6

I'm working on a system that uses UTF-8 characters in folder names for URLs. There's been no problem in navigating to these URLs and everything works as expected - except when issuing a redirect to another page on the site; whereupon the browser seems to encode the extended characters.

To give an example, I'm attempting to redirect to the following relative URL:

/geschäft/käfer/

If I visit that URL directly in the address bar, there's no problem. However if I change the location header to redirect the browser to this URL, it ends up at:

/gesch%E4ft/k%E4fer/

If I look in the response headers for the original page (It's a 301 redirect to the translated content) I can see this entry:

Location:/geschäft/käfer/

It seems the correct details are ending up in the header, but the browser's address bar is showing the encoded value with %E4 detailed above. I have attempted various ways of entering the URL into the location header, but all come out with the same result.

I am seeing this behaviour on Chrome 37.0.2062.120 m and on Firefox 32.0.2.

This is running on a dev box, Windows 7 Home with IIS7.5

EDIT: It seems this issue could be related directly to ColdFusion. If I use Javascript to redirect to the url, this works... with the caveat that the file has to be saved with a BOM. If I use cflocation, or if I use the pagecontext to insert the header manually, the issue persists regardless of whether or not a BOM is present.

I've also noticed a similar problem with using cfinclude in that these extended characters are displayed incorrectly unless the calling template is saved with a BOM.

Gary Stanton
  • 1,435
  • 12
  • 28

4 Answers4

3

I went to test this and didn't see the same result. But then I played with it a little more and tried using 

<cfprocessingdirective pageencoding = "utf-8"/>

Immediately, I was able to see the exact same issue that you did. It seems natural to include it in any page. This is very speculative, but CFAS may be doing some kind of URL encoding in the cflocation tag when used in conjunction with the pageencoding directive.

Assuming you have this in your code somewhere, try removing it for the redirect. If that works, then I would report this as a bug to Adobe.

Just FYI, I did this -- Output with encoding

<cfprocessingdirective pageencoding = "utf-8"/>
geschäft/käfer/

And I got 

geschäft/käfer/

But when I did this -- Relocation with encoding

<cfprocessingdirective pageencoding = "utf-8"/>
<cflocation url="geschäft/käfer/" addtoken="false" />

It relocated me to 

gesch%E4ft/k%E4fer/

And when I did this --Output without encoding

geschäft/käfer/
<cfabort>

I got 

geschäft/käfer/

But when I did this -- Relocation without encoding

<cflocation url="geschäft/käfer/" addtoken="false" />

Then I was relocated to 

geschäft/käfer/
Gerry Gurevich
  • 936
  • 6
  • 12
  • Amazing! That does indeed work... I have UTF-8 set all over the place as it's an internationalised site, so I actually have to explicitly add `` to the top of the page, and save it without a BOM, to get it to work... but work it does. This seems like it must be a bug... surely! Thanks for the answer anyway. I've been trying to resolve this for hours now! – Gary Stanton Sep 23 '14 at 13:07
  • 1
    I've submitted this as a bug. https://bugbase.adobe.com/index.cfm?event=bug&id=3830375 – Gary Stanton Sep 23 '14 at 13:33
  • Oddly enough, I still can't get this to work when the URL has been returned from a database... If I manually enter the text `geschäft/käfer/` I can get it to work as detailed above - with the iso-8859-1 setting - but if that exact same text is in a query object that has come from my database call, even though it outputs correctly on the screen, it _will not_ redirect correctly, no matter what I try. *sigh* – Gary Stanton Sep 23 '14 at 16:37
  • I'm going to guess the new problem is related to the encoding that goes with the query processing page. I think that the first cfprocessingdirective encountered is the one that holds. I started playing with GetPageContext().forward(); but I can't figure out how that works. It requires a relative path but mine seems to be relative to the CFIDE folder. An alternative might be to set up a translation of the encoding in the onMissingTemplate method of application.cfc (or the missing template hanlder if you are doing this server wide). – Gerry Gurevich Sep 23 '14 at 19:39
  • I'm not sure the issue is with the cfprocessingdirective in this instance. This is just on a test page and I can affect the change with the cfprocessingdirective tag, but not if the content comes from the DB. Having said that, the DB call itself is in a function so perhaps inlining the query might make the difference. I'll give that a go tomorrow. As it happens, all of this is intended to run in the onMissingTemplate method! I can't hardcode the values though, there are tens of thousands of URL possibilities across multiple languages. – Gary Stanton Sep 23 '14 at 20:11
  • So Adobe have unceremoniously and with absolutely no explaination whatsoever, refused to fix this bug. I'm furious. – Gary Stanton Dec 03 '15 at 15:44
2

I've tried the above but it couldn't get it to work. I ended up doing the cflocation "manually". Like this:

<cfprocessingdirective pageencoding = "utf-8"/>
<cfheader charset="utf-8" name="location" value="geschäft/käfer/">
<cfheader statuscode="302">

This worked like a charm for me.

birdy1980
  • 317
  • 1
  • 10
  • Thanks for sharing. Note for others, `cfprocessingdirective` is only needed if the special characters are hard coded into the script, like in this example. – SOS Dec 07 '17 at 17:27
1

This is still broken with ACF2018 - the CFHeader work-around does the trick, but... yuck..

It appears that Lucee (as of 5.3.3.62) also has the same problem.. I've reported it to them as well (LDEV-2456), we'll see what they have to say about it.

Tim Parker
  • 11
  • 2
1

Another potential work around for this is to URL Encode the portions of the URL that cflocation will break.

For example, I have some conditional testing that will 301 redirect if the combination of parameters are known to not work well together. One of those parameters is composed of Greek characters. Our solution is to use URLEncodedFormat() where needed.

<cfif Translation EQ "LXX" AND (URL.ot EQ "MGNT" OR URL.ot EQ "TR")>
    <cfset URL.word = URLEncodedFormat(URL.word)>
    <cflocation statuscode="301" url="/lang/lexicon/inflections.cfm?strongs=G#myStr.StrongsNum#&t=#URL.ot#&ot=#URL.ot#&word=#URL.word#" addtoken="No" />
</cfif>
Dan Davis
  • 524
  • 5
  • 9