Page MenuHomePhabricator

Broken Unicode links to Wikipedia at Commons
Closed, ResolvedPublic


Author: ausir

There's a problem at Wikimedia Commons with linking to Wikipedia articles with
Unicode characters. For example, a link to [[w:pl:Stanis%C5%82aw Lem]] links to
[[w:pl:Stanis]] instead. External links like
[ Stanis%C5%82aw Lem]] can of course be
used, but it'd be better if they were no-arrow interwiki links.

Version: unspecified
Severity: normal



Revisions and Commits

Related Objects

Event Timeline

bzimport raised the priority of this task from to High.Nov 21 2014, 7:08 PM
bzimport set Reference to bz563.
bzimport added a subscriber: Unknown Object (MLST).
  • Bug 645 has been marked as a duplicate of this bug. ***

From duplicate bug 645:

Dear friends,

at the
first remark named "link to [[w:ro:Discuţie
Utilizator:Gangleri]] ~ [[w:ro:User_talk:Gangleri]]" ilustrates
that some special characters as "ţ" cause problems in InterWiki

If this issue is already known please let me know where to read
about such "exceptions".

It maight be somehow related to too because it
shows the same effect.

Best regards Reinhardt

Looks like a problem with the way the redirect is handled through

jimes wrote: has a link "[[Juraj Beneš]]"
which displays correctly on that page as "Juraj Beneš" links incorrectly to
"Juraj Beneš" ("Juraj_Bene%C5%A1").

gangleri wrote:

I made a central place [[Wikipedia:Invalid article names]] in
[[:Category:Wikipedia maintenance]] where such broken links can be listed.
This is an equivalent to [[Wikipedia:Duplicate articles]]. I know some
others. I suppose they have been created with previous versions of WikiMedia
Regards Reinhardt

gangleri wrote:

Notes: some weeks ago I made some remarks at
rks#Invalid links (lists)]]. You may read some other related sections as well.

The problem is both the existence of pages which do not complay wit UTF-8 (?)
and the usage of such links in en.wikipedia and directed to en.wikipedia from
other projects.

It should be much easyer to verify the namespace then using a bot to scan for
such links inside the articles, talks, categories ... .
Regards ~~~~ (Reinhardt)

gangleri wrote:

Dear friends,

you will find two sort of links:
a) [[:ro:Discuţie Utilizator:Gangleri]],
:Gangleri|bg]], [[:he:משתמש:Gangleri|he]],

b) [[w:ro:Discuţie Utilizator:Gangleri]],
;:Gangleri|bg]], [[w:he:משתמש:Gangleri|he]],

Both work at

ONLY a) work at, ...

Regards Reinhardt

gangleri wrote:

Dear friends,

a) Do we need a seccond test environment?
b) Should InterWiki translation should be be fixed in 1.4-cvs?
shows that we are testing on an Unicode environment.

This means that issues related to InterWiki translations trough
en.wikipedia CAN NOT BE TESTED HERE.

My proposal:

Please make an environment and make the translations
from [[test:]] to [[xx:]] as [[🇫🇷]], [[:pl:]], [[:ro:]], [[🇷🇺]],
[[:he:]], [[:ja:]], [[:bg:]] ... trough because
problems regarding translations to those targets (except [[🇫🇷]]?) are

There are a lot of bugs which can be included then:

  • bug 563,
  • translations related to the anchor part of a link, where the anchor

contains special characters as (, ), ", ', Unicode characters and so on.

  • ???

Regards Reinhardt

gangleri wrote:

I was thinking on this again. It seems that three test environments are
needed in order to fix this / to emulate all combinations of final

B through which translations are made;
A and C in order to test:

  • A translated trough B to C
  • C translated trough B to A
  • A translated trough B to B
  • C translated trough B to C
  • B translated to A
  • B translated to C

If B will be a UTF-8 environment as it is now in 1.3.7. B should be an UTF-
8 too.

Regards Reinhardt

kjell.andre wrote:

(In reply to comment #3)

Looks like a problem with the way the redirect is handled through

When testing a few variants of interwiki links you can see that:

: [[w:pl:Stanis%C5%82aw Lem]] and [[:en:pl:Stanis%C5%82aw Lem]] does not work.
: but [[:pl:Stanis%C5%82aw Lem]] and [[de:pl:Stanis%C5%82aw Lem]] does.

Obviously the first (or only) prefix determines how the name is interpreted. If it is a language prefix for a UTF8
based wiki it works since the name is valid for that wiki, if it is a a prefix for a Latin-1 based wiki, the name is
not valid for that wiki and is instead cut of at the invalid character and hence a broken link.

To correct this, the normalization of names must be based on the last (or only) language prefix, not the first.

Best regards, Kjell ANDRÉ

Now that bug 65 is fixed this was pretty easy to add on top. Checked in and put live.

epriestley added a commit: Unknown Object (Diffusion Commit).Mar 4 2015, 8:20 AM