No.

I updated the patch to not contain {{eq}} anymore, which seems to be what everyone but Base wanted.

It's the template that's broken, not the duplication detection code. Passing a parameter with a number for a name doesn't cause that number to be skipped. Look at https://en.wikipedia.org/wiki/User:Jackmcbarn/positional and https://en.wikipedia.org/wiki/User:Jackmcbarn/positional_template for an example.

There's an unclosed <nowiki> tag on that page immediately after "També després vaig pensar amb el".

Yeah, that was me. The content model on that page needs fixed back to wikitext.

Use mw.title.new('whatever'):getContent() to get the text in a non-hacky way.

We already have mw.text.jsonDecode to decode JSON.

You can't edit that page because it doesn't exist and you don't have permission to create it. The Lua error is a red herring; you only see it because someone somehow broke MediaWiki:Noarticletext on that wiki.

mwInit isn't really intended to be loaded via require(). (Neither is most of the other stuff in lualib, now that I think about it.) Perhaps we should move libraryUtil, bit32, and luabit to a different directory and have that be the directory require() can see.

In T49137#1463102, @MZMcBride wrote:

• We already allow transclusion of Special:PrefixIndex.

Although Special:PrefixIndex can be transcluded, its contents are stripped, meaning that modules can't parse it.

I think the issue here is that if you move a page over a redirect, the old redirect is gone forever, rather than being moved to the archive table.

gsub returns 2 values and trim takes 2 parameters. It would be surprising if it threw the second one away for no good reason.

Wouldn't it make more sense to have pages list the gadgets they want to run on them, rather than gadgets listing the pages they want to run on? For example, https://en.wikipedia.org/wiki/MediaWiki:Gadget-switcher.js currently loads and runs on every page, but only does stuff if the page contains an element with class "switcher-container". If pages chose their gadgets, then pages that needed that could put something like {{LOADGADGET:switcher}} somewhere on them. It's not possible for that gadget to choose its pages, since any page could have a switcher-container element added to it, and it wouldn't be practical to update the gadget each time.

@Anomie I don't see how it exposes the Lua part of it.

I think https://gerrit.wikimedia.org/r/#/c/218116/ is a better way of accomplishing the same goal.

Indeed, that does appear to be it. Since we can't fix an IE bug (and can't even work around it in this instance), I'm closing this.

The fix is part of 1.26wmf8, which will be deployed to all WMF wikis by June 3rd.

Okay, this is trickier than I thought. The problem is that the easy way of fixing this will result in this breaking:

Template:Foo:
Something{{{1}}}. Something else{{{1}}}.

I see what needs to be done here. I'll get a fix written tomorrow.

That one is a much better example of the issue. It definitely needs a fix.

This is indeed new, caused by I0e136f952302b65f1abeeac5273bd842867e1dac. I don't consider this a big deal though, since something that contains a reference should never be used as an element's id anyway.

In general, we don't add things to core unless we expect that most of our users will use them.

What's the use case for this?

I don't think there's a bug here. It looks to me like the user tried to make the edit logged out, the filter stopped them, and they then logged in and successfully made the edit.

You're giving two backslashes on the end of the token. There's only supposed to be one. You probably did this because you copied and pasted the JSON quoted string without unescaping it.

@Anomie Lua actually does something a little more subtle than what your fix did:

=string.find('foobar', '[]]')
nil
=string.find('foo]bar', '[]]')
4	4
=mw.ustring.find('foo]bar', '[]]')
Lua error: malformed pattern (missing ']').

Apparently, if a ] is the first thing in the pattern, it considers it a regular character in the pattern.