Jump to content

Wikipedia:Village pump (technical)/Archive 118

From Wikipedia, the free encyclopedia

VisualEditor weekly update - 2013-10-03 (MW 1.22wmf20)

Hey all,

Here is a copy of the weekly update for the VisualEditor project. This is to make sure you you all have as much opportunity to know what is happening, tell us when we're wrong, and help guide the priorities for development and improvement:

VisualEditor was updated as part of the wider MediaWiki 1.22wmf20 branch deployment on Thursday 3 October. In the week since 1.22wmf19, the team worked on fixing bugs and stability improvements to VisualEditor.

References are now editable when they appear inside a media item's caption (bug 50459), thanks to a major re-write of the back-end (bug 52102) that will help us finally provide proper copy-and-paste support in the near future. We tweaked the wording for the warning message that appears for a blank reference list of the default group, and no longer talk about 'the group ""' (bug 51873).

Editing in the "slug" at the start of a page no longer shows up a pawn character ("♙") in some circumstances (bug 54791). In Firefox, blanking the page using select-all and then delete used to break editing; this is now fixed (bug 50947). Generated content blocks which get re-rendered as they are updated, like the experimental (and available on MediaWiki.org) formula editor now doesn't cause the inspector to judder to the begging and back when updating the image (bug 54584).

There were a number of urgent fixes made to the code behind VisualEditor during the week, to fix issues that developed. Firstly, there was a bug that meant that it was impossible to move the cursor using the keyboard away from a selected node (like a reference or template) once it had been selected (bug 54443). Another bug meant that links with a final punctuation character in them broke extending them in some circumstances (bug 54332). The "page settings" dialog was broken twice, first in a way that made removing categories impossible, though adding them worked (bug 54727), and later entirely, though only in a development branch (bug 54928). A broken deployment script meant that briefly VisualEditor was broken for all Wikipedias (bug 54935). Finally, 'snowmen' characters ("☃") no longer appear near newly-added references, templates and other nodes (bug 54712).

A complete list of individual code commits is available in the 1.22/wmf20 changelog, and all Bugzilla bugs closed in this period are on Bugzilla's list.

Following the regular MediaWiki deployment roadmap, this should be deployed here (for opted-in users) on Thursday 10 October.

Hope this is helpful! As always, feedback on what we're doing is gratefully received, either here or on the enwiki-specific feedback page. Please ping me using {{ping|Jdforrester (WMF)}} to make sure I see it promptly if you have any thoughts or corrections.

Jdforrester (WMF) (talk) 01:17, 5 October 2013 (UTC)

Lost text

Here's an odd one...

The template {{langx|lo|ສາມຫຼ່ຽມຄຳ}} in Golden Triangle (Southeast Asia) causes most of the opening paragraph to be lost (Win7/IE10). I can simulate it below with just {{langx|lo|ສ}}, which causes the entire paragraph to be lost (in case you can't see it either, go into Edit to check what's actually there).

Lao: Lorem ipsum dolor sit amet, consectetur adipisicing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat. Duis aute irure dolor in reprehenderit in voluptate velit esse cillum dolore eu fugiat nulla pariatur. Excepteur sint occaecat cupidatat non proident, sunt in culpa qui officia deserunt mollit anim id est laborum.

86.128.5.137 (talk) 01:25, 5 October 2013 (UTC)

Hmm, the above and the lead of the linked article (Golden Triangle) look fine to me. Firefox 24/Mac OS. Can someone with Win7/IE10 give it a try? – Jonesey95 (talk) 06:07, 5 October 2013 (UTC)
Confirmed the issue as described in IE 10. equazcion | 06:14, 5 Oct 2013 (UTC)

I have no idea what the problem is but after digging through the various transclusions it appears that any Lao letters in span tags with the lang="lo" attribute set makes whole paragraphs disappear in IE10, eg. <span lang="lo">ສ</span>. The complete text seems to appear for a split second at first but then disappears, as though it were hidden with CSS or script. In IE10's developer tools, if I disable the font-family style for that tag, which is set to "Phetsarath,sans-serif", the paragraph reappears (IE says this is an "inline" style, despite there being no inline style attribute in the span tag the source code does actually have an inline style attribute in the span tag with that font family specified, even though it's not present in the wikicode). Again, no idea what this is about, but just reporting what I learned. equazcion | 09:35, 5 Oct 2013 (UTC)
The "split-second thing" sound like what's happening in a bug I reported the same day here. EEng (talk) 13:29, 6 October 2013 (UTC)
If I create a .htm file containing just the text hello <span lang="lo">ສ</span> hello and open it in IE then it displays fine. However, if I paste the exact same thing into a Wikipedia page then the text disappears, so it seems to be something that Wikipedia is adding, or some interaction with some other Wikipedia elements, that is causing it to happen. 86.161.61.174 (talk) 12:35, 5 October 2013 (UTC)
I haven't figured out which embedded transclusion inserts the font-family tag, but I wonder if there is a typo in that section of code, like a tag that is opened but not closed. – Jonesey95 (talk) 19:03, 5 October 2013 (UTC)
The font-family style is most likely added by ULS. Please file a bug for it. Matma Rex talk 19:37, 5 October 2013 (UTC)
I don't actually see any font family specification. When I look at the relevant content in IE "view source" I just see:
<div lang="en" dir="ltr" class="mw-content-ltr"><p>hello <span lang="lo" xml:lang="lo">ສ</span> hello</p>
However, I am not completely sure what I'm doing, so I may be looking at the wrong thing. 86.160.82.72 (talk) 19:51, 5 October 2013 (UTC)
The attribute doesn't appear in "view source", which displays the code prior to any changes made by javascript -- it does show up when using IE10's "developer tools" (F12), or Firefox/Chrome's "inspect element". Looks like the inline style is added by a script, as Matma Rex alludes to, perhaps something to do with this part of the ULS extension. equazcion | 20:44, 5 Oct 2013 (UTC)
I see. Perhaps someone who is able to do so could file a bug for this as suggested above? 86.146.105.171 (talk) 13:13, 6 October 2013 (UTC)

Template: orphaned non-free revisions not functioning properly

{{Orphaned non-free revisions}} should work the same way as {{non-free reduced}}, but it doesn't, and I can't figure out what's the matter with it. What's supposed to happen is files with the template are supposed to be placed in the Category:Non-free files with orphaned versions more than 7 days old after the tag has been in place for a week, but they don't - the category is always empty. I have tried tinkering with the template, and examined the related category, and compared it to the similar (functional) template {{non-free reduced}}, and I am stumped. If anyone with template chops would like to have a look at it and get it to behave I would appreciate it. I have been manually looking through the files to do the maintenance task, but obviously it would be better if we didn't have to depend on that. Thanks, -- Diannaa (talk) 02:52, 5 October 2013 (UTC)

I believe {{non-free reduced}} works because Joe's Null Bot (talk · contribs) hits them all with a null edit once a day so that the software updates the category when seven days are up. Perhaps Joe Decker (talk · contribs) [ping!] can add Category:Non-free files with orphaned versions to the bot's list. -- John of Reading (talk) 06:15, 5 October 2013 (UTC)
I tried tagging File:2013ALDS.png with {{Orphaned non-free revisions|date=16:00, 27 September 2013 (UTC)}} and previewing, and Category:Non-free files with orphaned versions more than 7 days old came up in the hidden categories (I have hidden categories turned on in Preferences -> Appearence -- Category:Non-free files with orphaned versions, on the other hand, doesn't appear to be hidden). If images are failing to show up maybe the date parameter is being left out? Can you point out an example image that should be in the category but isn't? equazcion | 06:20, 5 Oct 2013 (UTC)
(e/c) {{Orphaned non-free revisions}} uses "Date" where {{non-free reduced}} uses "REVISIONTIMESTAMP". That seems to be the main difference that would prevent it from working entirely. {{non-free reduced}} also has an if statement that appears to act like an "else" statement, placing articles in a separate category when they have an invalid time stamp, which is probably helpful.
If I were super-bold, I would try using REVISIONTIMESTAMP in the non-working template (and be ready to undo the change if it set something on fire). I poked through the template's history, and it's not clear when/if the template was actually working as intended. – Jonesey95 (talk) 06:22, 5 October 2013 (UTC)
User:Theo's Little Bot seems to insert the date explicitly as needed. I'm not sure if changing the template would screw that up.... Although offhand my guess is no, it might be wise to contact the operator before changing it. equazcion | 06:27, 5 Oct 2013 (UTC)
Correction, the bot was adding the "reduced" tag in my diff, not the orphaned one. Although {{Orfurrev}} appears to insert a date as a parameter in orphaned tags. equazcion | 06:33, 5 Oct 2013 (UTC)
Files that were tagged Sept 27 and earlier should be showing up in the category. File:Sleater kinney.png, File:Specialteeshot.jpg, File:Swing lindbergh eagle.jpg, File:The Treasure of Tranicos.jpg, and many more qualify to be in the category. I don't think there's any in files Category:Non-free files with orphaned versions right now without date stamps. Something may have gone wrong with the template when it (and/or its related category) was moved in January 2012; I don't recall when it got broken, it's been a long time. Adding and removing the material that adds files to Category:Non-free files with orphaned versions with invalid timestamp doesn't seem to make any difference. This revision from May looks to be identical to the other template (only the names are different), and it didn't work. The associated Category:Non-free files with orphaned versions more than 7 days old seems to be built the same way as the functional Category:Rescaled fairuse files more than 7 days old. I've tried null edits and purging on both the template and the category. -- Diannaa (talk) 14:42, 5 October 2013 (UTC)
The category Category:Non-free files with orphaned versions more than 7 days old appears populated now when I look, for some reason. One of the things you did may have worked after all? equazcion | 14:52, 5 Oct 2013 (UTC)
Equazcion see my comment below. Werieth (talk) 14:53, 5 October 2013 (UTC)
I gave all the images in the category a swift kick, and 35 of them dropped into the sub-category. Werieth (talk) 14:51, 5 October 2013 (UTC)
Why are they not dropping automatically? I need to know that if I get hit by a car or something that things will carry on without manual intervention -- Diannaa (talk) 14:54, 5 October 2013 (UTC)
Its a design flaw that is caused by the need to cache pages. Link references, categories and other values are only computed when an article is saved (IE the parser processed the page). There are a few things that can trigger an update but those use the job queue and can take a prolonged period. The best thing to do is have Joe's Null Bot hit the category once a day. (I just did the equivalent) Purging the template and category pages themselves has no affect on the image description pages and those need purged/forcelinkupdate'ed separately. Werieth (talk) 15:00, 5 October 2013 (UTC)
  • (edit conflict) {{Non-free reduced}} uses {{REVISIONTIMESTAMP}} if the editor forgot to indicate a date. If {{non-free reduced}} is used without a date, then this means that the template thinks that the page was tagged at least 7 days ago if the latest edit to the page was at least 7 days ago. If {{{date}}} or {{{1}}} is specified, then that date is used instead. {{Orphaned non-free revisions}} used to support {{REVISIONTIMESTAMP}} too, but I see that it was removed by User:Diannaa in August. It would probably be a good idea to restore it.
When 7 days have passed, the files do not necessary move to the other category, due to category caching. To get a category update, you either need to make a null edit, or use mw:API to purge the page with certain options. From what I can see from the above, there is apparently a bot which automatically purges the reduced files but not the ones with orphaned versions, and for this reason, you only get quick updates for {{subst:furd}} but not for {{subst:orfurrev}}. --Stefan2 (talk) 15:04, 5 October 2013 (UTC)
Also, if I remember correctly, Mediawiki was recently changed so that you can't just make a null edit to the template; you need to make a null edit to each and every file in the category instead. --Stefan2 (talk) 15:12, 5 October 2013 (UTC)
I will restore the material I removed; it was only taken out in an effort to get the template working. I will ask the bot operator to add this task to his list of tasks for the bot. Thanks everyone for your help resolving this issue. -- Diannaa (talk) 15:17, 5 October 2013 (UTC)
I don't know much about the job queue, but my experience in fixing errors in templates during the last month has shown that articles transcluding those templates are not typically removed from the corresponding error category (Category:Pages using citations with old-style implicit et al.) for at least two to three weeks, maybe longer. From what I have read here on WP, the job queue essentially performs a null edit on every article periodically; it looks like that lag is longer than the 7-day trigger your maintenance work requires. – Jonesey95 (talk) 19:08, 5 October 2013 (UTC)
Thanks for the additional info. One weekend I didn't do the task, and the category still remained empty two weeks later, so yeah, it's not happening all that frequently. We can't afford to leave orphaned revisions lying around until the job queue does a random walk on them, as they violate the NFCC. -- Diannaa (talk) 03:25, 6 October 2013 (UTC)
The job queue is a bit slow. I have sometimes seen it taking weeks or months before a category is updated, but it will be updated eventually at some point. I agree that potentially waiting several months is too long for these NFCC violations. --Stefan2 (talk) 10:20, 6 October 2013 (UTC)

Removing unicorn from Rainbow article

Yesterday someone added unicorn-nonsense ("comes with a unicorn or leperchan with a pot of gold") to the rainbow article. The edit was immediately reverted by another user. Strangely, the unicorn-nonsense has disappeared only in edit mode. The unicorn-nonsense still persists in view mode. Force refreshing the browser doesn't help, nor does using another computer. My question is: (1) do other people see the unicorn-nonsense in the current rainbow article too? (2) How can it be removed? Ceinturion (talk) 08:53, 5 October 2013 (UTC)

Yes, I saw it as well. I gave the servers a kick by purging the page, and now it's gone. -- John of Reading (talk) 08:59, 5 October 2013 (UTC)
Thanks! Ceinturion (talk) 09:00, 5 October 2013 (UTC)

Speedy not a Speedy

Manawatu Rugby Union is showing up in Category:Candidates for speedy deletion and Category:Candidates for speedy deletion by user although I can't see an obvious reason why. Ben MacDui 13:42, 5 October 2013 (UTC)

[1]Werieth (talk) 13:45, 5 October 2013 (UTC)
Gotcha - I read the above thread, purged the page, and it seems to have sorted itself out. Ben MacDui 13:49, 5 October 2013 (UTC)

Inadvertant sockpuppetry

As I frequently do, I was editing Wikipedia and checking out Commons for images at the same time. I have different moniker at the latter. Engaged in an edit I found that when I wanted to save it I had been logged out as my usual self and logged in under the Commons pseudonym without any action that I am aware of on my part. I decided I was going to lose the edit if I logged out again so I just saved anyway per this diff, so creating a new WP editor into the process. Something similar happened about a week ago although I didn't press the save button. Anyone know what's going on and how to prevent recurrences? Ben MacDui 16:52, 5 October 2013 (UTC)

You don't actually need the Tarbert account, as you have one at commons:User:Ben MacDui (currently a redirect to the commons:User:W. L. Tarbert acct). The local Tarbert account was auto-created when you came over to en.wiki without logging out of that account name. See Wikipedia:Unified login. Now you have a wee sock :) -- Diannaa (talk) 18:39, 5 October 2013 (UTC)
Special:Log/W. L. Tarbert shows the account was actually created 30 September where you must have visited the English Wikipedia without logging out at Commons. Unified login makes it difficult to keep different usernames at different projects. I suggest you consider using the same name, for example with two requests at commons:Commons:Changing username to first change Ben MacDui to something else and then change W. L. Tarbert to Ben MacDui. PrimeHunter (talk) 21:12, 5 October 2013 (UTC)
Many thanks for explaining this. As my first preference is to remain as W. L. Tarbert on Commons and there not to be a editor with that name here I am not sure what to do. I think I will put a note on en:Wiki W. L. Tarbert to say that this is a computer generated sockpuppet of my account and redirect the talk page to User talk:Ben MacDui. I suppose I just have to log out of the MacDui en:wiki account if I want to check my Commons watch list or upload a file. I have just logged on to Commons using a different browser - wonder who I'll be when I sign this message. Ben MacDui 09:29, 6 October 2013 (UTC)
See also bugzilla:14409 (Allow user accounts to be 'linked' to a master account) and bugzilla:48892 (Ability to link accounts). Helder 12:48, 6 October 2013 (UTC)

Math formula fails sporadically with unknown error

While editing the Rapidity [2] article, I notice a very unusual bug: most of the formulas failed to display for unknown reasons (in PNG mode), about 1/6th of the time. I could "fix" the error by inserting or removing spaces at the end of the formula, but it just didn't make any sense. My hypothesis is that somewhere there's a caching bug that's local to one of the servers, but other than that I couldn't tell why it rejected the markup.

An example is shown below. Make sure it's in PNG mode, and you may need to refresh the page a couple times. Notice how the codes are identical except for the last additional space character)

Renders fine:

Fails occasionally:

--Freiddie (talk) 01:04, 6 October 2013 (UTC)

This is pretty interesting. Every once in a while when I load this page the second instance does indeed show a red error. equazcion | 05:35, 6 Oct 2013 (UTC)
I also see this as an error - for 4 tries, at least, it was fully reproducible, both by viewing the article and by previewing edits with the space added or deleted. (This was with current Firefox; allowing javascripts did nothing to change the error) Wnt (talk) 06:03, 6 October 2013 (UTC) LOL, I just finished this and was ready to walk away with a second edit about the scripts, and then the error magically went away. :) And now there's consistently no error. Wnt (talk) 06:07, 6 October 2013 (UTC)
This is a bug that is being tracked in Bugzilla. – Jonesey95 (talk) 06:25, 6 October 2013 (UTC)

Equals-sign glitch in efn note

I wrote a footnote, using efn. When I previewed it, I was told that I had put refs with no content. There was no new ref, and the footnote had lots of content. The footnote included some math symbols, including an equals sign. After some experimenting, I found I could make the thing behave properly by enclosing this sign in nowikis. Weird. Check what I just did in the Parabola article to see this. DOwenWilliams (talk) 04:00, 6 October 2013 (UTC)

It's because of the equals sign, which makes it parse the text before the equals as a parameter name. There's a few ways around it, and the way you did it is fine.—Kww(talk) 04:12, 6 October 2013 (UTC)
This section title is rather amusing when read aloud. I picture you throwing something. — kwami (talk) 04:31, 6 October 2013 (UTC)
Those efn equals signs get me pretty riled up too. equazcion | 05:21, 6 Oct 2013 (UTC)

ClueBot archive index weirdness

I recently set up ClueBot III archiving at Module talk:Convert (diff). ClueBot has made an archive, but there is a lot of junk in the index (see Module talk:Convert where there is a list of 500 items in the archives box, and see User:ClueBot III/Master Detailed Indices/Module talk:Convert and User:ClueBot III/Indices/Module talk:Convert). I assume that is a ClueBot glitch? The same weirdness is at Template talk:PD-CAGov, and I'm asking here in case there are others that need fixing. Johnuniq (talk) 09:24, 6 October 2013 (UTC)

I don't see anything wrong with the invocation, so it does seem a bit mysterious. I've left User:Cobi (the bot operator) a message on their talk page, so hopefully they can come and take a look. — Mr. Stradivarius ♪ talk ♪ 14:52, 7 October 2013 (UTC)

How to edit Help:Contents

How would I edit Help:Contents to flip the appearance of two sections? I want the currently top right portion of "I want to edit Wikipedia" to appear in the top left and the portion "I want info about browsing or searching" to appear in its place. Thanks. Biosthmors (talk) pls notify me (i.e. {{U}}) while signing a reply, thx 11:51, 6 October 2013 (UTC)

You would grab everything from the first |style="{{/cellstyle}} until the next, and paste it just above the subsequent |- . See User:Equazcion/sandbox4 for what the code looks like (it looks messed up there though because it's missing subpages). No opinion on whether to actually make the edit. equazcion | 12:17, 6 Oct 2013 (UTC)

Category:Use dmy dates from May 2013

An article i use to update is in Category:Use dmy dates from May 2013. Should i do something for it? Is that a problem? Xaris333 (talk) 18:06, 6 October 2013 (UTC)

No. It is just an informational category used to denote that dates used within the article should use the day month year format. Theoretically, a bot could go around correcting any entries that came along and switched to mdy dates, but you don't have to do anything. Resolute 18:14, 6 October 2013 (UTC)
Thx! Xaris333 (talk) 19:04, 6 October 2013 (UTC)

Very quiet today

Let me guess -- everyone's at church? EEng (talk) 19:47, 6 October 2013 (UTC)

Saturdays and Sundays are generally "quieter" around the English Wikipedia for some reason. This tool seems to corroborate this at least for the last few weeks. I'm sure there is a better tool to examine it over a longer period. Killiondude (talk) 22:34, 6 October 2013 (UTC)
It's because Wikipedia is really just for doing when you're supposed to be working. Otherwise there's not much point. equazcion | 00:16, 7 Oct 2013 (UTC)
I've looked on holidays, and pageviews don't seem to drop much then. French Wikipedia seems to have smaller weekly variations than English, but the Spanish version varies a lot more. Saturdays tend to have smaller numbers of pageviews than Sundays, but both are less than weekdays. Maybe it's because, around here anyway, electricity is cheaper on weekends than weekdays, so people spend more time doing housework and laundry, instead of playing on the computer. Or maybe they just sleep more. DOwenWilliams (talk) 03:34, 7 October 2013 (UTC)

How to use "Header" of "Collapsing" templates on wiki-site.com

I tried to use these templates, but they don't identify as templates. I was advised to enter java-code from https://en.wikipedia.org/wiki/MediaWiki:Common.js , but I don't know how to make it work (usual "<script> </script>" tag doesn't work). What can I do? NAME XXX (talk) 21:30, 6 October 2013 (UTC) .

Are the additions of the image and image caption parameters to the company infobox correct?

I noticed that in the company infobox, image and image caption parameters had been added. Should this be reverted? Please advise. Aeroplanepics0112 (talk) 00:13, 7 October 2013 (UTC)

I don't see why. It seems to work fine. Jackmcbarn (talk) 00:21, 7 October 2013 (UTC)

Special:Log not loading?

I'm getting a 503 error if I go through someone's contributions and navigate to their logged edits:[3]. Not having any problems getting through the main Special:Log interface, however: [4]. Marcus Qwertyus (talk) 06:15, 7 October 2013 (UTC)

See #timeouts. Legoktm (talk) 07:32, 7 October 2013 (UTC)

Parser profiling data

maybe it's been there for a while, and i did not notice, and maybe it's a new feature with 1.22wmf19 (see Special:Version).

either way, i just noticed something that can be very useful for people who care about performance: when editing (this works with wiki-editor, but unfortunately not with VE, as far as i could tell), when you hit "preview", you'll notice at the bottom of the page a new "Parser profiling data". as far as i could see, this contains the same information that's been included as a comment in the page source, under "NewPP" for quite some time now, but it's much nicer to get it directly from the "preview" page. in addition, the information in "preview" pertain to the part you are actually editing, so it allows us to profile either a single section or the whole page, which may help find performance bottlenecks.

they say that "The Appetite is Sharpened by the First Bites", so here is some additional requests from the profiler, and from preview tools in general:

  1. the profiler tells us how many "expensive" parser functions the page uses, but it does not tell us which. would be nice to know that.
  2. when running preview (or edit) for the whole page, we get a list of all the transcluded templates. however, when a template is transcluded through another (and sometimes there are several levels of transcludion - i think i saw something like 4 or 5 in enwiki), it would be very helpful to be able to get the transclusion tree, not just the "flattened" list of transclusions. it's sometimes pretty hard to find how something is transcluded.

peace - קיפודנחש (aka kipod) (talk) 17:41, 7 October 2013 (UTC)

This looks similar to bugzilla:3241 (which is about the "What links here" feature) but I didn't find an equivalent bug specifically about the "Templates used" feature. Helder 17:53, 7 October 2013 (UTC)
maybe "similar", but logically speaking, this is really the mirror image of "what links here": the templates list is not a list of pages that link or transclude this one, but rather a list of templates this page transscludes. peace - קיפודנחש (aka kipod) (talk) 18:09, 7 October 2013 (UTC)

Some help by a technically skilled user is needed there. Thank you in advance. --Leyo 09:01, 20 September 2013 (UTC)


Testing new version of GettingStarted experience

Hey, for those interested, soon we're going to be running a controlled experiment (an A/B test) of a new version of the GettingStarted experience delivered to new users right after they register. This means that for about a week, the experience when you register will differ depending on whether you have an odd or even user id. There's a thread on this talk page, and I would really appreciate any detailed feedback people might have. Thanks, Steven Walling (WMF) • talk 02:20, 4 October 2013 (UTC)

BTW, this test is live now. Steven Walling (WMF) • talk 00:20, 9 October 2013 (UTC)

Page view statistics broken (again)

Henrik's tool produces an "HTTP 500 Internal Server Error" on any wiki page. Just wondering if anyone can help. Thanks in advance, XOttawahitech (talk) 04:39, 6 October 2013 (UTC)

Oops, should be fixed now - the database had fallen over, someone needed to push it upright again. Stats for yesterday should be up soon, in an hour or two. henriktalk 07:30, 6 October 2013 (UTC)
Still not up, Henrik. Do you have an idea when it may be working again? Thanks--أخوها (talk) 19:25, 8 October 2013 (UTC)
It's working for me, at the moment. equazcion | 19:29, 8 Oct 2013 (UTC)

Saving talk pages

Is it just me, or is anyone else finding that it's taking two minutes or so to save a talk page post? They're opening slower than most pages too. Peridon (talk) 11:24, 7 October 2013 (UTC)

I've been getting intermittent slowness when previewing and saving in general, ever since that was first reported around a week ago. It keeps fixing itself and reappearing. equazcion | 11:29, 7 Oct 2013 (UTC)
The issue is probably the fact that most talk pages haven't been archived in about a week since the MiszaBot's went down. Legobot went on a archiving spree today, so most pages should be a bit faster. Legoktm (talk) 00:02, 8 October 2013 (UTC)
I noticed Legobot taking over archiving for a lot of Misza-configured pages today. Thanks for doing that. I'm not entirely convinced page length is actually the root of the particular speed issue we're seeing, but that is still quoite noice. equazcion | 01:39, 8 Oct 2013 (UTC)
Talk pages have always been slower than articlespace pages for me; how much slower varies. - The Bushranger One ping only 17:49, 8 October 2013 (UTC)
  • My Twinkly menu keeps going in and out. Wikipedia in general has been slow as thick shit through a funnel (as the Dutch say) the last few days. Drmies (talk) 17:30, 8 October 2013 (UTC)

Server issues part MCMXVI

Just noticed the servers are going cough-sputter-hack, with HotCat not loading sometimes (sometimes taking five-to-six refreshes before correctly loading) and pages on both en.wiki and Commons occasionally loading in an "unstyled" state. - The Bushranger One ping only 17:30, 8 October 2013 (UTC)

Anyone Else Noticing Wikipedia Loading Problems?

WP has been acting wonky the past few minutes. Pages load slowly--look like HTML 4, early-web renderings--no sidebar, just hyperlinks and text. And I'm not seeing my Twinkle tabs. I've tried purging and doing a force-refresh (Ctrl+Shift+R) but no luck. Anyone else experiencing this? -- Veggies (talk) 17:33, 8 October 2013 (UTC)

Yep, just posted about this above (and got the "unstyled" format the first time I clicked 'edit' here). Just keep refreshing until it works... - The Bushranger One ping only 17:35, 8 October 2013 (UTC)
I'm getting problems intermittently, but refreshing a few times usually fixes them. Specifically it seems to be loading slower and forgetting to load parts of the pages for me. Ks0stm (TCGE) 17:36, 8 October 2013 (UTC)

There was a minor issue with the bits caches so you may have seen some unstyled content for a bit. All should be resolved now. ^demon[omg plz] 17:51, 8 October 2013 (UTC)

Speak up about the trademark registration of the Community logo.

Article taking ridiculously long time to save

Hi, I've been making some edits to Panama Canal, and the article is consistently taking a ridiculously long time to save, even when I am only editing a relatively short section. Is there anything obvious about the article that is unusual or inefficient or broken? 86.128.4.151 (talk) 02:05, 9 October 2013 (UTC)

Without and with {{Panama Canal map}}
Without With
CPU time usage 7.456 seconds 12.373 seconds
Real time usage 7.875 seconds 13.258 seconds
Preprocessor visited node count 37173 67762
Post-expand include size 329993 750995
Template argument size 69012 155049
I would say {{Panama Canal map}}. There's something happened to the processing of WP:RDTs recently, unrelated to template or module changes, that has slowed them right down. The {{Panama Canal map}} RDT itself hasn't been edited since 16 June 2013, and of the subtemplates, the only edit since 10 July 2013 is this one, which is insignificant. However, RDTs handle a lot of small images, such as , and images are held on a different server than the wikitext. If that server is slow, RDTs will be slow, and so will pages that use RDTs.
By way of experiment, I did a full-article preview of Panama Canal both without and with the {{Panama Canal map}}, and I found that although the highest expansion depth just went from 29 to 31, some of the other figures were significantly greater, see right. Notice in particular that the last two rows are more than doubled. --Redrose64 (talk) 09:23, 9 October 2013 (UTC)
Thanks for taking the time to look. 86.160.83.13 (talk) 11:14, 9 October 2013 (UTC)


Overlapping portal boxes in Safari 6.0.5

Could someone help out at Portal talk:Australia#Error on portal page. please? -- John of Reading (talk) 18:31, 8 October 2013 (UTC)

Teahouse looks strange

The Teahouse questions at the top of the page are indented but the ones at the bottom are all the way to the left against the border.— Vchimpanzee · talk · contributions · 21:24, 9 October 2013 (UTC)

There was a stray closing </div> causing it that I removed. Chris857 (talk) 21:52, 9 October 2013 (UTC)

Disallow VE software changes from the WMF for the foreseeable future

After testing the new release of VE, announced here in #VisualEditor weekly update - 2013-09-26 (MW 1.22wmf19), it became quite obvious that I am apparently the first one to test these changes before they are implemented live, since the predictable bugs were very easy to find and will create loads more problems (luckily, with the opt-in, the amount of VE edits has dropped dramatically).

I have posted these problems in the above thread (where one developer responded) and at Wikipedia:VisualEditor/Feedback#Error reports before they even happen!, where no one from the WMF could be bothered to reply.

I then filed Bugzilla 54737, which was closed as invalid. Apparently, we first need to get consensus that theior product and new release sucks before they can act upon it, instead of using their brains and testing it for themselves. Furthermore, even if we get a consensus, "on average, they take a month or two to process"[5]. But anyway, here we are.

I would like to see whether people agree that, considering the current state of VE and the quality of the weekly releases (buggy, badly and incorrectly described, ...), it would be a lot better if we didn't get any VE updates or releases until most of the major bugs are solved and we get an actually working product. The WMF is still trying to use the Wikipedia's as their forced "community testing ground" (see nice older pages like [6]: "The VisualEditor project aims to create a reliable rich-text editor for MediaWiki. It is a top priority for the Wikimedia Foundation and it is available for testing on the English Wikipedia."), whether we want this or not.

This has to stop. They can test it on their own pages if they want to, or on testwiki, or on Wiki-versions that explicitly agree to be a testing ground; but they shouldn't bother us with it, and they should stop pushing new, untested, faulty releases to us. Fram (talk) 07:37, 30 September 2013 (UTC)

Do these updates affect the opt-in nature? equazcion | 07:44, 30 Sep 2013 (UTC)
... you told them to not push any updates for 3 months in the bug! 1) You're not even giving them a chance to fix the bugs that exist, 2) that is not what Bugzilla is for. --Rschen7754 07:45, 30 September 2013 (UTC)
As long as it's opt-in only and being used by people that are using it with an eye to evaluating the results and the interface, I can't get too excited about releases, Fram. Even well-run projects have the occasional regression. If I see evidence that people testing Visual Editor are damaging Wikipedia and not cleaning up after themselves, my excitement level will quickly rise.
Conceptually, I agree with you that until they get tables, complex templates that include styles and table formatting, and tables themselves working, there's no actual reason for anyone but themselves to test the code.—Kww(talk) 07:52, 30 September 2013 (UTC)
Don't you think you're getting a little bit extremist? You want them to fix VE's problems by... doing nothing. Sure. Adam Cuerden (talk) 07:58, 30 September 2013 (UTC)
No, he wants them to fix VE's problems by fixing VE's problems. As Fram notes, they don't need the live English Wikipedia community in order to find problems, since many problems are already apparent. A beta software release for live testing should really only be done once obvious inadequacies are taken care of and further bugs aren't becoming readily apparent via genuine efforts at non-live testing. That said, personally, I don't much care if they want to continue plodding down this half-assed path, as long as VE remains opt-in and doesn't cause problems for non-testers. equazcion | 08:09, 30 Sep 2013 (UTC)
Like Equaczion said, I want them to fix bugs at Testwiki, and at every Wiki that actively welcomes VE and its updates, be it Mediawiki or another language version or Wikisource or whatever. I don't want them to push clearly untested updates to here and everywhere. I have left feedback at Mediawiki as well, but no one seems to care that their release notes are utterly wrong. I seriously doubt that anyone had tested any of the VE "improvements" of the version 19 before they implemented it at Mediawiki and announced the rollout here for this week, and I doubt as well that anyone but me has independently tested it but me (WMF developer Qgil has tested my bug reports and confirmed them, thanks for that). Why would we allow them to push their releases when their quality is way sub-par? Whether it will affect the opt-in, I don't know, since I'm unable to test that (and I doubt they have done). I don't see the value in rushing a new release every week when there are so many major bugs left and so little enthusiams here to test or use it, or to clean up after those that still use it. Fram (talk) 08:30, 30 September 2013 (UTC)
I'm afraid I don't see the point in this request. I think, although WMF should have withdrawn VE entirely after it was determined that there were serious encyclopedia-damaging bugs, that there's no point in withholding updates. There is no indication that WMF is likely to restore previously existing serious bugs, and I consider the possibility that an update adds new serious bugs likely, but unlikely to be detected without using a live Wiki. I could be wrong, — Arthur Rubin (talk) 09:57, 30 September 2013 (UTC)
Well, I tested the new release V19. There are, at the moment, a number of serious bugs in file handling (someone should really tell the WMF that it is about "files", not "images", since quite a few years), e.g. with the moving of files. Moving files works very poorly and causes more problems than it solves. So when this new version doesn't address this bug, but instead opens up this "functionality" for templates and other things, then yes, the WMF is actually expanding known bugs to new possibilities, and testing in their own pages at Mediawiki were more than sufficient to find these problems (and a few others to boot). Pushing these changes (with the accompanying incorrect description, which no one is allowed to correct apparently) to all wikipedia instead of testing them locally (at Mediawiki and Testwiki) is irresponsible behaviour and only makes Wikipedia worse. It isn't too bad for us, now that we have opt-in, but we could at least send the message that theit approach is totally wrong. Fram (talk) 10:05, 30 September 2013 (UTC)

Woah, woah, you want developers to spend their precious time testing software before releasing it? This is Web 2.0/"agile"/"waterfall"/insert buzzword here! Your users are your testers! Regression testing, test suites, fuzzing, what are those, some new websites or somethin? To be fair to the WMF, this problem is hardly limited to them. The trendy thing in software development these days seems to be letting your end users find the mistakes you made, and then bitching at them for not providing a ready-to-go patch to fix your errors, or if you're a large corporation potentially calling the police on them for finding bugs with security implications. After all "open source" means "other people do my work for me," right? I mean, who's got time to test software anymore? That's what they do when they write actually important software, right, like software that can kill people if it screws up (medical devices, military hardware, etc.), just throw the latest HEAD revision onto the device and ship it when the deadline comes. --108.38.191.162 (talk) 16:49, 30 September 2013 (UTC)

Cute, but even open-source developers only recommend live use after major issues are dealt with. The difference between the old and new-age open-source ways is that works in progress aren't "closed" quite as often, so instead of limiting them to a hired or even select group, they're often made available for all to test, should they wish to and be able to find the test release. Things still aren't generally implemented live or even put in plain sight until they're pretty reliable. By even the most relaxed industry standard, VE is in the alpha stage, and in order to participate in testing, users should need to have the requisite knowledge to navigate to a non-live testing ground. equazcion | 18:12, 30 Sep 2013 (UTC)
Oh I'm aware of that. I'm a programmer myself. My point was mainly in regards to the belief some people have that "open source" is magic pixie dust that will fix all the bugs in your software. This is more common in the corporate world, where some suits seem to think a legion of programmers will materialize to fix all your bugs for free. As for testing, if the WMF wants to learn how to do things right, they should look at projects like the Linux kernel, Perl, and distributions like Debian and Gentoo, all of which have automated test suites, groups of dedicated testers, and in fact require testers to sign off on code before it's marked for release. --108.38.191.162 (talk) 19:33, 30 September 2013 (UTC)
  • Fram, I get that you dislike VE and the way it's been deployed. That's great; it's certainly not a rare opinion! But this suggestion seems to be deliberately pointed - everyone else has to not be told about ongoing work because you dislike the project?
Given one of the major community complaints about VE, and every other major technical project, is a lack of communication, demanding less communication is going to be massively counterproductive and serve to make things just that bit less pleasant for everyone involved further down the line. Andrew Gray (talk) 19:45, 30 September 2013 (UTC)
    • No, that's not what I am suggesting. It's not that I don't want to hear about ongoing work (assuming that their reports are actually correct, the current one isn't), but that I don't want them to deploy the actual updates until a few months have passed, most of the major bugs have been fixed, and things have thoroughly been tested at testwiki and at any wiki that agrees to be used for that purpose (e.g. Mediawiki). It seems that I haven't explained my proposal very well, or that my current disagreement about the latest release notes have been mixed up with this proposal. But I repeat: I call here for a moratorium on deployments, further VE code updates, not for silent deployments. apologies for the misunderstanding! Fram (talk) 19:58, 30 September 2013 (UTC)
  • Fram, this is one of those cases of watching an inevitable result. Now that WMF has lost English Wikipedia as an involuntary testbed, their first few releases will inevitably represent quality degradations to the point that many of the projects that have it marked as "opt-out" will move to "opt-in". They will then get even less feedback, so their quality will degrade further. WMF will either have to learn to test code on their own, or the project will collapse. I'm hoping for the former. One way or the other, us raising a bigger fuss about it probably won't help.
I'm personally hoping that the VE development team starts to use WMF's QA team to evaluate releases (right now, my understanding is that they don't, something which doesn't surprise me but does sadden me). It probably is reasonable to ping Mr. Forrester and ask him what QA stages VE goes through. The last quote I have from him on the topic is "<+James_F> Elitre1: We test the fixes, yes, but clearly not enough. :-( We've spoken with the QA team about working with them, but currently we don't do any significant work with them, no."—Kww(talk) 20:05, 30 September 2013 (UTC)Pinging Mr. ForresterKww(talk) 22:49, 2 October 2013 (UTC)Pinging Mr. ForresterKww(talk) 06:23, 4 October 2013 (UTC)It's a reasonable and specific question, Mr. ForresterKww(talk) 03:59, 6 October 2013 (UTC)Perhaps it's some kind of bug in WP:Notifications, Mr. ForresterKww(talk) 20:05, 8 October 2013 (UTC)
@Kww: Hey, sorry, indeed Notifications doesn't let you just endlessly ping people to avoid people abusing it. I didn't see this until pinged by Scott. I talked in some detail about testing strategy at the IRC office hours last week - was there a specific question you wanted to ask? Also, per WP:DENY, please don't use a banned user's hate site as a reference, even if it seems reasonable to assume that the log is accurate. Jdforrester (WMF) (talk) 21:48, 8 October 2013 (UTC)
Reading IRC logs is always painful, so maybe I missed what I was looking for, Jdforrester. I saw mention of different kinds of tests, but I didn't see mention of who does the testing. When are changes specifically evaluated by someone that didn't write the code? Is there a specific set of regression tests that all releases must pass before they are released to a production version of WIkipedia?—Kww(talk) 22:10, 8 October 2013 (UTC)
@Kww: Changes are tested by the developer, the review, several computer (see the unit tests, the integration tests and the browser tests, as well as the dirty-diff testing) and by me. The test are run before, during, and after the code is merged. Yes, code cannot be merged (let alone deployed) without passing the unit and integration tests; we hope to be able to add the browser tests into this list when they are more stable). Jdforrester (WMF) (talk) 22:38, 8 October 2013 (UTC)
Note that they are currently hiring a full-time VE tester. Andrew Gray (talk) 20:17, 30 September 2013 (UTC)
It's either amusing or sad or both that among the "pluses" for a candidate to possess (but not requirements) are "Knowledge of wiki markup and experience as an editor of Wikipedia or another MediaWiki wiki" and "Not only know how to test VisualEditor but also why to test". Similar requirements were applied to the development team, I'm guessing. EEng (talk) 19:44, 6 October 2013 (UTC)
Shouldn't they have filled that position six months ago?—Kww(talk) 20:22, 30 September 2013 (UTC)
(ec)Any reason to believe whatever Jdforrester tells us? His track record isn't exactly spotless over the last couple of weeks... He asked for feedback about the status report, but can apparently not be bothered to read it or act on it. He has replied to my request about where he got the "edit summary" figures from, but his response (on his talk page) is seriously unconvincing. Fram (talk) 20:23, 30 September 2013 (UTC)
It's always good to at least know what the official statement is, regardless of whether you personally find it credible.—Kww(talk) 20:35, 30 September 2013 (UTC)
The feedback that the VE team/WMF should be getting is not on the level of bugs in this or that feature, but that they have failed in how they have managed this project. Until they have demonstrated some organizational changes, and demonstrated (on a suitable test bed) that they can deliver adequately tested software, they should stop deploying it. Alternately: the fuss they have gotten is the feedback. Do they need more? ~ J. Johnson (JJ) (talk) 21:28, 8 October 2013 (UTC)

User:Username/common.js page not creating

I seem to be unable to create the page User:Antiqueight/common.js - when I try I get the page up with no text box (everything else seems to be there) and when I click on save it seems to. But hours later there is still no page (and I purged cache). I was going to try out a script. I have Twinkle turned on. I don't know if this is related as when I try to go to the page above it has a note there pointing me to where to go to change the Twinkle preferences. I don't know where to look for answers so I thought I'd try here. No rush as I'm about to keel over for the night. But if you can HELP :-) --Antiqueight confer 22:48, 8 October 2013 (UTC)

What do you mean by "no text box"? Is there no edit box to type content into? It's not possible to create a page with no content. A page creation must contain at least 1 byte in order to save. PrimeHunter (talk) 22:54, 8 October 2013 (UTC)
Well - that was it. With wikiEd enabled there was no text box at all - so you are right - there was nothing to save (I hadn't thought of that). When I turned wikEd off the text box appeared. Thanks for the prompt. I went back and looked at the page and turned off everything I could think of and then up popped the text box. I've added the code and it all seems to work. I should probably leave playing with it til tomorrow!--Antiqueight confer 23:07, 8 October 2013 (UTC)
It looks like wikiEd isn't compatible with CodeEditor (which was recently deployed to WMF projects to add syntax hilighting and such to all code pages) - probably because CodeEditor just replaces the entire edit thing with its own UI. Not sure who would be able to fix/address this, but maybe if I mention it here someone else will magically know. -— Isarra 19:44, 10 October 2013 (UTC)
On T55914, Cacycle said that the versions of WikiEd as of a few days ago should be not activating on pages with CodeEditor. Have you tried bypassing your cache? Anomie 21:38, 10 October 2013 (UTC)

WMFlabs.org

Does anyone know why http://tools.wmflabs.org/enwp10/cgi-bin/list2.fcgi?run=yes&projecta=Medicine&importance=Top-Class&quality=C-Class is not working? The link was generated is from the table to the right. I was trying to help a WP:Student assignment for medical students know which articles they should pick for maximum-impact. In this case I clicked on the top-importance C-class articles. Thanks. Biosthmors (talk) pls notify me (i.e. {{U}}) while signing a reply, thx 09:57, 9 October 2013 (UTC)

Biosthmors Worked fine for me. Graeme Bartlett (talk) 10:42, 9 October 2013 (UTC)
Weird. Thanks Graeme Bartlett. So no one gets "Internal Server Error. The server encountered an internal error or misconfiguration and was unable to complete your request. Please contact the server administrator, [email protected] and inform them of the time the error occurred, and anything you might have done that may have caused the error. More information about this error may be available in the server error log. Additionally, a 500 Internal Server Error error was encountered while trying to use an ErrorDocument to handle the request."? I guess I should send an email, then? Biosthmors (talk) pls notify me (i.e. {{U}}) while signing a reply, thx 10:56, 9 October 2013 (UTC)
I get the same "500 Internal server error", so you're not alone. Fram (talk) 11:08, 9 October 2013 (UTC)
I am getting that error now. Also on other tools like https://tools.wmflabs.org/copyvios/. So it must be intermittent but widespread. Graeme Bartlett (talk) 11:29, 9 October 2013 (UTC)
I sent an email and linked to this discussion. Biosthmors (talk) pls notify me (i.e. {{U}}) while signing a reply, thx 12:47, 9 October 2013 (UTC)
  • The issue is known, and should be fixed today. The problem is that Tool Labs has been more popular faster than expected and some less-well behaved tools are eating up resources and breaking other tools. I'm going to put some partitioning in place to help alleviate the issue. — MPelletier (WMF) (talk) 13:23, 9 October 2013 (UTC)
Thanks, MPelletier. Can you please post here when you believe the problem is fixed, so that we can test? I'm seeing the same Internal Error since yesterday when trying to use a saved Catscan2 bookmark. – Jonesey95 (talk) 17:40, 9 October 2013 (UTC)

Looks like the server is in trouble again. I have just been trying http://tools.wmflabs.org/catscan2/catscan2.php, but there is no connection. No response either at http://tools.wmflabs.org/ --BrownHairedGirl (talk) • (contribs) 23:17, 10 October 2013 (UTC)

It's planned maintenance that turned out to take much longer than was planned. See http://lists.wikimedia.org/pipermail/labs-l/2013-October/001748.html. Anomie 01:10, 11 October 2013 (UTC)
Catscan is probably another issue. Nikkimaria (talk) 18:08, 11 October 2013 (UTC)

Updating Dashboard

I attempted to update WP:Dashboard by removing historical pages and adding The Teahouse, but my edits were reverted by Legobot (see here). Help‽ ʍw 13:11, 9 October 2013 (UTC)

Chances are Legobot needs to be informed of which boards to include/exclude, and simply removing them from the page will just trigger an "update" where it puts things back the way it thinks they should be. I'm sure @Legoktm: can fix it, and perhaps modify the behavior so that we can include/exclude via templates on the dashboard pages? equazcion | 13:58, 9 Oct 2013 (UTC)
  • Perhaps create wp:Dashboard/new for progress: Until all the bugs in Legobot can be fixed, try updating wp:Dashboard/new to make upgrades without artificial restraints, and then other users can read that page to review the enhancements. Per the "wiki" concept, nothing derails progress like pages which require permission to update. Also it is not just problems with Legobot, but many Bots have trouble, such as date-bots removing date links from calendar(!) pages, where any fool would know a page with 365 date links is probably not "over-linked" for dates. Perhaps Bots should be called "idiOTs". However, even intelligent people can write bad software due to all the complex factors to consider. Meanwhile, focus on a better version, as wp:Dashboard/new, until the bugs in Legobot can be corrected. We are months behind in updates to valuable tools, due to being sidetracked by the planned VE disruption fiasco. -Wikid77 (talk) —Preceding undated comment added 20:20, 9 October 2013 (UTC)
    • I'm actually rather appreciative myself of those who maintain bots, as they provide some pretty important behind-the-scenes stuff that would otherwise be a pain. Especially Legoktm, who took it upon himself to take over for more than several retired bots. Developers tend to get it in the arse from the audience because there are always flaws, no matter how much care a programmer tries to devote, and this particular problem is really nothing to get bent out of shape about. equazcion | 04:15, 10 Oct 2013 (UTC)
@64.40.54.7: Makes sense.
@Equazcion: Your first comment is pretty much what I thought. It might be nice to be able to add or remove noticeboards through the templates at the dashboard, though it's certainly not a necessary feature (it's not that often that new noticeboards are created or old ones are closed). If such functionality is difficult/impossible to add, there should at least be a notice that all updates and modifications to those pages must go through Legoktm (although I half-agree with Wikid77 that that's a bit contrived).
@Wikid771: By all accounts, this it not the result of "bugs" in Legobot, though it is an unusual procedure to have to go through to update the pages. The biggest problem here seems to be that Legotktm has not been being notified of the needed updates to the templates/bot (until now). I have no idea what you're getting at with "WP:Dashboard/new"; what would that be, a sandbox? We're not talking about sweeping upgrades and modifications to everything at the dashboard, just removing links to defunct noticeboards and adding a new one. There really isn't that serious of a problem with the pages as they are, they're just a little bloated with all the links to historical boards (the most serious problem is the lack of a link to the Teahouse).
ʍw 12:20, 10 October 2013 (UTC)
Resolved
See here. ʍw 22:46, 10 October 2013 (UTC)

Inserting tables

The toolbar above the edit box has a button to insert a table, with options for header row, style with borders, making it sortable, and number of rows and columns. I suspect that most new tables added to articles must be made from it. The header row is in a slightly darker grey color.

But shouldn't the first column be a header column as well, as detailed Wikipedia:Manual of Style/Accessibility/Data tables tutorial#Overview of basics? Or at least included as another option in the gadget? Cambalachero (talk) 14:46, 10 October 2013 (UTC)

The section you linked to is just documentation on how to make header rows and columns, among other things, rather than a recommendation. You'll notice most of the table examples on that page lack a header column, and from experience I'm pretty sure most article tables don't have one (although I could be mistaken). equazcion 15:51, 10 Oct 2013 (UTC)

timeouts

Anyone else getting timeouts on log pages for administrators? Special:Log/Drmies and Special:Log/Kww both get 504 errors for me. If I choose a non-admin, say Special:Log/Technical_13 or Special:Log/PantherLeapord, I get a response. — Preceding unsigned comment added by Kww (talkcontribs) 19:32, 4 October 2013 (UTC)

I haven't seen a timeout, but I have noticed that viewing my log takes a lot longer than usual. ​—DoRD (talk)​ 20:04, 4 October 2013 (UTC)
Kww, yes. Long page loads in general, too. Killiondude (talk) 20:10, 4 October 2013 (UTC)
Hmm, I can get each individual log, but I can't get "all logs". I wonder if there's just a raw capacity limit. I've been an admin for a while, so I have a healthy logbook, but I wouldn't think that either me or Drmies would be record-seting. DoRD, when you say you didn't get timeouts, do you mean that you are able to get your own, or that you are able to see mine?—Kww(talk) 20:36, 4 October 2013 (UTC)
I was talking about my log, but now it is timing out as well :\ ​—DoRD (talk)​ 21:03, 4 October 2013 (UTC)
Add me to the list - I can't get Drmies's or mine. Peridon (talk) 09:50, 5 October 2013 (UTC)
I'm able to get DoRD's logs after about a 2-3 minute delay. Drmies and Kww give me the timeout. equazcion | 09:59, 5 Oct 2013 (UTC)
I just got Peridon's logs after a delay similar to DoRD's. equazcion | 10:03, 5 Oct 2013 (UTC)
I just got DoRD's after a long wait, but mine timed-out. Peridon (talk) 12:00, 5 October 2013 (UTC)
Me too. Using the limit parameter doesn't help. There seems to be an issue somewhere :) -- zzuuzz (talk) 13:18, 5 October 2013 (UTC)
Hopefully fixed in gerrit:87168. Legoktm (talk) 18:52, 5 October 2013 (UTC)
Still getting 504 for Drmies. Peridon (talk) 10:08, 6 October 2013 (UTC)
So the commit I linked above has been merged into MediaWiki core, but not deployed it, so the errors will still occur. It should be deployed to enwiki in less than 2 days. Legoktm (talk) 01:17, 9 October 2013 (UTC)
Thank you it seems to be working now. -- zzuuzz (talk) 05:30, 11 October 2013 (UTC)
Yes, for me too. JohnCD (talk) 11:31, 12 October 2013 (UTC)

Cite error formatting

The bottom of Talk:Miller–Rabin primality test currently displays the weird notice . While the cite error is real, it should not be processed by MathJax.—Emil J. 16:07, 10 October 2013 (UTC) [EDIT: Since many people were confused by an unrelated error affecting the <math> markup I used to depict the error message, I replaced it with a screenshot. Also, let me stress that the notice is only visible with MathJax enabled.—Emil J. 13:22, 11 October 2013 (UTC)]

Narrowing it down, <strong class="error">a b c</strong> also gets processed by MathJax. Something is wrong either with MathJax setup, or with CSS for this element-class combination.—Emil J. 16:16, 10 October 2013 (UTC)

Looks like that might be this as reported above. I'm not always seeing the error in your post on every load of this page, but intermittently. equazcion 16:24, 10 Oct 2013 (UTC)
I’ve only put the message in <math> tags here to imitate what I see at the linked talk page, but that’s not how it is actually coded. The bug you mention affects PNG output, whereas the misformatted error message I’m talking about only shows up if you enable MathJax (otherwise it is invisible).—Emil J. 16:48, 10 October 2013 (UTC)
I actually do see the error here intermittently, despite your coding choice and my having the default preference chosen. I also note that whenever a page load here shows the error in this section, it also shows up in the section above -- and when the error does not show up here, it does not show up above either. equazcion 16:51, 10 Oct 2013 (UTC)
It looks to me as if the issue is at least partly caused by copying a quote from the article page, including a ref tag, with no reflist on the talk page. I have commented out the ref tag. DES (talk) 17:04, 10 October 2013 (UTC)
Sigh. I intentionally left the ref unfixed so that the error message is visible. The problem I’m reporting is about the error message, not about the ref that triggered the message. The error message needs fixing whether this particular instance of the error is corrected or not.—Emil J. 17:13, 10 October 2013 (UTC)
Fine I will remove the comment symbols. I would have had no problem if you had done so. DES (talk) 17:39, 10 October 2013 (UTC)
(edit conflict) Ah okay I think I misunderstood. EmilJ is saying that even when the error doesn't show up, this ref shouldn't be getting formatted as a math expression (at least I think that's what they're saying). Sorry about that. I'm not very familiar with math coding and how references within them are supposed to be handled, so I'll let someone else comment on this. equazcion 17:06, 10 Oct 2013 (UTC)
No. I’m saying that when the error does show up, the error message shouldn’t get formatted as a math expression.—Emil J. 17:08, 10 October 2013 (UTC)
(ec) That makes sense, i failed to read the above fully I think. I agree that even where there is a cite error, it shouldn't be parsed as part of a math expression. DES (talk) 17:11, 10 October 2013 (UTC)
Although it seems to me that whatever text a ref produces will get processed based on the tags surrounding it, no? It could be probably be gotten around by changing the CSS to override purely in cases of errors, or for all refs if that's what people want, but I don't think this is actually a bug. equazcion 17:18, 10 Oct 2013 (UTC)
User:Trappist the monk might have some insight. equazcion 17:21, 10 Oct 2013 (UTC)
Missing reflist errors are only emitted after the whole page is processed, so they cannot depend on tags surrounding <ref>’s even if they wanted. Anyway, there doesn’t seem to be anything unusual about how the message is formatted in the HTML, as you can see for yourself in the HTML source (note that the message is always there, even if it’s made invisible by the default CSS). As I already wrote above, the problem can be triggered anywhere by <strong class="error">a b c</strong> a b c. By default, this shows up in boldface red (as expected), but with MathJaX enabled, it is processed as if it were a math expression (making it show up in a math italic font without spaces, black, or worse if the text happens to include LaTeX commands).—Emil J. 18:12, 10 October 2013 (UTC)
With the default PNG preference, I'm seeing the error text as a PNG, so it would appear this isn't dependent on mathjax, and that the math tags are processed after the ref tags are expanded. I'm not sure if a ref tag would ever realistically expand into something containing latex code, but I guess that's something to consider. I'm not sure what you mean exactly about the span tag -- if you're saying spans classed with "error" are processed as math even without math tags, the following code: <span class="error">a^{n-1}=1\pmod n</span> displays as ordinary red text with no math, at least for me, no matter which preference I have chosen (I tried both): My mistake, you said strong tags, not span tags -- with mathjax enabled, strongs classed with "error" do indeed show math: a^{n-1}=1\pmod n. Interesting... equazcion 18:38, 10 Oct 2013 (UTC)
There are no math tags anywhere close to the ref in question. Are you really saying that if you go to Talk:Miller–Rabin primality test, and scroll to the bottom of the page, there is a PNG containing the error message that I copied above? That certainly did not happen in my tests.—Emil J. 18:52, 10 October 2013 (UTC)
No, actually, I was about to post another correction and edit conflicted with you: I only see a PNG when there is no ACTUAL error -- the PNG I referred to is actually from the PASTED error text you posted (which makes this confusing :) ). When the actual red error shows up and my preference is set to PNG, the error shows as bold red text, NOT as an image -- so you were right about that as well. equazcion 18:55, 10 Oct 2013 (UTC)
I don't think I have much to contribute to this conversation. The error message at the top of this section has, except for one brief instance, always appeared as bold red text. It did briefly appear as italicized black text in a .png image but if I backtrack through the page history I can't see a black italic .png version. I have not seen the error message in either guise at Talk:Miller–Rabin primality test.
Even though it's apparent that this problem has nothing to do with citations, I would add that whenever including references in a talk page, it's appropriate to use {{reflist-talk}} and set |close=1.
As I preview this edit I'm seeing the error message as black italic .png.
Trappist the monk (talk) 19:10, 10 October 2013 (UTC)

Beware intermittent loss of CSS format styles

There have been problems this week with file-cache failures for bits.wikimedia.org or upload.wikimedia.org (for the math-tag png cache images), and in some cases, none or only partial CSS styles have been applied in pages. Hence, it is possible that a math-tag style was active in some cases, as when the cite-references error message was formatted. Some users have noted to refresh the page several times and take-your-pick of the results. Unfortunately, with all the hundreds (thousands?) of new CCS classes in span-tags, sometimes failing to process in the browser, the overall interface to Wikipedia has become "WYSIHUH?" or "WYSIWTF?" and perhaps we should write an essay "wp:WYSIHUH" to explain the convoluted results which can often appear in pages. -Wikid77 (talk) 05:50, 11 October 2013 (UTC)


Right-to-Left failing

When trying to see the next 200 pages in a category in a right-to-left language, it fails. It was working as recently as a few weeks ago.

See: [7] and click on the next 200 button - for non-Farsi readers using Chrome with translation on will assist you. Carlossuarez46 (talk) 21:21, 10 October 2013 (UTC)

must be something transient - tried it just now (and also with "uselang=en", "uselang=he", and "uselang=fa"), and found no problem. can you try it when logged out? if logging out solves the problem, it may be something in your account - maybe some preference, a gadget, or something in your "common.js" or <skin>.js. peace - קיפודנחש (aka kipod) (talk) 06:00, 11 October 2013 (UTC)

Can't access tools.wmflabs.org

RESOLVED: Was offline maintenance, back now. -Wikid77 14:40, 11 October 2013 (UTC)

I just clicked "Articles created" at the bottom of my Special:Contributions page and got this: "Unable to connect. Firefox can't establish a connection to the server at tools.wmflabs.org." --Anthonyhcole (talk · contribs · email) 05:38, 11 October 2013 (UTC)

  • Perhaps re-try in 30 minutes: Still stuck 25 minutes later. There are still some occasional connection failures with wp:WMFLabs, even though the system is much faster than the old toolserver system, but I have found a re-try within 15 minutes often works, as compared to multi-hour waits to re-connect with the old toolserver. Perhaps there is a current WMFLabs upgrade in progress now. -Wikid77 (talk) 06:05, 11 October 2013 (UTC)
The servers are undergoing maintenance, and it's taking longer than expected. See #WMFlabs.org above. – Jonesey95 (talk) 06:12, 11 October 2013 (UTC)
  • WMFLabs back online again: I just confirmed the "articles created" button works again under Contributions. Perhaps there might be a better hour to run server maintenance, to allow extra time. -Wikid77 14:40, 11 October 2013 (UTC)

Are there statistics for access to articles by direct Wikipedia search?

This question was asked on the Help Desk.— Vchimpanzee · talk · contributions · 19:15, 11 October 2013 (UTC)

VisualEditor weekly update - 2013-10-10 (MW 1.22wmf21)

Hey all,

Here is a copy of the weekly update for the VisualEditor project. This is to make sure you you all have as much opportunity to know what is happening, tell us when we're wrong, and help guide the priorities for development and improvement:

VisualEditor was updated as part of the wider MediaWiki 1.22wmf21 branch deployment on Thursday 10 October. In the week since 1.22wmf20, the team worked on fixing bugs and stability improvements to VisualEditor.

When you delete or backspace over a node (like a template, reference or image), the node will first will become selected before a second press of the key will delete it, making it more obvious what you are doing and avoiding accidental removals of infoboxes and similar (bug 55336).

If you hold down the ⇧ Shift key whilst resizing an image, it will now snap to a 10 pixel grid instead of the normal free-hand sizing. A number of improvements were made to the transactions system which make the undo/redo system able to cope better with real-time collaboration, where multiple users will be able to edit a page at the same time in one session.

The save dialog was re-written to use the same code as all other dialogs (bug 48566), and in the process fixed a number of issues. The save dialog is re-accessible if it loses focus (bug 50722), or if you review a null edit (bug 53313); its checkboxes for minor edit, watch the page, and flagged revisions options now layout much more cleanly (bug 52175), and the tab order of the buttons is now closer to what users will expect (bug 51918).

The code for the action buttons on the right (RTL environments: left) of the toolbar was re-written slightly to improve its flexibility. The display of the help and edit notice menus is now improved, including the addition of a close button (bug 52386). The width of the format drop-down was made adjustable so that long labels don't cause it to break (bug 54870), and a bug that caused the toolbar's menus to get shorter or even blank when scrolled down the page in Firefox is now fixed (bug 55343).

A complete list of individual code commits is available in the 1.22/wmf21 changelog, and all Bugzilla bugs closed in this period are on Bugzilla's list.

Following the regular MediaWiki deployment roadmap, this should be deployed here (for opted-in users) on Thursday 17 October.

Hope this is helpful! As always, feedback on what we're doing is gratefully received, either here or on the enwiki-specific feedback page. Please ping me using {{ping|Jdforrester (WMF)}} to make sure I see it promptly if you have any thoughts or corrections.

Jdforrester (WMF) (talk) 21:03, 11 October 2013 (UTC)

version 1.22 wmf20 - miniscule breaking change

so in version 1.22 wmf20, the JS variable wgVectorEnabledModules is no more. this is not a huge deal - this variable was marginally useful when "vector" skin was deployed as an extension, and some of its modules (specifically "collapsiblenav") were considered optional: this variable indicated wheterh a specific module was installed.

since then, vector was integrated into core, its modules are no longer "optional", and this variable became superfluous.

there is nothing in mediawiki namespace that makes use of this defunct variable, but there are some 30-odd pages in userspace that seem to refer to it, maybe half of them are legit scripts. some of these are broken regardless, but at first glance it looks as if the removal of this variable have borked a single-digit number of scripts in userspace.

so if something that used to work for you stopped working, you may want to check to see if you are using one of those scripts, and if so, suggest to the script owner to remove the reference to this defunct variable. peace - קיפודנחש (aka kipod) (talk) 21:33, 11 October 2013 (UTC)