issue in Webkit [20:07:51] same goes for table of contents (basically it seems my situation grabs the worst of the worst of the double load bounce). but let me know when you're read to discuss what we can do to make it better....! [20:08:00] ready that is. [20:08:06] adam_miller: you still here? [20:08:15] yea [20:08:35] TrevorParscal: I don't know of other breaking bugs other than

wrapping on Webkit [20:08:38] wondering how it's going on the template capsules. or the edit page gunk. [20:08:53] And maybe the fact that encapsulateSelection puts newly inserted content in
s rather than

s [20:09:13] (and it positions the cursor in a weird way post-insert, but that's not horrible) [20:09:37] parutron: i haven't touched either in a while, last week was all bug fixes, some ltr fixes and a few small styling things for hannes [20:09:38] RoanKattouw: the display bug in IE8 that causes

to have too much space [20:09:49] i've nailed it down to a CSS issue [20:09:58] and it's NOT to do with the highlighting marks [20:10:04] cause it does it with or without them [20:10:04] let me know when you're done with bug fixes etc..... [20:10:15] *TrevorParscal contacts parutron in 2 years [20:10:29] TrevorParscal: Cool. Inserting
s inside

s should be fine though, right? [20:10:44] TrevorParscal: Why don't I poke at

s while you poke at that CSS thing? [20:10:49] teeeeee heeeeeee [20:11:59] RoanKattouw: sounds good [20:12:05] OK [20:17:02] where's this damn comma! [20:17:17] RoanKattouw: I'm totally rewriting the 'stylize' function, that shouldn't conflict with anything you're doing, but lemme know if you end up changing stuff in there [20:18:03] nimish_g: I'm poking at something totally unrelated and more time-critical right now, will inform you when I move to your issue [20:18:13] Which may not be for days, as we want to push this release first [20:18:52] the latest invoice you sent doesn't have the total hours. [20:18:56] oops [20:18:59] wrong window [20:19:22] nimish_g: I advised you to comment out a few lines last week (see logs), that should at least fix single-line template calls. Multiline is probably still broken though [20:20:40] brb [20:20:41] food [20:21:16] RoanKattouw: it didn't fix it for single-line template calls [20:22:12] MEh [20:45:22] RoanKattouw: http://img192.imageshack.us/img192/6237/picture1qa.png [20:45:32] I can't find any trailing commas yet [20:45:45] Hm [20:46:10] *TrevorOmNomNom starts reverting to previous revision to see when it broke and who's to blame [20:46:36] http://www.mediawiki.org/w/index.php?title=Special%3ACode%2FMediaWiki%2Fpath&path=%2Ftrunk%2Fextensions%2FUsabilityInitiative%2Fjs%2Fplugins [20:47:17] *RoanKattouw starts using his editor's regex search with ,$ [20:47:25] Or ,\s*$ rather [20:47:50] Hm I can't find Kate's regex search mode [20:47:53] Good old vi then [20:48:32] ok - the IE issue begins with this revision http://www.mediawiki.org/wiki/Special:Code/MediaWiki/61798 [20:50:52] Maybe type: 'foo' is bad? [20:51:06] You know, type could be one of those reserved words [20:51:22] AH! [20:51:35] Same for default, ran into that before [20:51:43] You'll have to quote it like 'type': 'foo' [20:51:50] ok, I've been sort of hinting at this to people on the team, but maybe we should all agree to ALAWAYS put quots around JSON keys [20:51:50] And use foo['type'] instead of foo.type [20:52:18] Yeah that makes sense [20:52:27] (highlight.js around line 155 is where .type is used) [20:52:38] do you want to fix your revision? [20:52:54] I have pretty big local changes [20:53:02] Or wait [20:53:06] Maybe those files are clean [20:53:15] Nah that'd still mess up the combined files [20:53:28] I'll do it after this commit if you don't do it first [20:53:34] k [20:53:46] i will do it, and your stuff can diff to mine [20:53:56] if you think that won't make a mess [20:54:00] TrevorParscal RoanKattouw - I can't get any new line characters to be returned from getSelection in opera. is there a way to fix that or is that just how it is? [20:54:39] adam_miller: Can you debug getSelection() to see which branch Opera takes? [20:54:58] it takes the first, the firefox/opera branch [20:57:35] OK [20:58:02] Can you dump the value of retval after retval = context.$iframe[0].contentWindow.getSelection(); ? [21:00:25] it's the selection without any linebreaks [21:03:25] It's not an object? [21:04:51] in firebug it show the object [21:05:01] in opera, i believe it is, but it logs it out as a string [21:05:22] maybe that's all their javascript console can handle though [21:06:39] Try this: for ( var i in retval ) { console.log( i + ': ' + retval[i] ); } [21:07:53] TrevorParscal: Can you re-instruct me about Webkit's

wrapping? What kind of edge cases do I need to look out for? [21:08:17] aside form it using div instead of p for new lines [21:08:32] it sometimes will wrap multiple divs, brs or ps in a div [21:08:49] let me mess with it to give you a better answer [21:08:50] hld [21:10:26] RoanKattouw: that dumps all properties and functions of the object [21:10:27] OK [21:10:36] adam_miller: Yes, which is what I'm interested in [21:10:39] Can you pastebin the output? [21:11:04] it will do things like, take a
and wrap it in a div [21:11:10] lol [21:11:20] yeah [21:11:25] so, it's the same as just a br [21:11:33] but we need to detect that [21:11:36] RoanKattouw http://pastebin.com/m1970ca8f [21:12:06] now that we are using

webkit is actually going ahead and using

tags for new lines [21:12:12] unless you are at a
[21:12:30] in which case the new line is a
but the previous now is a

[21:12:32] go figure [21:13:23] pasted content seems to be getting placed into the body directly, and everything below it gets wrapped in a div [21:13:29] that's the most insane one [21:14:02] OK so what I'm currently doing is just blindly treating

s as if they were

s [21:14:10] So that's not good enough for

[21:14:17] But what else is it not good enough for? [21:14:21] Does any nesting occur? [21:15:01] adam_miller: Could you do the same for retval.getRangeAt( 0 ) ? [21:15:50] pdhanda: + echo "WHY ARE WE HERE "; [21:15:58] pdhanda: Run svn diff and glance over the output before committing ;) [21:16:37] pdhanda: Also try to keep consistent whitespace style within files [21:16:50] pdhanda: And do not hardcode English text ever [21:16:53] that's an awesome debug message to accidentally commit [21:16:58] http://img23.imageshack.us/img23/6475/picture2plo.png [21:17:43] RoanKattouw: http://pastebin.com/m71fe3816 [21:17:54] sigh [21:18:19] TrevorParscal: OK so basically all occurrences of

that you have found should simply be stripped and ignored? [21:18:43] pdhanda: I have more comments, should I write a comment at CodeReview or do you prefer private chat? [21:19:28] CodeReview is fine [21:19:32] thanks [21:19:34] OK [21:20:01] RoanKattouw: yeah [21:20:18] OK that's easier than what I've been doing [21:20:24] I can revert most of my local changes actually [21:23:54] and can you please take a look at that revision that's making IE freak out [21:27:29] Yes [21:27:40] Right after I finish writing a comment for Priyanka [21:28:04] be nice [21:28:06] she's new [21:29:55] I know [21:32:30] if i was not new, you would not be nice to me? [21:32:33] ;) [21:33:21] No, it's because you're a new committer that I scrutinize your code more :) [21:33:38] If you have questions or need help with any of the things I address in my comment, just ask me or #mediawiki [21:33:46] Specifically the things about internationalization and database functions [21:34:14] Documentation about those things is not what it should be, but a lot of devs know these things by heart [21:34:21] i was kidding, I appreciate it, thanks [21:34:29] heh [21:35:02] Seriously though I do scrutinize new committers more, as they (understandably) make more mistakes [21:35:28] RoanKattouw: even after changing the .type to ['type'] and type: to 'type': it was still causing issues [21:35:45] Bleh so maybe that's not it after all? [21:35:52] Another line that looks dubious is the assignment in the return [21:36:30] TrevorParscal: + return context.htmlToTextMap[origHTML] = leading + retval + trailing; [21:36:34] Maybe that's what IE doesn't like? [21:36:40] hmm [21:36:42] perhaps [21:36:47] what the hell is going on there? [21:36:54] why is there a + at the begining? [21:37:01] Because I copypasted from the diff ^^ [21:37:16] Clearly IE thinks there's a syntax error *somewhere* [21:37:27] Maybe use jslint to find it? [21:38:20] in fact [21:38:27] type isn't a reserved keyword [21:38:49] jslint doesn't like us very much [21:38:51] hehe [21:38:55] Hey wait a minute [21:38:59] ? [21:39:03] If

s should just be ignored [21:39:16] Then what is there left to do? Shouldn't stuff Just Work? [21:39:22] As .text() presumably doesn't add newlines for them [21:39:53] yeah, unless webkit gets confused and starts using divs instead of ps [21:39:58] trying to get it to break now [21:40:00] Oh so that DOES happen? [21:40:42] Please don't tell me I need to reapply all the changes I just destroyed :) [21:40:43] well, it did before we switched to

format to start [21:40:46] Yeah [21:40:54] I think we are good now [21:41:03] did you find the IE syntax error? [21:41:07] should I be looking for it? [21:41:18] No I didn't find it, I was hoping jslint would [21:41:18] was the leading + it? [21:41:22] *RoanKattouw downloads jslint [21:41:25] No, that was from the diff [21:41:37] You know, added lines in diffs start with a + [21:42:02] yeah [21:42:03] sorry [21:42:17] i'm getting all out of sorts over here, clearly [21:43:17] *TrevorParscal blames lack of sleep, exercise and nutrition [21:44:10] *pdhanda blames everyone else :) [21:44:37] *TrevorParscal blames pdhanda [21:49:26] Problem at line 728 character 38: Expected an identifier and instead saw 'class' (a reserved word). [21:49:28] Aha [21:50:15] what did it end up being? [21:50:39] class is a reserved keyword? [21:50:46] RoanKattouw: i think safari is breaking on the same thing right now [21:50:55] Probably [21:50:58] Safari ~= IE [21:51:00] this line? 'beforeSelection': function( class, strict ) { [21:51:17] http://www.quackit.com/javascript/javascript_reserved_words.cfm [21:51:24] class is part of the 3rd table [21:52:54] Yeah probabky [21:52:59] I have a fix in my local code [21:53:04] Plus some more fixes jslint suggested [21:53:14] Most of it is crap, and it even complains about stuff inside comments [21:54:04] yeah, that fixes IE [21:54:25] jslint it cool, but a little too strict imo [21:56:40] Committed fix [21:58:44] TrevorParscal: So does IE choke on type or doesn't it? [21:59:09] nope [21:59:13] type is totally fine [22:00:27] OK [22:00:30] brb [22:00:37] all better in safari as well [22:01:10] RoanKattouw: did you have anymore thoughts on getting linebreak characters from getSelection in opera? [22:01:32] adam_miller: I asked you to dump retval.getRangeAt(0) earlier [22:01:37] i did [22:01:53] http://pastebin.com/m71fe3816 [22:02:00] Oh must've missed that sorry [22:02:07] its cool [22:02:33] I assume .getRangeAt(0).toString() isn't any better? [22:03:22] no it's the same output [22:03:45] just the text selected with linebreaks stripped out [22:04:06] What about .getRangeAt(0).cloneContents().textContent ? [22:04:58] Or $j( retval.getRangeAt(0).cloneContents() ).html() or .text() ? [22:06:56] first one is teh same deal, i'll try the jquery move... [22:09:09] Also try context.fn.htmlToText( jquery thing with .html() ) [22:09:49] i get an error $j( retval.getRangeAt(0).cloneContents() ).html() [22:10:02] err. i get an error WITH that stamenet i mean [22:10:36] same deal with the htmlToText [22:10:44] type mismatch error [22:12:02] Interesting [22:12:17] Lemme poke at this stuff a bit [22:12:54] i was looking at it because the splitlines option does not work in opera [22:13:04] and that would explain why [22:15:53] Yeah [22:15:56] I think I may have a fix [22:16:12] This only happens in Opera and not in other browsers? [22:17:22] adam_miller: Quick sanity check: does $j.browser.opera return true in Opera? [22:17:59] it does [22:18:19] i haven't seen it happen in any others [22:20:37] adam_miller: Try inserting http://pastebin.com/m3f9c144f in the Firefox/Opera branch after retval = context.$iframe[0].contentWindow.getSelection(); [22:22:01] color me impressed [22:22:16] works beautifully [22:22:30] Cool [22:22:33] I'll commit it then [22:28:03] TrevorParscal: Ping. Have you been successful at reproducing

failures in Webkit? How do you feel about pushing to test? [22:28:20] I'm happy with things right now [22:28:28] OK so wanna push? [22:28:36] pdhanda: head's up [22:28:38] yes [22:28:38] I do [22:28:49] OK I first need to merge stuff then [22:29:04] the IE8 double-line thing is still being worked on, but it doesn't cause any functional problems [22:29:09] just display [22:29:10] Good [22:29:33] I'll be merging the current state of /trunk/extensions/UsabilityInitiative into /branches/wmf-deployment/extensions/UsabilityInitiative_alpha [22:30:16] Then I'll sync the UsabilityInitiative_alpha dir out to srv124 (the test server), add facilities to enable it in CommonSettings.php, then enable it for testwiki [22:30:30] I'd normally sync it out to the entire cluster but I'm a bit wary of running scap these days [22:32:39] *werdna waves [22:32:49] Hey werdna [22:33:12] I hardly talk to you anymore now that you're asleep when I'm awake :) [22:34:25] indeed [22:34:31] I just woke up [22:34:39] crazy timezones [22:34:54] Is Naoko around? [22:37:53] *RoanKattouw waits for a slow svn merge to complete on a VM with a throttled internet connection [22:40:16] werdna: howdy [22:40:22] when are you visiting us? [22:40:48] There's another tech meeting in March, right? [22:44:30] yes [22:46:05] I don't know -- ask naoko / erik :) [22:51:14] RoanKattouw: what do you think about adding the old [22:51:14] to the meta? [22:51:48] IE has this display issue, and I'm not the only one who's noticed it, and nobody has a solution [22:52:00] That forces IE8 to go into compat mode? [22:52:11] yes [22:52:15] Hm [22:52:19] I know [22:52:24] it's a bit - lame [22:52:26] but it works [22:52:33] and we have how much time to make magic happen? [22:52:37] oh yeah, less than a month [22:52:39] so, hmmm [22:52:41] Fact is we're doing some workarounds in JS for IE7 that are potentially slow [22:52:42] sounds tasty [22:52:45] Plus well it's generally ugly [22:53:17] hmmm [22:53:20] What's the issue again? [22:53:31] we could put JUST the iframe in IE7 mode maybe [22:53:35] since that's where the issue is [22:53:36] testin [22:53:37] g [22:54:16] hmm [22:55:29] doesn't seem to work... [22:55:40] So what is being misdisplayed again? [22:57:52] a

gets 2x as much spacing as it should [22:58:08] and it's not a CSS issue [22:58:12] it's an IE8 issue [22:58:14] a bug in fact [22:58:48] So this is about stuff like

Line 1

Line 2

? [22:59:12] yes [22:59:16] and the diff is clean [22:59:19] it's just the display [22:59:44] if I give colors to the br and p tags, you can see the extra lines between the br and p tags which there's no accounting for [22:59:52] and it happens with or without highlighting [22:59:58] so it's not the divs for headers [23:00:02] Yeah just checking what kinda HTML provokes that [23:00:46] http://img534.imageshack.us/img534/1290/picture1y.png [23:04:35] TrevorParscal: Is it possible to use a CSS selector like br + p here? [23:04:53] hmm [23:04:56] interesting idea [23:16:01] RoanKattouw: no css hacks work [23:16:10] because if the BR is at all visible, it takes up 2 lines [23:16:14] Dammit [23:16:18] even if it's set to height: 1em or something [23:20:25] pdhanda: I still think you can rewrite that query to not use a subquery: http://pastebin.com/m4cfb5fa2 [23:21:10] ok [23:21:16] I MIGHT have a css hack [23:21:18] let me try that Roan [23:22:27] pdhanda: Disclaimer, I didn't test if those queries are identical [23:22:41] Also, I think we should add some indexes to speed up that query and the one you seem to have copied this from [23:22:46] RoanKattouw: but now the TOC IS messing it up [23:22:47] grrr [23:23:18] The TOC-inserted

s you mean? [23:23:20] RoanKattow: watch out for spam [23:23:37] mysql> SELECT COUNT(cr_id) AS rev_count [23:23:37] -> FROM code_rev, code_paths [23:23:37] -> WHERE cp_rev_id = cr_id [23:23:37] -> AND cp_repo_id = 1 [23:23:37] -> AND cp_path LIKE '/trunk/extensions/UsabilityInitiative%' [23:23:38] -> AND cp_rev_id > -1 [23:23:38] -> AND cr_id > -19987; [23:23:39] +-----------+ [23:23:39] | rev_count | [23:23:40] +-----------+ [23:23:40] | 5333 | [23:23:41] +-----------+ [23:23:41] 1 row in set (0.63 sec) [23:23:57] that ain't right since it is counting each path [23:24:01] what i need is [23:24:11] mysql> SELECT COUNT(cr_id) AS rev_count FROM `code_rev` WHERE cr_id in (SELECT cp_rev_id from code_paths WHERE cp_repo_id = 1 AND cp_path LIKE '/trunk/extensions/UsabilityInitiative%' AND cp_rev_id > -19987); [23:24:11] +-----------+ [23:24:11] | rev_count | [23:24:11] +-----------+ [23:24:11] | 1277 | [23:24:11] +-----------+ [23:24:12] 1 row in set (0.00 sec) [23:24:28] *TrevorParscal gives pdhanda lecture about pastebin [23:24:32] Ah it's a one-to-many table? [23:24:38] there may be a way to do a better join without the subselect [23:24:45] sorry :/ [23:24:54] pdhanda: GROUP BY cr_id mayeb? [23:25:28] one sec [23:25:50] TrevorParscal: How are the TOC-inserted

s messing it up? [23:25:56] *pdhanda looks sideways at TrevorParscal and decides to use pastebin [23:26:01] ha ha [23:26:28] RoanKattouw: basically, i have this selector that's like "if you are a p following a br, have a negetive top margin of 1.5 (the line-height) [23:26:33] and that works [23:26:44] until the TOC comes allong and drops in these divs [23:26:45] pdhanda: In this case pointing me http://www.mediawiki.org/wiki/Special:Code/MediaWiki/61818#c5562 would've sufficed :) [23:26:49] then it all falls apart again [23:26:56] Ah yes [23:26:59] so, it's actually BOTH working against me at this point [23:27:23] And there's no way you can do it on the other side? Like manipulate the p before the br? [23:28:53] Why can't we have section header highlighted? http://toolserver.org/~dispenser/view/Dab_solver#Finding_pages_to_disambiguate [23:29:38] oh [23:29:55] Dispenser: The link seems unrelated [23:30:04] Something about disambig pages [23:30:04] it breaks it, because once you insert the div, it no longer is a p following a br, it's a p following a div class="wikiEditor-heading" [23:30:12] Yeah of course [23:30:19] That's why I asked you if you could do it from the other side [23:30:47] ok I have yet another silly question [23:31:09] there is no way to get a standard list of templates that are commonly used? I have to manually export them from, say wikipedia? [23:31:16] Yes [23:31:29] RoanKattouw http://pastebin.com/m27bb8e7f [23:31:40] You can export pages and templates with Special:Export and check Include templates to recursively export dependency templates [23:31:54] ah, I didn't see the recursive option [23:31:56] pdhanda: Oh of course, I'm stupid [23:32:03] that's the part that seemed EXTRA retarded [23:32:14] Not sure that's the same thing [23:32:15] how about this CSS rule for ya? br + p, br + div.wikiEditor-toc-header + p [23:32:34] TrevorParscal: Hack. At least do div.wikiEditor-highlight [23:33:03] yes [23:33:05] true [23:33:16] Can't we do like $('p + br').each(function() { $(this).prev().css('dostuff'); }); [23:34:22] Although we'd have to do that all the time to keep up with newly added
s, bleh [23:34:35] RoanKattouw: I've come here since communication on the usability wiki was lacking [23:34:41] came* [23:35:11] sorry if this is a bad time to consider new stuff [23:35:20] Dispenser: I get that, but at least give us a link that we understand. The link you gave me seems totally unrelated [23:35:33] Oh wait [23:35:44] You're talking about the yellow highlighting of the visited section header [23:35:50] yes [23:35:56] That's very easy to do in [[MediaWiki:Monobook.css]] and Vector.css [23:36:15] That's how visited ref highlighting works as well [23:36:16] and its very helpful navigation aid [23:36:29] wait till you see this commit [23:36:37] IE8 is a POS to be certain [23:36:47] Dispenser: You can steal the CSS rule from the toolserver wiki and bug local communities about implementing it [23:37:03] RoanKattouw: could you go ahead and merge that too please? [23:37:19] Sure [23:37:24] :) thanks [23:37:34] Although I hate it cause it's not a durable solution [23:37:35] that's not all, there's also anchor so there's an easy perma link to the section [23:37:49] Oh wait it's got .wikiEditor-highlight on it [23:37:52] Then it may not be that bad [23:38:15] Dispenser: They probably do that with local JS as well [23:38:59] ok [23:39:53] RoanKattouw: lets push to test [23:40:17] I already did that [23:40:20] oh [23:40:21] Merging your rev in now [23:40:30] l [23:40:31] k [23:40:36] naoko will be happy [23:40:40] unless it blows up [23:42:11] pdhanda: Does COUNT(DISTINCT cr_id) without the GROUP BY work? [23:43:46] It seems to work with some quick queries on the cluster [23:45:57] TrevorParscal: I'm noticing something similar to what you just fixed in IE8 in firefox [23:46:43] really [23:47:01] working on a bug related to applying the indent to each line of selected text. seems safari and firefox are outputting the same exact thing but it looks like there are extra spaces added in firefox [23:47:17] finally figured out it's just empty p tags being given heights in firefox [23:48:04] TrevorParscal: We should be on test now [23:48:05] the markup inserted is exactly the same, but firefox is still giving a line of space for each of the empty p tags making it look like there's extra
tags being shoved in [23:48:18] haha blank page on test, classic [23:49:01] Aha I forgot to merge core revs [23:51:23] I'm still getting problems with IE's TOC jump to after doing some editing [23:52:27] Problems as in? [23:53:00] after the first few sections each section link jumps to the top [23:53:25] Interesting [23:53:45] Can you verify the relevant anchors exist in the DOM? [23:54:21] Also, once upon a time there was a feature automatically scrolling the highlighted section into view in the TOC [23:54:28] I should dig up its code in SVN and restore it [23:55:22] so.. you made test die? [23:56:45] For now yes [23:56:50] Should be back up shortly [23:58:15] RoanKattouw... it does. Cool! and for some reason i thought i tried that [23:58:45] alright i need to head home. I think the route of the multi line indent issue is that we seems to extract text out of p tags, the empty p tags get left in the iframe and then we reinsert the text around them [23:58:59] if anyone cares about multiline indenting not working perfectly :) [23:59:05]