 [17:15:45] Cause there's some git switchover prep work left to do, like reverting ~120 extension revisions [17:15:51] hexmode: looking at https://www.mediawiki.org/wiki/Wikimedia_engineering_20%25_policy it's Ian Baker, talking with Gabriel re tomorrow, RoanKattouw, and a teensy bit Nikerabbit & Tim? [17:16:02] RoanKattouw: that reversion work is IMO the most urgent of course [17:16:08] hexmode: what do you think? [17:16:31] RoanKattouw: so many? [17:16:57] <^demon> The 7 outstanding fixmes on phase3 also need poking by someone [17:17:01] <^demon> To either revert/fix [17:17:18] Yeah that too [17:17:21] RoanKattouw: want to work on js issues for mlwiki? [17:17:35] RoanKattouw: it's being deployed to production? [17:17:39] Nikerabbit: getting to zero unreviewed/zero fixme. we hadnt' been harsh enough about enforcing the code slush, so Roan was harsh for us :) [17:17:59] hexmode: Ah, no? git switchover? [17:18:00] \o/ roan! [17:18:13] Thehelpfulone: AFTv5 has been deployed for a long time, I'm just doing a bugfix update [17:18:23] RoanKattouw: sorry, forgot that was your thing today :) [17:18:29] oh I'm thinking of something else, yeah of course I've seen it [17:19:32] <^demon> hexmode: If you're looking for something to do or put on someone's plate--there's 25 outstanding fixmes (7 for phase3) [17:19:43] Yeah, those fixmes need dealing with [17:19:51] Revert if disentangle-able, fix up if not [17:19:59] Roan, can we test on en.wiki? [17:20:16] yoni_omniti: You can test on test.wp.o in a few minutes [17:20:23] ^demon: RoanKattouw: just revert? [17:20:23] hi raindrift [17:20:26] oh ok, thanks [17:20:41] hexmode: If possible that's preferred. Or you can triage them and I can do the reverting, I'm mass-reverting other stuff anyway [17:20:59] <^demon> hexmode: Well, case by case. But I'd encourage reversion rather than going down rabbit holes :) [17:21:07] pre-branch fixmes are less urgent [17:21:13] Yes, case by case [17:21:16] And that too [17:21:22] We might also decide to just turn a blind eye [17:21:23] <^demon> As are non-deployed extensions, since they're not blocking migration [17:21:26] esp if it's pre-branch [17:21:35] Yes, only core and WMF-deployed extensions [17:21:40] I am completely ignoring non-deployed things [17:22:05] hmm [17:22:35] where are we at with the postgresql stuff? [17:23:06] that's a pretty big chunk of the fixmes [17:23:30] I would like to just revert all of saper's Postgres saga [17:23:45] RoanKattouw: does it have a negative impact on MySQL? [17:23:55] (potentially?) [17:23:55] AFAIK it has zero impact on production [17:24:10] But I think we should assign Aaron or Sam and let them be the judge of that [17:24:23] Hello everyone! Look forward to working with RoanKattouw and OmniTI this morning, to deploy new AFT5 features to testing, then production. [17:25:29] Reedy: ^ see Roan's comment [17:26:12] RoanKattouw: I'm not so much worried about impact on production as I am impact on people using the installer on top of MySQL [17:26:19] I've no idea with regards to the postgres stuff [17:26:25] (or sqlite, or other not postgresql database) [17:26:37] Presumably it should be self contained enough not to cause problems for othe stuff, but I've not reviewed any of the commits [17:26:39] Today, we aim to release new AFT5 Feedback links (Options A and E, with an 'X' close button for logged-in users), as outlined here: http://www.mediawiki.org/wiki/Article_feedback/Version_5/Feature_Requirements#Feedback_links_on_article_pages [17:26:59] it looks like Saper does modify the base class at least a little bit [17:27:56] saper has said he won't have time to really finish that up right now, so go ahead and back it out [17:28:15] OK [17:28:23] It may require a bit of surgery [17:29:02] I think there is some good stuff there, but it kind of worked w/o it so it isn't a crippling loss [17:29:10] get this patient to the OR, stat! we need fluids? what are the vitals?! [17:29:20] It'd just be a temporary revert anyway [17:29:22] heheh [17:34:51] We also plan to deploy today a change to the central Article Feedback Activity Log, as well as the Supression log. The Suppression log will only display oversight-related events to oversighters and staff, as specified in our feature requirements page: http://www.mediawiki.org/wiki/Article_feedback/Version_5/Feature_Requirements#Related_Features [17:35:29] fabriceflorin, yoni_omniti: Alright, we should be live on test.wp.o now [17:35:51] <^demon> RoanKattouw: Thoughts on http://www.mediawiki.org/w/index.php?title=Special:Code/MediaWiki&path=%2Ftrunk%2Fextensions%2FRSS&status=new ? We use it on wmf.org. [17:36:26] ^demon: I'm gonna revert all new revs. My to-revert list is tag/gerritmigration [17:36:34] <^demon> *nod* [17:36:37] (sans revs with status==reverted) [17:36:45] And that's also where you should put anything you revert [17:36:53] So they can be resubmitted into Gerrit tomorrow [17:37:23] <^demon> Just new/fixme with that tag will be reverted right? [17:37:26] <^demon> Or ok too? [17:37:29] ok too [17:37:39] That's because they're dependent on a new/fixme rev [17:39:09] OK, I'm gonna start processing the extensions revert queue [17:40:26] Lastly, we aim to release a few AFT5 bug fixes, as well as new click tracking for the feedback link options A and E, and AFT4 bucketing changes. [End of update]. [17:40:41] fabriceflorin: Are you guys testing yet? [17:41:41] ^demon: Actually some of those OK/resolved revs will get reprieves because they've been reviewed overnight, I'll have to reassess those [17:41:56] Anyway, working on that now until the AFTv5 folks finish testing [17:42:01] <^demon> Yeah I reviewed some late last night and this morning before you were around. [17:42:43] RoanKattow, sorry I missed you note that we are on testing now. Going to it right now. yoni_omniti, auroraeosrose and chrismcmahon, have you started testing there already? I will send a URL to that testing site now. [17:42:58] i have [17:43:10] No worries, I have plenty of work to keep myself busy [17:43:34] I'm on test.wm.o , not seeing AFT for random pages? [17:44:38] s/wm/wp/ [17:45:06] chrismcmahon: PM with fabriceflorin and he'll enlighten you [17:45:15] thanks RoanKattouw [17:45:58] roan, could there be a different configuration value for $wgArticleFeedbackv5LinkBuckets in test.wp? [17:46:06] i can't get link E to show up [17:46:10] Probably not different from prod [17:46:15] But I didn't change it [17:46:18] Let's see what it is [17:46:18] i mean different form trunk [17:46:34] should have -,A and E set to 33 each [17:46:55] Hi RoanKattow, I cannot see Option E either. I am going to this test link: http://test.wikipedia.org/wiki/Golden-crowned_Sparrow?bucket=1&debug=true&aftv5_link=E [17:47:11] yoni_omniti: http://pastebin.com/6rhLmhEN [17:47:22] Ooh wait [17:47:26] That var is exported to JS [17:47:30] * RoanKattouw purges config cache [17:47:38] looks legit... [17:47:50] <^demon> hashar: Running `git fsck` now but otherwise done rewriting history. Stepping to the store for a few minutes while fsck does its thing. [17:48:06] ^demon: ok [17:48:11] ^demon: have you fixed empty messages? [17:48:18] <^demon> Nope. [17:48:28] ^demon: so just extraparsertests and ogg files right? [17:48:43] <^demon> No I hadn't done the rebase yet, I was doing the filter-branches on extensions [17:48:46] <^demon> *That* is what just finished [17:49:13] ^demon: I uploaded my notes about rebase somewhere; [17:49:43] ^demon: will be back in 2 hours. give me a call if needed [17:49:53] <^demon> *nod* [17:50:05] yoni_omniti: Try now with a hard refresh? [17:50:28] trying... [17:51:09] nope [17:51:27] me neither fwiw [17:51:43] Not sure if this helps, but I noticed that the Article Feeback Category had a red link, so I added the same info to this article category page as on the english Wikipedia, in case this helps enable the AFT5 feedback links. http://test.wikipedia.org/wiki/Category:Article_Feedback_5 [17:56:36] When I go to the test page for AFT feedback link option A after emptying my caches, I still do not see the Option A text link below the article title: http://test.wikipedia.org/wiki/Golden-crowned_Sparrow?bucket=1&debug=true&aftv5_link=A [17:57:00] Hmm [17:57:13] Maybe the OmniTI devs that wrote that code should look into that [17:57:24] I can tell you that wgArticleFeedbackv5LinkBuckets is exported correctly [17:57:24] ^demon: RoanKattouw : this needs to be updated for the git migration http://wikitech.wikimedia.org/view/How_to_deploy_code [17:57:35] tfinc: To some degree yes [17:57:43] thats what my engineers look at to make sure to know about updated changes [17:57:44] tfinc: We're not actually migrating the deployment branch to git today [17:57:46] awjr: --^ [17:58:03] good. so before we do lets update the doc [17:58:04] s [17:58:14] ^demon: I'm gonna do my mass revert on Ext:RSS now [17:58:22] Yes, we'll change the docs when things change [17:58:26] … but I see that there is a funny URL below the article title, where the Wikipedia slogan and 'Improve this article' link should be: http://test.wikipedia.org/wiki/Special:ShortUrl/m1 [17:58:28] And notify private-l [17:58:31] Only step 1 needs changing really [17:58:45] fabriceflorin: That's because we're testing the ShortURL ext on that wiki, you can ignore that [17:58:54] oh, and svn up -> git pull [17:59:33] No [17:59:39] 1.19wmf1 will remain in SVN for the time being [17:59:43] might the two extensions be conflicting? [17:59:54] auroraeosrose: Maybe... I can disable ShortURL and we'll see [17:59:55] Thanks, RoanKattouw - yoni_omniti and auroraeosrose, any ideas why we're not seeing the feedback links on testingg? [18:00:05] RoanKattouw: just verified that it works on both prototype and my local [18:00:19] what revision is deployed? [18:00:50] trunk [18:00:57] I literally copied trunk over [18:01:38] <^demon> tfinc: It won't change too much, but yeah :) [18:01:54] mw.config.get( 'wgArticleFeedbackv5LinkBuckets' ).buckets on the test page gives me the right values.... [18:02:10] ^demon: RoanKattouw : sure. but whatever you choose needs to be mailed out to the other deployment engineers [18:02:13] so that were all clear [18:02:21] Yes [18:02:23] even if its just a couple of weeks [18:02:30] thats a huge amount of time within the mobile team [18:02:35] i've given robla a heads up as well [18:02:38] <^demon> "No more deploys, everrrrrr!" [18:02:43] haha [18:02:49] right .. not going to happen ;) [18:03:26] <^demon> Well the basic premise is we'll continue to have a wmf branch, so once changes are merged there, you just `git pull && git submodule update` on fenari. [18:03:29] <^demon> Then scap as usual [18:04:01] Yes, but that's the situation as it will be a few weeks from now, not tomorrow [18:04:32] Roan, i see that configuration and code are the same across the environments, however it does not work on test.wp, but works on others. any ideas what may be different on test.wp that causes that? [18:04:38] I have no idea [18:04:47] I suggest the people that wrote the code in question debug it [18:05:07] ^demon: makes sense but lets mail out about it. not everyone is paying attention to this conversation [18:05:09] Roan, the same code works on other envronments [18:05:19] robla is going to help draft that [18:05:41] Still, your devs are better able to debug than I am [18:06:15] Even if the environment is to blame, debugging it should turn up what part of the env is causing it [18:06:34] I agree, and we are debugging - but we can't debug on test.wp, and it behaves correctly where we are able to debug [18:06:57] How is it that you can't debug on test? [18:07:16] Have you tried ?debug=true and using a debugger? [18:07:23] yes [18:07:30] i am in debug=true [18:10:31] RoanKattouw and yoni_omniti, let's put our heads together to solve this problem and figure out what might be causing it. Thinking of previous deployments, do we remember experiencing any issue like this? Could it be possibly related to the bucketing system? or cookies? or some special category that would need to be enabled? [18:11:17] On a more positive note, I am pleased to report that I just tested the central activity log on testing, and it works as intended, separating the oversight and non-oversight events between it and the suppression log: https://test.wikipedia.org/w/index.php?title=Special%3ALog&type=articlefeedbackv5 [18:11:24] i think i know what the problem is… tneed a few minutes to verify [18:11:31] <^demon> Heh, git fsck goes through commits alphabetically rather than sequentially. I guess that's faster :) [18:11:35] I would normally help you, but today I simply don't have time to debug someone else's code, sorry about that [18:11:51] <^demon> A fsck doesn't care about history, so it just pulls them off disk in order probably. [18:14:24] ok [18:14:43] Roan, in jquery.articlefeedbackv5.js that is being served, the function selectTriggerLinks does not exist [18:14:48] server caching? [18:15:19] <^demon> RoanKattouw: Did your gerrit ui project get a public address yet? [18:16:04] ^demon: gerrit-dev.wmflabs.org [18:16:11] yoni_omniti: Oh, hah [18:16:20] yoni_omniti: Let me purge some more caches here [18:16:35] thanks! [18:16:41] done, try now [18:17:11] <^demon> Login doesn't work? [18:17:16] ^demon: Use 'become' [18:17:21] It's in dev mode [18:17:38] auroraeosrose: Do you need https://www.mediawiki.org/wiki/Special:Code/MediaWiki/114386 to be included in this deployment? [18:17:39] <^demon> Ahh, I'll have to make an account [18:18:07] <^demon> Will do later, was just curious. [18:18:10] <^demon> I like the colors so far. [18:18:13] I have to rip out some of the JS there [18:19:48] roan, i'm still getting the wrong file :( [18:19:57] That's strange, lemme investigate [18:20:40] RoanKattouw: not sure - without it the activity log doesn't pick up the suppress items for oversighters in the feedback special page activity log [18:21:13] That's all Chinese to me, I'll defer to fabriceflorin [18:21:29] hehe - yeah, it's probably fabriceflorin's call [18:21:31] RoanKattouw: do the dutch actually use that line? [18:21:36] because we, of course, use dutch :P [18:21:39] At Roan's suggestion, I just tested both feedback links A and E on Chrome's 'incognito' mode (click Command-Shit-N to do that). Roan thought that might be a good way to see if this is related to bucketing cookies. Neither A nor E show up in that 'incognito' mode. http://test.wikipedia.org/wiki/Golden-crowned_Sparrow?bucket=1&debug=true&aftv5_link=E [18:22:31] fabriceflorin: yeah we think we found the root of the problem, roan is investigating [18:22:44] selectTriggerLinks appears in my debug=false response at least [18:22:57] $.articleFeedbackv5.find( '#articleFeedbackv5-find-feedback' ).elastic(); [18:23:00] Thanks, yoni_omniti, keep us posted. [18:23:02] is not a function [18:23:04] Oh hey [18:23:18] didn't elastic get taken out? [18:23:29] yeah but its irrelevant its a function in an object that is never used (deprecated bucket) [18:23:34] should be removed in future release [18:24:08] wait, wtf [18:24:10] js execution should not stop b/c of it - unless we're on IE6 [18:24:13] Yeah that doesn't appear in trunk [18:24:33] missing some commits? [18:24:40] no way [18:24:45] let me double-check [18:24:52] No, it doesn't appear in the latest version [18:24:59] It was caching [18:25:04] RoanKattouw, what auroraeosrose is referring to is that we are now placing oversight-related events in the Supression log, as discussed in our meeting with you on Monday. This is an important request, which both Oliver Keyes and Philippe Beaudette feel strongly should be implemented ASAP. Can you help us deploy that feature today, please? [18:25:18] OK now when I visit https://test.wikipedia.org/wiki/Golden-crowned_Sparrow?bucket=1&debug=true&aftv5_link=E the console says "Using link option E" [18:25:19] so the answer RoanKattouw is yes ;) [18:25:23] fabriceflorin: Weren't we already doing that? [18:25:28] OK I'll include that rev then [18:25:32] to be sure - i am still working with SVN until EOD today? [18:25:33] and the one after ;) [18:25:42] (I typo'd... sigh) [18:26:03] "Using link option E" is exactly what is should be [18:26:27] https://www.mediawiki.org/wiki/Special:Code/MediaWiki/114387 too or it'll break [18:26:41] works for me now [18:26:43] Yeah got it [18:26:46] sweet [18:26:57] RoanKattouw, up until now all AFT5 events were only appearing in the Article Feedback Activity Log -- today, we are moving oversight-related events to the Suppression log, but keeping all other events in the AFT log: https://test.wikipedia.org/w/index.php?title=Special%3ALog&type=articlefeedbackv5 [18:27:42] Roan, thanks! I'm happy we resolved that. I verified both A and E to work properly. [18:27:43] No, this is another change, this makes one of the views aggregate stuff from the AFT5 and suppression logs if the usr has perms [18:27:46] yay [18:27:52] AFK 10 minutes [18:28:06] fabriceflorin: yeah, that was a fix for the activity log stuff [18:28:39] on the feedback page - we moved the stuff into suppress log and suddenly it wasn't showing up there ;) I had to change the query so if you have permissions you can see all activity including the suppress stuff [18:28:46] Success! I can also see the Option E and Option A fine now. What did you guys do to fix it? http://test.wikipedia.org/wiki/Golden-crowned_Sparrow?bucket=1&debug=true&aftv5_link=E [18:29:05] Just purging stuff [18:29:07] It was caching [18:29:36] OK auroraeosrose 's fix should be in too [18:29:59] RoanKattouw, was it caching on the front-end or the back-end? Just trying to figure out what went wrong, so we avoid that in the future. [18:29:59] I can't check that one - can you fabriceflorin? [18:30:32] Probably Squid, probably test.wp.o-specific [18:31:39] auroraeosrose, I can confirm that recent oversight options are now going into the suppression log, which is only available to oversighters and some staff members (Philippe gave me temporary staff rights). Is that what you wanted me to check? [18:31:53] can you go to the feedback page [18:31:57] and click on the activity link [18:32:09] and make sure the suppress stuff IS showing up there if you're an oversighter? [18:32:16] I can verify it's NOT if you're not [18:33:36] auroraeosrose, I can confirm that the article activity log on the feedback page correctly displays oversigher events. Thanks so much for fitting in this last minute request! [18:34:07] This will make Ironholds very happy ;o) [18:34:18] heh, I had forgotten that the suppress change would break stuff there too - sigh [18:34:22] no problem [18:34:38] only oversighters will get full activity though [18:34:46] auroraeosrose: you drink scotch? [18:34:47] not sure how you want to handle the count totals for that... [18:35:04] scotch is good ;) tequila is better! [18:35:51] should you come to Wikimania in DC (I do not know why, maybe you are bored at the time!) or make it to SF at the same time I do, tell me and I will buy you the most glorious bottle of single malt known to humanity [18:36:10] HAHA, all righty [18:36:41] <^demon> RoanKattouw: Oh, you might also want to hop on http://etherpad.wikimedia.org/Git-migration. Kind of a running log of what Antoine/I have been doing [18:36:47] OK [18:36:59] Ironholds is the Santa Claus of good scotch [18:37:28] I am [18:37:38] I just turn up on alotted dates and dispense it to everyone who has been good [18:37:45] people who have been bad get blended horsepiss [18:37:53] I believe you guys call it bourbon [18:37:55] Hehe Ironholds ;o) [18:39:27] So from my standpoint, the oversighter / suppression log functions work as requested, and as summarized in my last report to Philippe and Geoff. [18:40:02] are we good to deploy on en.wp? [18:40:14] I'm good to go if you are [18:40:21] fabrice? [18:40:39] are the counts correct? this seems off: https://test.wikipedia.org/wiki/Special:ArticleFeedbackv5/Golden-crowned_Sparrow/34 [18:42:57] 20 feedback posts, 16 accounted for in Showing: selectbox, 1 displayed. [18:44:47] I don't know if the sql script to fix the counts was ever run there - they were all out of whack for awhile [18:45:16] Probably not [18:45:18] Link? [18:46:19] I'm fine with it if y'all don't think it's a big risk. [18:47:42] Roan and auroraosrose, the next thing I would like to do today is to change the variable that sends the email address for oversight requests to oversight-en-wp@wikipedia.org (instead of aft@wikimedia.org), as documented on this Bugzilla ticket: https://bugzilla.wikimedia.org/show_bug.cgi?id=35020 -- would that be possible? [18:48:18] I think we are good to deploy to en-wiki. thanks for taking the time to test all this! [18:48:28] For sure [18:48:39] Mind if we do that after the code update? It's a trivial config change but let's separate them [18:48:50] <^demon> Sweet. Got the repo down to 355678 objects and only 100M. I win \o/ [18:48:54] yay [18:49:01] <^demon> And I haven't even rebased out ExtraParserTests yet. [18:49:10] ^demon: There are more reverts coming that you'll have to import as well, mind [18:49:13] RoanKattouw, I agree we should do that after the code update, but wanted to mention it now, in case you wanted to do it before. [18:49:18] fabriceflorin: OK deploying to enwiki now [18:49:32] <^demon> They're all post-history-rewrite so it'll just be a simple import && repack [18:49:36] <^demon> No need to re-filter. [18:49:41] Deploy running, this'll take a while [18:49:43] * AaronSchulz can't wait till stuff is switched over [18:49:44] OK, excellent [18:56:04] Thanks, RoanKattouw ! I can now see the Option A button on En-Wiki, but it is missing the '|' pipe divider before "Improve this page" which Brandon / Jorm requested yesterday, and which OmniTI implemented yesterday. Did we not include that code revision? http://en.wikipedia.org/wiki/Golden-crowned_Sparrow?bucket=1&debug=true&aftv5_link=A [18:56:38] I have no idea, ask OmniTI which rev itwas [18:57:11] lemme check [18:57:30] 114271 [18:58:04] !r 114271 [18:58:04] https://www.mediawiki.org/wiki/Special:Code/MediaWiki/114271 [18:58:04] https://www.mediawiki.org/wiki/Special:Code/MediaWiki/114271 [18:58:10] grr [18:58:29] lol [18:58:44] Yeah that should've made it in [18:58:55] LOL [18:59:24] fabriceflorin: Also note that the deployment isn't done yet, enwiki is probably in a limbo state right nwo [19:01:16] Got it. That explains why I am still not seeing the new dark blue feedback button Option E on production. Should I clear my cache? http://en.wikipedia.org/wiki/Golden-crowned_Sparrow?bucket=1&debug=true&aftv5_link=E [19:01:37] No you should wait until the deployment is finisihed [19:01:39] It's still running [19:02:05] <^demon> RoanKattouw: I've marked the pgsql stuff resolved. It looks fine to me (although I've not personally tested), and the phpunit-postgres builds are all passing. [19:02:15] OK, thanks [19:02:25] If any of them were tagged gerritmigration, please untag [19:02:49] BTW big kudos to the i18n team [19:03:00] There were like 30 of their revs that were on my to-revert list, and they were all reviewed overnight [19:04:25] <^demon> Ok, out of 5 fixmes on phase3, 3 are CSS/JS/RL related. [19:04:31] <^demon> Last one is the new log entry thing. [19:06:04] Yeah I need to look at the RL ones [19:06:20] That's probably just gonna be a question of migrating the remaining breakage to a bug report [19:06:29] fabriceflorin: OK, deploy done [19:06:36] ^demon: what is the correct path for gerrit.wikimedia.org:29418/test/mediawiki/extensions/MobileFrontend.git [19:06:42] If there's any remaining issues after purging your cache, poke me [19:06:57] <^demon> preilly: No test/ [19:07:34] ^demon: ah [19:07:52] <^demon> If we need to write out that library from the history, the time to do it is *now* [19:08:00] <^demon> Once we've migrated, changing history is a big no-no. [19:08:31] ^demon: what library? [19:08:38] <^demon> WURFL [19:08:49] ^demon: is it NOT in the git repo? [19:08:58] <^demon> No, it already is. [19:09:05] <^demon> Because I've been mirroring extensions for the past ~2 weeks [19:09:16] <^demon> And I just heard about this issue ~24 hours ago [19:10:19] I can now see Option E on production, but it only says "Improve this" on Chrome, not "Improve this page" - Yoni_omniti, I thought you guys were using a graphic for this, rather than plain text. Do you want me to send you a screenshot? I cannot see it on Firefox or Chrome, even after I empty my cache for these browsers. I am using this URL, is it correct? http://en.wikipedia.org/wiki/Golden-crowned_Sparrow?bucket=1&debug=true&aftv5_link=E [19:10:50] ^demon: so, when do we stop using SVN? [19:10:53] <^demon> preilly: So once we've flipped the switch and people start cloning, it's too late to rewrite the history. So the time to do it is now when we're soft-launched and few clones exist. [19:11:00] <^demon> Tonight, weather permitting. [19:11:17] ^demon: so, should I remove WURFL in SVN or git? [19:11:43] <^demon> Go ahead and remove it in SVN. I'll take care of rewriting the history in git. [19:11:52] ^demon: okay [19:12:00] fabrice, testing in production [19:12:16] and, no, were not using a grahpci [19:12:20] *graphic [19:12:30] <^demon> RoanKattouw: Patrick's going to remove WURFL from MobileFrontend before we migrate. That's ok to allow through. [19:12:41] OK [19:12:55] I should have brought my hammer into the office today [19:13:03] Bang on my desk every time I revert stuff [19:13:36] this is why i'm doing svg and video tests today. ;) no revert worries ;)) [19:13:42] I can also see Option A on all browsers, but the little "X" icon does not show up when I am logged in. It is correct for it to not show up for logged out state, but should appear in logged in state. http://en.wikipedia.org/wiki/Golden-crowned_Sparrow?bucket=1&debug=true&aftv5_link=A [19:14:18] options E appears and behaves correctly for me across safari, ff and chrome [19:14:38] option A too [19:14:46] fabrice, clear cache in your browser [19:15:29] I emptied my cache on all browsers. I am going to shut down my browsers and restart them, in case they are causing the problem (I have way too many windows open). Stand by. [19:15:50] restarting the browsers should not have any effect [19:20:27] fabriceflorin: for option A, when you mouseover the "improve this page" the X appears for me [19:20:50] but not until mouseover [19:21:01] for option E it's working fine for me... [19:21:07] OK, I shut down all my browsers, re-emptied my cache on all 3, and I am still not seeing Option E at all, even on Chrome -- and Option A is still not displaying the 'X' close box in logged in mode. Is this the correct URL? http://en.wikipedia.org/wiki/Golden-crowned_Sparrow?bucket=1&debug=true&aftv5_link=E [19:21:40] that is working fine for me [19:21:43] ... weird [19:21:43] Maybe it's just me. chrismcmahon, what are you getting on your end?Should I restart my computer? [19:21:50] ^demon: see https://www.mediawiki.org/wiki/Special:Code/MediaWiki/114393 [19:22:18] it's working for me in all browsers here on windows [19:22:43] how much the GIT ETA? [19:24:29] fabriceflorin: this http://en.wikipedia.org/wiki/Golden-crowned_Sparrow?bucket=1&debug=true&aftv5_link=A and this http://en.wikipedia.org/wiki/Golden-crowned_Sparrow?bucket=1&debug=true&aftv5_link=E show me exactly the same AFT box, the one with the box to enter text [19:25:05] I am on Mac. It's working fine for Heather, who is also on a Mac, but it only started working for her a few minutes ago. [19:25:17] caching…. [19:25:21] tough to get around it [19:25:45] ^demon: do we NOT have a '.gitreview' file found in the core repository? [19:26:21] God did it disappear AGAIN [19:26:29] I must've put in that .gitreview file at least twice [19:26:29] <^demon> mediawiki/core.git hasn't had one [19:26:33] <^demon> I keep rewriting history [19:26:36] <^demon> So don't add it :) [19:26:39] OK [19:26:47] You add it when you're done then [19:26:50] <^demon> Yeah [19:26:54] Just steal the one from the puppet repo and tweak it [19:27:20] Is there a better solution than emptying our cache? None of this is working for me properly right now, despite repeated attempts to clear my cache. Should I restart my entire computer? [19:28:37] I think we're seeing server-side caching. [19:28:43] not browser caching [19:28:54] Possibly [19:28:55] yup, sounds like it [19:29:00] What's broken? [19:29:02] have a proxy somewhere? [19:29:08] "a proxy"? Yeah we have lots [19:29:36] hehehe [19:31:47] chrismcmahon: I see the same boxes as well, though only the second link shows an 'improve this page' box [19:34:59] OK, we found the cause, purging the right caches now [19:35:08] &aftv5_link= is only respected in debug mode [19:35:19] However, AFTv5 and ResourceLoader have differing opinions about what debug mode is [19:35:38] So if you request &aftv5_link=E&debug=true , the link option is respected but you're being served stale JS [19:35:52] If you just set &aftv5_link=E , you get up-to-date JS but the param is ignored [19:36:00] nice [19:36:08] What did work is &aftv5_link=E&debug=false , because AFTv5 doesn't recognize 'false' correctly :D [19:36:24] To clarify: Option A is supposed to display a small text link below the article title, while Option E is supposed to display a larger, dark blue button at the bottom right of the window. I just spoke to Roan, who could reproduce the problem that Chrismcmahon reported. He thinks that he needs to purge more caches on the server side. In the meantime, he thinks that one third of the users are getting the correct function, and instructed me to [19:36:24] keep trying to join that one-third by using the 'incognito' mode in Chrome (Cmd-shift-N). I will do this now and report back. [19:36:28] <^demon> preilly: https://gerrit.wikimedia.org/r/gitweb?p=mediawiki/extensions/MobileFrontend.git;a=commit;h=7428c57ab30ec36fc0da13fe6134a2edac8144b1 [19:36:30] That allowed me to figure out which URLs to purge; if you hard-refresh the debug=true&aftv5_link=E URL now it should work [19:37:12] ^demon: cool [19:38:00] <^demon> The entire repo is only 1.3M now, with history :p [19:38:35] o.O [19:38:38] that doesn't seem right [19:38:50] Is that the MF repo only? [19:39:32] <^demon> MobileFrontend :) [19:39:35] <^demon> Not core. [19:39:57] I haven't run the commands on that yet [19:40:07] <^demon> No, the local copy I just rewrote and pushed. [19:40:11] ah [19:40:16] ready for me to run the commands there? [19:41:49] still don't think I'm seeing E at http://en.wikipedia.org/wiki/Golden-crowned_Sparrow?bucket=1&debug=true&aftv5_link=E, in either Chrome incognito or normal view [19:42:51] I was able to display the Option E by repeatedly trying the Chrome 'incognito' method which Roan proposed. It worked as intended in logged out mode. I then logged in, and it correctly displayed the 'X' close button, as well as showed the flyover, as intended. However, I am still not able to see E on Firefox or Safari. [19:43:28] and now I'm getting a really long wait from bits.wm.o [19:43:43] <^demon> Ryan_Lane: Yeah, sorry [19:44:21] done [19:45:26] Alright, Collection and then it's time for lunch [19:46:06] aha, now I see the dark blue 'improve this page' box, after two really long waits from bits.wm.o [19:46:52] <^demon> RoanKattouw: Do you think it's worth going through all the steps https://www.mediawiki.org/wiki/Git/Conversion/rewriting#ExtraParserTests.php_8MB_string to rewrite out ExtraParserTests.txt? The 8MB mistake? [19:47:23] "all the extra steps" [19:47:34] You're rebasing one file out of existence, that's not hard [19:47:51] Ah, oh, you're editing the file, well still [19:47:57] <^demon> The rebase doesn't like our empty commit messages. [19:48:01] boo [19:48:55] <^demon> I'll make a copy. This could very well go wrong [19:49:23] Just keep an extra branch pointer to the original commits [19:49:37] i.e. run git branch rescue right before git rebase -i [19:50:20] * robla starts drafting an email at the bottom of the Git-migration scratchpad....happy to move elsewhere if it's in the way [19:50:54] <^demon> I'm gonna make a "Done" section at the bottom we can move stuff down to [19:56:46] OK, that's collection done, lunch time [19:57:13] I am happy to confirm that the Article Feedback Activity Log is working as intended on production. The oversight-related items have been correctly moved to the Suppression Log, as they should. [19:57:23] bits.wm.o extremely slow for me [19:58:11] and of course I have just missed roan :-] [19:58:16] hello [19:58:35] ^demon: back :) [19:59:04] chrismcmahon: might just be your local connection. [19:59:10] <^demon> Welcome back. I got the repo down to ~100M. I'm rebasing out the extra parser test now [19:59:28] hashar: could be. good time to break for lunch I think :) [19:59:38] chrismcmahon: probably :) [19:59:48] ^demon: wonderful! [20:00:11] ^demon: were my instruction any helpful ? Wanna share a screen or something ? :-) [20:03:18] <^demon> Hmm, I think I missed it somewhere. [20:03:22] <^demon> $ git rebase --continue [20:03:22] <^demon> maintenance/ExtraParserTests.txt: needs merge [20:03:22] <^demon> Working tree is dirty [20:03:51] <^demon> That was after getting back to r67091 and removing the test case. [20:05:29] <^demon> Ugh no no NO. [20:06:49] I don't think I ever met the message "Working tree is dirty" [20:06:56] <^demon> Neither have I. [20:07:22] though at some point you will have some conflicts [20:07:44] cause change are made to ExtraParserTests referring to the big files and hence git rebase can not apply the diff [20:08:20] <^demon> I'm skipping that. It's not as big a deal. [20:08:37] the ExtraParserTests ? [20:08:43] <^demon> Yeah, I give up on that. [20:08:48] I can do it if you want: ) [20:08:48] <^demon> It's a timesink and I've got bigger problems. [20:09:13] what is left to do anyway ? :-) [20:09:16] <^demon> svn-all-fast-export did something wrong. [20:09:19] * hashar looks at Git-migration etherpad [20:09:32] <^demon> My core dump is messed up, I'm not sure how I'm going to pull the last ~9 hours of core commits in [20:09:58] <^demon> Ah, duh. I'll resume-from then push them [20:15:28] Yay! Thanks to RoanKattouw for showing me how to reset my cookies manually in Firefox, so that I could confirm that the bucketing is working as intended. (On a Mac, go to Preferences, then Cookies, then scroll to en.wikipedia.org, then find the cookie called ext.articleFeedbackv5@0-display-bucket, check its content to see which bucket you are in - 1, 2 or 3, then remove the cookie.) Next time we refreshed the page, you are assigned one of th [20:15:28] 3 buckets at random, and we got Option A, which worked as intended. Whew! I will now do the same for Option E, to make sure all is well there too. [20:15:45] ahh [20:16:08] ^demon: dumping a new one? [20:17:21] <^demon> Yep [20:18:00] <^demon> It's not a fast-forward, ugh [20:18:08] <^demon> It's just a dump with only 4 log entryes [20:18:10] <^demon> *ies [20:18:12] <^demon> blah [20:19:00] <^demon> Ah, maybe this will work [20:20:28] <^demon> Down to 89.01M downloaded on clone... ~100 for the whole tree. I'm not going to micro-optimize that any more. [20:23:33] can you allocate me half an hour to try to remove thatExtraParser test ? [20:23:37] will cp from the repo [20:23:40] try rebase [20:23:45] if that does not work I give up [20:23:53] if that works. Round of beer for the whole team :-] [20:23:58] A couple more notes about the cookie update process described above: on a Mac, click on the Privacy tab of Firefox preferences, then click 'Show Cookies'. Also, to refresh the page, you need to use the standard URL, without the debug code: en.wikipedia.org/wiki/Golden-crowned_Sparrow -- that said, I seem to be stuck with Option A on Firefox, so there may be another cookie in there somewhere. Will try to do the same now on Safari. [20:32:18] <^demon> The hell happened? [20:32:28] <^demon> I fucked up. [20:33:16] What happened? [20:33:47] <^demon> I completely destroyed the core's history. [20:33:56] oh men [20:33:59] <^demon> That's why I left -f off. It shouldn't have done that if it wasn't a simple fast-forward. [20:34:15] I have made a copy a few minutes ago if you want [20:34:21] How far back does this throw you? [20:34:28] <^demon> hashar: Where? [20:34:44] in my home but it is rebasing right now [20:34:47] :/ [20:34:58] That's OK, it'll keep the old commits [20:35:16] /home/hashar/core-fetch on formey [20:35:22]