[00:03:34] hellekin: It's mostly because of OS/2 [00:03:51] Which is a legitimate name for a Wikipedia article, and also not a subpage of anything [00:04:15] 03(mod) Cannot delete tags anymore - 10https://bugzilla.wikimedia.org/45597 +upstream; +comment (10Chad H.) [00:04:17] RoanKattouw: LOL, that's the best answer I could expect :] [00:04:23] 03(NEW) adding "notify translators" link to translate extension - 10https://bugzilla.wikimedia.org/49326 enhancement; MediaWiki extensions: Translate; () [00:04:29] 03(mod) For git server, rewrite http urls to https - 10https://bugzilla.wikimedia.org/49305 +comment (10billinghurst) [00:05:38] 03(mod) For git server, rewrite http urls to https - 10https://bugzilla.wikimedia.org/49305 +comment (10innocentkiller) [00:05:44] mangler: Like where the contents of MediaWiki:Sitenotice is displayed? [00:05:51] Chad: not sure what is the issue then [00:06:20] mangler: Try SiteNoticeAfter or SiteNoticeBefore [00:07:11] bawolff: Yes. that would be almost exactly where I need it. I'm trying to extend AdManager and it's only using the sidebar right now, but I have a requirement to put in a banner add [00:07:50] RoanKattouw: thank you for your explanation. Do you know whether %2F would then be interpreted as / (subpage separator) then? [00:08:05] I think so [00:08:16] But the treatment of %2F is generally black magic and depends on your webserver [00:08:31] However, note that '/' being a subpage separator is a MediaWiki title thing, not a URL thing [00:08:40] So URL-encoding won't really get you around it [00:08:48] arf. The joys of the web [00:08:58] when do we make it a science? ;o) [00:09:00] If for some reason you're creating a page with '%2F' in the title, then that won't be a subpage separator (and will show up in the URL as %252F) [00:09:23] Is that even valid? [00:09:32] It should be, I think [00:09:49] %252F is valid, it just looks broken because 90+% of the time you see it it's because someone screwed up [00:09:50] moment of truth, /me tests [00:10:02] "The requested page title was invalid, empty, or an incorrectly linked inter-language or inter-wiki title. It may contain one or more characters that cannot be used in titles. " [00:10:02] great. Thank you very much for the headstart RoanKattouw [00:10:06] bawolff: Hmm although I'm not sure that % is allowed in page titles, that would be good to test [00:10:07] so doesn't work [00:10:10] Meh OK [00:10:17] So you can't have % in page titles, sorry [00:10:25] and bawolff thank you for testing :] [00:10:30] I think % is allowed, provided it doesn't look like you are url escaping something [00:10:33] Klaas: ping [00:11:07] Cat%25Dog is valid in my test [00:11:10] 03(mod) Have a different link for page patrol and revision patrol - 10https://bugzilla.wikimedia.org/49115 +comment (10krinklemail) [00:11:25] I guess the way around would be to use a unicode character in the private space to display / that is not a slash. [00:12:37] 03(mod) Have a different link for page patrol and revision patrol - 10https://bugzilla.wikimedia.org/49115 +comment (10bawolff+wn) [00:14:06] bawolff: thanks for the help. that SiteNoticeAfter should work perfectly. I will return later if I have more hook questions [00:14:28] mangler: cool. Good luck with your extension [00:14:47] /w/Test/SubTest%2FSubSubTest gives me 404 [00:14:59] hellekin: I'm sure unicode has look alike character. Fractional slash or something [00:15:21] I just committed a change to unit tests that used fractional slash character in a class name [00:15:38] +1 [00:15:42] @_@ [00:16:12] DanielFriesen: I swear it makes sense in context :P [00:16:19] lol [00:16:50] Btw, for the rewrite obsessed mind. Fixing our system to use real subpages would be another option. [00:17:07] Could even add support for non-nested subpages. [00:17:34] wikivoyage has a hack that semi does that i believe [00:17:56] DanielFriesen: non-nested subpage? Do you mean... A page? [00:17:57] Either a page_parentpage(ref) or a page_parentnamespace(int) and page_parenttitle(string) [00:18:28] hellekin: I mean being able to assocate [[Bar]] as a subpage to [[Foo]] instead of only allowing [[Foo/Bar]] to be a subpage of [[Foo]]. [00:18:45] meh, I imagine use case for that sort of thing is rather low [00:18:59] The / based subpages would go out the window besides a bit of back compat. Existing subpages would be grandfathered in. And there would be a new UI. [00:19:04] Perhaps. [00:19:24] 03(mod) [Regression] Special:Log/patrol doesn't show "(automatic)" for autopatrols anymore - 10https://bugzilla.wikimedia.org/35565 +comment (10krinklemail) [00:19:53] But you don't have to turn that part on. It uses the same base code as we'd use to drop the / semantics while allowing Foo/Bar to be a subpage of Foo. [00:20:13] I think a stricture for sub-categories would be more useful. [00:20:39] A stricture? [00:21:08] * bawolff learns a new word [00:21:09] 03(mod) For git server, rewrite http urls to https - 10https://bugzilla.wikimedia.org/49305 +comment (10billinghurst) [00:21:13] The point of all this being that you can have both a [[Foo/Bar]] that IS a subpage of Foo (because it has a page_parent*) and a [[A/B]] that IS NOT a subpage of A (because it has no page_parent*) [00:21:26] hehe, wiktionary ftw. [00:21:44] ;) then we can just get rid of this ridiculous disabling of subpages in specific namespaces. [00:22:17] or the simpler approach - new magic word __NOTASUBPAGE__ [00:22:39] Mm, can you do circles then, DanielFriesen? [00:23:50] Magic word would be less flexible and isn't very backwards compatible. [00:24:18] ;) it's also ugly and not forwards compatible with things like VisualEditor [00:24:21] bawolff: in something like File: ns, why would we want subpages? [00:24:51] sDrewth: moot point, you can't have a / in filenames anyways so it doesn't matter if it's on or off. [00:24:56] sDrewth: variants of an image? [00:25:05] sDrewth: In order to conflict with using the subpage as what page number the pdf is :P [00:25:20] hehe [00:25:43] I've actually thought it would be interesting to use subpages as source files or dependant files [00:25:54] upload a png, and then upload the xcf as a subpage [00:25:55] *meta* data. [00:26:16] or a pov ray source file on the subpage, etc [00:26:16] I know that it is used for djvu files in WS. ThomasV initially had page ns: disabled for subpages, adn they were strings, when the last change was made from 100 to 111, fortunately it didn't break things [00:27:54] Amgine: I do love me my metadata [00:29:07] Me too, but I'm also aware that playing with/supporting metadata can (often does) replace the primary goal of software projects. [00:31:04] The other day, I was contemplating the idea of a parser function that returned meta information about the metadata for a file (metametadata) [00:31:17] I have since decided that was a stupid idea :P [00:31:45] I do think MediaWiki should do a better job of metadata stuff [00:31:49] Thank you. [00:32:06] As a bare minimum, should be able to retrieve file metadata from wikitext [00:32:15] DanielFriesen: in Dokuwiki, namespaces are separated by : and when you use [[namespace:page]] the link only shows 'page' [00:32:32] and should be able to edit a file's metadata (the very common fields anyways) inside the wiki [00:32:33] Of course it should. But doing the metadata better should not detract from the primary mission (however you want to describe that.) [00:33:32] I'm defining the primary mission as "to create a database of freely usable media files to which anyone can contribute" [00:33:55] [00:34:23] They only have 17,246,898 files so far. Not big enough [00:34:26] 03(mod) Bug in showing Arabic scripts content - 10https://bugzilla.wikimedia.org/49113 +comment (10Florence) [00:34:28] Also some of them aren't free ;) [00:34:31] Size queen. [00:35:09] 03(mod) Bug in showing Arabic scripts content - 10https://bugzilla.wikimedia.org/49113 +comment (10dramir2wiki) [00:35:41] that is a *lot* of media... just sayin'. [00:36:15] All the media [00:36:33] except maybe not more penis images, I'm told they already have enough [00:39:28] and perhaps a couple more videos. Watching today's video of the day - it aint exactly going to win an oscar [00:51:37] 03(mod) Install Extension:Widget to English Wikivoyage - 10https://bugzilla.wikimedia.org/47400 +comment (10singaporemaps) [00:51:53] 03(mod) For git server, rewrite http urls to https - 10https://bugzilla.wikimedia.org/49305 +comment (10innocentkiller) [00:54:03] 03(mod) Install Extension:Widget to English Wikivoyage - 10https://bugzilla.wikimedia.org/47400 +comment (10sam) [01:05:51] 03(mod) [Regression] Special:Log/patrol doesn't show "(automatic)" for autopatrols anymore - 10https://bugzilla.wikimedia.org/35565 +comment (10robinhood70) [01:06:40] 03(mod) Install Extension:Widget to English Wikivoyage - 10https://bugzilla.wikimedia.org/47400 +comment (10Bawolff (Brian Wolff)) [01:08:41] Which is the current best-practice for object caching? APC, Xcache? [01:09:16] 03(NEW) overlapping text after moderation on permalink-page of feedback - 10https://bugzilla.wikimedia.org/49327 trivial; MediaWiki extensions: ArticleFeedbackv5; () [01:13:04] ResourceLoader question: in my extension's javascript, can I _rely_ upon having useAjax already loaded by default by the ResourceLoader? It appears so, but I don't know where I find a list of loaded-by default variables. [01:17:08] * RoanKattouw wonders where useAjax comes from [01:17:08] * RoanKattouw greps [01:17:33] I can't find the string "useAjax" in core [01:17:42] Wikinaut: What is useAjax and where does it come from? [01:17:54] Amgine: memcached for object caching, APC for PHP bytecode caching [01:18:30] Thanks RoanKattouw [01:19:22] coming from wgUseAjax from DefaultSettings ??? This is, what I suppose [01:19:29] 03(mod) VisualEditor: Move back to "normal" Save dialog with optional rather than mandatory Review - 10https://bugzilla.wikimedia.org/49258 +comment (10James Forrester) [01:19:38] /** [01:19:40] * Enable AJAX framework [01:19:42] */ [01:19:44] $wgUseAjax = true; [01:20:08] Wikinaut: afaik ResourceLoader will inititalise regardless of wgUseAjax [01:20:19] Oh that, hmm [01:20:27] however having said that I think it is a silly variable that should be removed. it doesn't make sense anymore. [01:20:53] Wikinaut: What are you thinking of using it for? [01:21:24] Wikinaut: Unless you need to interact with something else that currently loads only if wgUseAjax, you can write your own code regardless of wgUseAjax. [01:21:32] extension/AJAXPoll/ajaxpoll.js detects that and then uses AJAX or REST response [01:21:37] e.g. ext.foo.ajaxwhateveryouwant doesn't need to check wgUseAjax if you don't want to. [01:22:01] Wikinaut: How does it detect that currently? [01:22:36] see http://dpaste.com/1218507/ [01:22:39] Also I noticed your original question includes "having useAjax already loaded". I'm not sure what you mean. useAjax isn't a library it is a boolean configuration. [01:22:45] you can't "load" useAjax or something [01:23:07] Wikinaut: Where is that variable useAjax set? It looks like a local variable [01:23:56] Wikinaut: $ret = ' [01:23:59] one moment, I am testing with mungling the variables names and looking if the change appears in the RL-loaded JS [01:24:07] Krinkle: of course [01:24:07] It is set by AJAXPoll itself, it is not from MediaWiki [01:24:18] I am currently changing that to use the RL method [01:24:30] Krinkle: OK [01:24:40] your are correct, [01:24:53] (solved) [01:25:17] If you don't want to change the logic but just refactor to use RL, I'd recommend you change that to do $outputPage->addJsConfig( 'wgAjaxPoll', array( 'useAjax' ... [01:25:36] then use if ( mw.config.get( 'wgAjaxPoll' ).useAjax ) in JS [01:26:41] actually, it looks like it is a non-issue. The reason AJAXPoll has that condition is because it uses sajax_do_call [01:27:13] Assuming you will get rid of that in favour of $.ajax you no longer need that condition because jQuery is loaded always (whereas sajax_do_call (legacy.ajax.js) is only loaded if wgUseAjax is true) [01:27:25] * Krinkle let's you do it :) [01:28:07] Krinkle:two questions: i) do we care of loading such extension-specific javascript-variables on _all_ pages ? [01:28:24] ii) do you mean, I could "drop" the non-ajax part ? [01:28:28] yes [01:29:45] 03(NEW) Create Wikipedia Tuvan - 10https://bugzilla.wikimedia.org/49328 enhancement; Wikimedia: Language setup; () [01:29:47] 03(mod) Wikis waiting for creation (tracking) - 10https://bugzilla.wikimedia.org/16976 (10MF-Warburg) [01:30:21] Krinkle: yes for i), ii) or both) ? [01:30:31] I already answered both before you asked [01:32:00] Krinkle: what you said above should be documented as example in the RL manual page/s [01:32:03] Assuming you will get rid of that [sajax_do_call] in favour of $.ajax, you no longer need that condition [if(useAjax)] because jQuery is always loaded (whereas sajax_do_call (from legacy.ajax.js) is only loaded if wgUseAjax is true) [01:32:04] above = If you don't want to change the logic but just refactor to use RL, I'd recommend you change that to do $outputPage->addJsConfig( 'wgAjaxPoll', array( 'useAjax' . [01:32:18] Krinkle: I do understand now, ty [01:33:46] 03(mod) VisualEditor: Toolbar button text doesn't line up vertically - 10https://bugzilla.wikimedia.org/49249 (10James Forrester) [01:33:58] Wikinaut: I added that to the docs earlier today [01:34:25] 03(mod) VisualEditor: [Regression] Arrow keying around multibyte unicode characters fails after an inline node - 10https://bugzilla.wikimedia.org/49246 (10James Forrester) [01:34:50] https://www.mediawiki.org/wiki/ResourceLoader/Migration_guide_for_extension_developers#Configuration_variables [01:34:52] bawolff: uh, apparently, I did not (yet!) have that page/s in my watchlist, because I didn't yet made a change there... will check rc [01:34:57] uh ^very quick [01:34:59] tsy [01:35:01] ty [01:35:07] bawol [01:35:25] Krinkle should probably double check it, since I'm not a resource loader expert, but i essentially said the same thing he just said [01:35:48] bawolff: bravo. that was, what I yesterday suggested to you, wasn't it ? [01:36:04] Krinkle: bawolff: and it will be good to have examples there [01:36:15] yes [01:36:24] for both cases: global, and per-page loaded variables [01:42:35] hmm, is there an easy way to count how many numeric indexes there are in a php array if the array has both associative and non-associative indexes? [01:43:47] repetez, s'il vout plait, en anglaise? [01:44:38] * bawolff blames the canadian public school system, but me no speaky french [01:44:56] The extent of my french knowledge is that I understand that sentence :P [01:45:22] * bawolff is just going to do isset checks until I come to an element that isn't set [01:45:53] repeat please in english [01:45:54] ;) [01:46:05] [01:46:29] Reedy: in ye-olde-english? [01:46:40] 03(mod) Revision::fetchFromConds SELECT ... FOR UPDATE invalid in Postgres - 10https://bugzilla.wikimedia.org/47055 +comment (10fred) [01:46:49] SPEAK ENGLISH GODDAMNIT [01:47:07] Cool man, I'm down wi' dat. [01:49:42] 03(mod) Create a VisualEditor plugin tool to add/edit maths blocks - 10https://bugzilla.wikimedia.org/43058 +comment (10mybugs.mail) [01:49:47] Krinkle: I tried to add $wgOut->addJsConfig( 'wgAjaxPoll', array( 'useAjax' => $wgUseAjax ) ); in static function AJAXPollRender() but does not work [01:50:17] mmm, i tried installing luasandbox, and mostly succeeded... but half the time, if i try the debug console, i get "Error sending API request: Bad Gateway" [01:51:01] krinkle: https://git.wikimedia.org/blob/mediawiki%2Fextensions%2FAJAXPoll/d5804f11347b2168021857cae1b67134009dde94/AJAXPoll_body.php [01:51:28] ^ link to the AJAXPollRender() fucntion as such [01:51:35] i.e. without my patch [01:51:49] Wikinaut: I know why it doesn't work from there, but before we get on a tangent, you don't need to mess with anything of "useAjax" [01:51:56] You said you understood it now ;-) [01:52:01] Let me clarify again [01:53:00] the reason AJAXPoll has this "useAjax" in the current non-RL code is because the js uses sajax_do_call (a legacy function from mediawiki.legacy.ajax, skins/common/ajax.js). Those legacy functions are only loaded by MediaWiki core if wgUseAjax=true is set; [01:53:25] Krinkle: because you wrote above "If you don't want to change the logic but just refactor to use RL, I'd recommend you change that to do $outputPage->addJsConfig( 'wgAjaxPoll', array( 'useAjax'" [01:53:29] Therefor AJAXPoll had to check for it because if it wouldn't it fail on sajax_do_call() because that function won't exist if wgUseAjax=false by the local wiki [01:53:43] Wikinaut: Yes and then I said "actually, no need for any of that because ... " [01:54:22] But since sajax_do_call is old and you should get rid of that in favour of $.ajax, there is no need for the useAjax check because jQuery (which provides $.ajax) is loaded always and cannot be disabled. [01:54:43] first, i wanted to make the quick fix by adding $outputPage->addJsConfig( 'wgAjaxPoll', array( 'useAjax' [01:55:02] later, to get rid of sajax_... [01:55:20] If you get rid of sajax_ first, you won't need to migrate useAjax [01:55:40] they don't depend on each other, one makes the other obsolete. two for one, win-win :) [01:55:42] ok, then pls. can you tell me, how to recode the sajax... in ajaxpoll.js [01:55:56] Reedy, I branched to origin/REL1_21, the install screen reported MW 1.22Alpha. I'm assuming this means I really *didn't* successfully branch? [01:55:57] delete sajax_* calls, introduce $.ajax in their place. [01:56:21] Wikinaut: Also, what you should probably do before sajax_ and useAjax is get rid of index=ajax and turn that into an api.php module [01:56:32] uh, no [01:56:32] Because that is the lowest level problem of that extension [01:56:38] Amgine: Most likely not [01:56:44] too much at once [01:56:58] too much changes at once [01:57:01] Wikinaut: Not all at once, one thing at a time [01:57:10] currently I am working onthe Html-class change [01:57:13] see .. [01:57:18] Wikinaut: What I mean is I recommend you do that first otherwise you'll be doing double work, migrating things only to get rid of then the next commit. [01:57:21] https://gerrit.wikimedia.org/r/#/c/67002/ [01:57:40] I thought you were working on migrating to ResourceLoader [01:57:41] ^first priority work [01:57:58] [01:58:00] for this, i only need (please help me) the o $outputPage->addJsConfig( [01:58:32] hook AJAXPollRender returns the HTML [01:58:44] Wikinaut: Why? If you're only looking to get rid of raw html, just stick to » » » » array(), 353 [02:00:24] that is in the commit [02:00:37] but bawolff recommended to better use the resource loader [02:01:02] Just to clarify, I meant its better to do that as a separate issue. You don [02:01:08] 't need to do both at once [02:01:11] ok, then I will work on the present solution as in https://gerrit.wikimedia.org/r/#/c/67002/17/AJAXPoll_body.php [02:01:23] You can fix the raw html stuff first, and do the other stuff later [02:01:36] Wikinaut: yes, bawolff recommended right, but that doesn't have to be in the same commit. [02:02:04] * bawolff apologizes if I gave the wrong impression of you needed to do RL stuff in that commit [02:02:36] My -1 on that commit was basically just for the $wgContentLang->normalise thing [02:02:40] It is best done later because that "other" thing (using ResourceLoader) depends on 3 other things that need to be fixed first. that's (among other reason) why the simple $out->addJsConfig didn't work. [02:04:53] bawolff: and the $wgTitle issue, I don't know, what you guided me away from my (working) $wgTitle-> version to $wgParser-> [02:05:30] no, don't use $wgParser->getTitle(), that won't work in some contexts [02:05:49] you told me that before [02:06:18] Just if you're from a parser hook, you should be using the getTitle() method of the parser object being passed as an argument to your parser hook [02:07:17] hte extension was using 'action' => $wgTitle->getLocalURL(), for ages [02:07:23] Wikinaut: Rule of thumb - anything that starts with $wg, that isn't a static config variable you would set in LocalSettings.php, is probably not the currently recomended approach [02:07:26] can I switch back to that [02:08:06] If its between that and something that doesn't work, obviously you can go with the one that works :) [02:08:55] $wgTitle, for reference, has been effectively deprecated since 1.18 [02:09:28] what shall I use ? as you know, this part is also called via ajax [02:10:39] meh, from ajax you might have to use $wgTitle, I'm not sure if there is another option there [02:13:58] 03(mod) Create a VisualEditor plugin tool to add/edit maths blocks - 10https://bugzilla.wikimedia.org/43058 +comment (10James Forrester) [02:18:15] Krinkle: bawolff: perhaps we need Html::element with a further parameter double_encode (default: true) [02:18:32] Wikinaut: Why? [02:18:38] parameter is passed to the Html-internal htmlspecialchars() [02:18:47] Krinkle: I _do_ need this [02:18:51] to avoit [02:18:51] Where [02:18:53] to avoid [02:19:00] I don't think its good practise in general to have content that needs to be escaped, but not entities [02:19:06] my local escapeContent [02:19:30] bawolff: he, I do have characters in my $content [02:19:50] yes, but I'm not clear on the why and how? [02:19:51] and they are output escaped in _your_ version [02:20:03] bawolff: are you the maintainer, or I ? [02:20:19] it was my question of the week [02:20:21] Wikinaut: Where do those characters come from? Why are they there? Is this text or html input? [02:20:36] database? message? user form? [02:20:40] inside ... [02:20:48] user input [02:21:03] Can it contain wikitext? [02:21:06] no [02:21:12] Can it contain html? [02:21:21] wikitext: I wanted to add this later, perhaps [02:22:04] * bawolff thinks proper solution is to allow it to be wikitext, and then escaping but not entities is a non-issue [02:23:44] Wikinaut: Why not stick with strip_tags for now? It looks like any solution (including the new non-standard escapeContent you created) has side-effects that change the behaviour of the extension [02:26:34] to answer your questiion above (html?) [02:26:35] $input = trim( strip_tags( $input->getText() ) ); [02:27:09] $input lines between tgas is later added to the database [02:27:18] shoiuld be using Sanitizer class [02:27:56] Krinkle: bawolff but current patchset, can you +1 with that ? [02:28:11] https://gerrit.wikimedia.org/r/#/c/67002/ [02:28:12] Wikinaut: Assuming this is still the commit 'code cosmetics', try to avoid changing the behaviour in any way as that wouldn't be merely cosmetic clean up [02:28:30] ? [02:28:44] I am not changing behaviour of the code [02:28:55] I mean, don't change to Sanitizer or htmlspecialchars instead of strip_tags since that does change things. [02:29:06] yes, correct [02:29:18] that is, why I now want to merge the present PS [02:29:23] if you _can_ agree [02:29:29] at least for 99,99% [02:29:34] or 99% [02:29:37] I've pointed out the issues in gerrit [02:29:43] new ones ? [02:29:51] I already commented your one [02:30:00] quarter of an hour ago [02:30:04] I suggest we move to #wikimedia-dev so you see the notifications without refreshing the browser :) [02:34:55] Wikinaut: You don't need $double_encode to permit character entities in text. [02:35:19] set to false, that is exactly my point! [02:35:30] but Html::element doesn't do this [02:35:35] No, you don't need the ability to set that. [02:35:40] it_double_encodes" [02:35:50] I have to set to false, basta [02:36:07] otherwise the blanks comes as escaped in the rendered page [02:36:31] Wikinaut: Pass text you want to support character entities in through Sanitizer::decodeCharReferences before escaping. [02:36:35] "something ?" [02:36:41] instead of [02:36:45] "something ?" [02:36:52] this is only one example^ [02:37:35] difference between Html::element(..., content )and Htlm::rawElement(..., content) [02:37:48] Sanitizer::decodeCharReferences [02:37:50] of course, rawElement is to be avoided [02:38:11] DanielFriesen: let's try that... [02:39:24] yep. seems to be working [02:40:09] That's how Linker's comment formatting works. [02:40:53] It decodes character references to actual characters before html escaping. [02:41:48] DanielFriesen: it was a long way to that. I asked around, the first answer was use htmlspecialchars .... [02:41:48] ;) and as a bonus your output gets smaller and normalized so things can be gzipped easier. [02:41:53] new ps [02:43:50] * DanielFriesen wonders if he should start an RFC on moving wmf branches out of mediawiki/core [02:44:12] <^demon> Not an awful idea. Would help keep repo size down. [02:44:15] I am now on #wikimedia-dev [02:44:19] bye here [02:44:43] ^demon: it would also make looking at the repo easier. [02:44:52] DanielFriesen: after you have done the mw release proposal? [02:45:33] No no... someone else can do that on, ;) I have a mw chapter proposal to do instead. [02:45:43] <^demon> DanielFriesen: What we could basically do is mirror core, and only put wmf branches on the mirror one. [02:45:51] Yeah that's the idea [02:46:03] 03(NEW) Package installation request: libcache-memcached-fast-perl - 10https://bugzilla.wikimedia.org/49329 normal; Wikimedia Labs: tools; () [02:46:11] DanielFriesen: wouldn't the releases fall under the chapter though? [02:46:12] wmf branches in core are really a svn holdover [02:46:45] p858snake|l: Only if someone doing releases wants to be hired by the chapter. [02:47:30] The chapter plans are people focused not project focused. [02:48:04] The chapters projects are influenced by what types of people join. Not the other way around. [02:48:16] +' [02:48:21] DanielFriesen: We could also get rid of wmf branches and create tags locally on the deployment host in wmf pointing at master commits. [02:48:37] Wanna change something? Commit to master :P [02:48:40] <^demon> That'd make it harder to cherry pick and keep an eye on what's deployed. [02:49:34] <^demon> I'm fine with scaling back wmf branches, but let's not do something harder :) [02:49:49] wmf/a and wmf/b ? [02:50:06] ^demon: Oh, BTW, I had a question yesterday. [02:50:09] <^demon> Could. [02:50:12] We should also get rid of the silly "wmf live hack" that gets cherry picked for almost a year now [02:50:27] I highly doubt it is that wmf specific. Deal with it. [02:50:32] ^demon: Is it OK to kill REL1_19/etc. branches of VisualEditor (they don't mean anything - we're still in alpha), or will that break things. [02:50:56] James_F: Breaks ExtensionDistributor [02:51:05] Krinkle: I was going to leave master. :-) [02:51:07] <^demon> Nope, you can kill them if you don't need them. ExtensionDistributor checks for existence before displaying. [02:51:14] <^demon> Only one that has to exist is master. [02:51:15] Krinkle: If you drop the incremental notion of wmf branches something a little more git flow might be better than wmf/{a,b} [02:51:17] James_F: They're old, old versions can be unsupported. MediaWiki has old versions too. [02:51:44] Krinkle: Special:ExtensionDistributor tells people that VE REL1_20 is "stable". This is a laughable lie. [02:51:53] Krinkle: It's actively unhelpful to users. [02:51:53] Speaking of REL1_## branches. The naming pattern for those is a svn holdover I wish we could change too. [02:51:55] DanielFriesen: I assume git-flow is not opaque? Where does it keep track of which is which? [02:52:22] Anyway. {{Done}}. [02:52:26] <^demon> DanielFriesen: Nobody has to use those anymore. I just kept them for consistency at conversion time. If you've got any alternatives feel free to RFC. [02:52:41] ^demon: Oh no, I mean the ones in mediawiki/core. [02:52:47] The specific naming pattern. [02:53:05] <^demon> Again, we can pick a new name going forward if anyone wants. [02:53:09] <^demon> I don't rightly care. [02:53:10] The REL1_21 pattern comes from svn. [02:53:20] I'd really like release/1.21 [02:53:37] * bawolff hates change. Long live REL1_21 [02:53:37] Branches are first-class in git. [02:53:47] and svn had no problem with / either [02:53:49] <^demon> I wouldn't mind changing how we do releases. [02:53:54] I assumed those came from cvs [02:53:55] maybe release/1.21.x to indicate it is the working-copy of the release in general [02:54:03] rolling [02:54:05] James_F: 1.19 is on LTS support iirc, but if VE supports it is another question [02:54:10] <^demon> When we're working on a release, we should really commit directly to the stable branch, then merge to master. [02:54:22] <^demon> p858snake|l: For whatever LTS is supposed to mean. [02:54:33] And some git UIs have really nice ways to view properly named branches. [02:54:45] 03(mod) New extension: EXIF parser functions - 10https://bugzilla.wikimedia.org/41498 +comment (10bawolff+wn) [02:54:46] long term support or something >.> its hexmode's baby ask him [02:55:48] When you use a / in git branch names you really do end up with a nice folder structure. And one UI I know of, SourceTree rightly expresses that fact by displaying foo/bar in a collapsable tree structure. [02:56:08] ^demon: LTS = Long Term Support! like Ubuntu! [02:56:11] btw, who came up with the gitblib url scheme? It is highly unusable. Aside from the unfortunate url escaping of '/' in repo names (undertandable from a generic tool point of view, most git hosts don't support / in repo names like Gerrit), aside from that it also does it to the paths of file trees. [02:56:19] In other words, no sane human will type those urls. [02:56:29] And the it redirects 404s instead of serving 404, so you can't correct a mistake [02:56:58] <^demon> I'm ok with improving url structures, everything's puppetized if someone wants to come up with rewrite rules. [02:57:05] Hence if we used release/ then release/1.21, release/1.22, etc... would be beautifully group them all together into a list of 1.21, 1.22, etc... that you could collapse out of your way when you don't need it. [02:57:07] <^demon> That's all default right now. [02:57:59] Gitblit's url scheme is strange. It looks like an unholy combination of TWO different ways of disambiguating the / in urls. [02:58:22] It urlencodes the / AND appends .git to the end of the name. [02:58:41] You only need to do one of those. [02:58:46] <^demon> Krinkle: Also, we can customize the 404 page. Could maybe do something nicer? [02:59:03] ^demon: Having it not hijack the url would be an improvement [02:59:20] I ended up with /p/mediawiki/core.git/commit/... in Gareth. [03:00:11] p858snake|l: VE might possibly support 1.19 at some point, but it will require significant backporting of MW core functionality and VE still isn't stable to consider it. [03:00:34] p858snake|l: By the time it is, we're likely to have a new LTS version. [03:01:09] <^demon> Anyway, gitblit improvements are doable...upstream guy is nice and really responsive to ideas. [03:02:02] * DanielFriesen should go shopping before he touches any more code [03:04:35] DanielFriesen: Dare I say... GitHub... [03:04:44] huh? [03:05:49] <^demon> Krinkle: Are you referring to like https://git.wikimedia.org/summary/?r=mediawiki/omg.git (substitute summary for log, tree, etc), or like https://git.wikimedia.org/omg? [03:06:16] no, though that's also od [03:06:17] odd [03:06:24] It redirects to https://git.wikimedia.org/repositories/ [03:06:36] https://git.wikimedia.org/commit/mediawiki%2Fcore.git/e7651f865d31dd620laaal3828cf509383da2ea7515c3 [03:06:39] urls like those [03:06:47] boom, url gone [03:06:53] <^demon> Ahh, gotcha. [03:07:24] I suppose outputting that entire list is also somewhat expensive, thoguh that's not my reason [03:07:52] on 404 it shows the entire overview homepage with an error stamped on top [03:08:06] <^demon> That's managed in a couple of places. [03:08:17] <^demon> But probably best, could redirect to like log or summary or somesuch at least. [03:08:21] <^demon> If the repo's at least real [03:10:17] <^demon> Hmm, error(String, bool), bool is redirect. [03:10:24] <^demon> Wonder what happens if we don't redirect. [03:11:07] ^demon: nuclear reaction? [03:11:18] <^demon> I'm gonna recompile and let's test. [03:11:26] Yay for testing in production [03:16:54] <^demon> Hrm... [03:30:12] <^demon> Krinkle|detached: I'm just gonna file a bug :p [04:04:13] 03(NEW) Jenkins doesn't run unit tests - 10https://bugzilla.wikimedia.org/49330 normal; Wikimedia: Continuous integration; () [04:30:40] 03(mod) Meta-Wiki name space request for "Programs" - 10https://bugzilla.wikimedia.org/49312 +comment (10b) [04:48:06] I should make RequestContext default to a FauxRequest instead of WebRequest [05:00:18] I wonder if we need something like User::newFromText [05:03:26] I run into this over and over. [05:03:49] "Given an ip address string, how do I get a User instance for that IP" [05:17:04] 03(mod) Package installation request: libcache-memcached-fast-perl - 10https://bugzilla.wikimedia.org/49329 +comment (10Peter Bena) [05:30:23] 03(mod) Provide a central logging service for tools - 10https://bugzilla.wikimedia.org/48846 +comment (10benapetr) [05:35:27] 03(mod) Provide a central logging service for tools - 10https://bugzilla.wikimedia.org/48846 +comment (10benapetr) [05:41:16] Susan: add emus to !b 34911 ? [05:51:05] !b 34911 [05:51:06] https://bugzilla.wikimedia.org/34911 [05:52:09] Sad one QuestyCaptcha is the strongest captcha... [05:53:08] *when [06:33:04] 03(NEW) Click on "X" and still getting sent to board elections page - 10https://bugzilla.wikimedia.org/49331 normal; MediaWiki extensions: CentralNotice; () [06:33:31] 03(mod) Click on "X" and still getting sent to board elections page - 10https://bugzilla.wikimedia.org/49331 (10Sarah Stierch) [06:34:23] 03(NEW) #property gives empty string, but property is set at item on wikidata - 10https://bugzilla.wikimedia.org/49332 normal; MediaWiki extensions: WikidataClient; () [06:35:37] 03(mod) Click on "X" and still getting sent to board elections page - 10https://bugzilla.wikimedia.org/49331 +comment (10sarahstierch) [06:45:55] 03(mod) Empty "nodes" are printed as arrays in JSON even if objects are used for them normally (= when non-empty) - 10https://bugzilla.wikimedia.org/49201 +comment (10db) [06:45:56] 03(mod) API JSON formatter returns [] as an empty return value - inconsistent with {} for non-empty values - 10https://bugzilla.wikimedia.org/10887 +comment (10db) [07:12:33] is it possible to get the size of an edit from the database? [07:13:12] rev_len is the size of the page after the edit [07:15:10] 03(mod) [New UserLogin] Convert Wikimedia-specific messages to be blank in core by default - 10https://bugzilla.wikimedia.org/47801 (10spage) [07:39:57] 03(mod) Jenkins doesn't run unit tests - 10https://bugzilla.wikimedia.org/49330 +comment (10hashar) [07:50:55] 03(mod) Down arrow missing from keyboard icon in RTL - 10https://bugzilla.wikimedia.org/49317 +comment (10niklas.laxstrom) [07:57:46] 03(mod) Jenkins doesn't run unit tests - 10https://bugzilla.wikimedia.org/49330 normal->04CRIT; +comment (10Antoine "hashar" Musso) [07:58:08] 03(mod) Automatic test runs and merging broken - 10https://bugzilla.wikimedia.org/49294 +comment (10hashar) [07:58:08] Can we PLEASE fix the spam filter on wikitech-l. [07:59:31] If the email says From: wikitech-l@lists.wikimedia.org and it's being sent To: wikitech-l@lists.wikimedia.org then it's clearly, obviously, undeniably, irrefutably, SPAM! [08:00:59] well, it could be that the list became sentient [08:01:46]