Bug 47686 - Setting SSI variables appear to assume ISO-8859-1 character encoding
Summary: Setting SSI variables appear to assume ISO-8859-1 character encoding
Status: RESOLVED FIXED
Alias: None
Product: Apache httpd-2
Classification: Unclassified
Component: mod_include (show other bugs)
Version: 2.2.13
Hardware: All All
: P2 regression with 6 votes (vote)
Target Milestone: ---
Assignee: Apache HTTPD Bugs Mailing List
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2009-08-12 05:34 UTC by Björn Wiberg
Modified: 2010-09-19 09:03 UTC (History)
4 users (show)


Attachments
Testcase that shows the problem, SSI has to be enabled! (214 bytes, text/plain)
2009-11-02 11:34 UTC, Rafael Gattringer 		Details
View All Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description Björn Wiberg 2009-08-12 05:34:34 UTC 
Setting an SSI variable to a value containing UTF-8 characters and then echoing it yields output with HTML entities for each of the two bytes of each UTF-8 character.

E.g.:

<!--#set var="DEFAULT_APPLICATION_DESCRIPTION" value="den CAS-tjänst du besökte" -->
<!--#set var="APPLICATION_DESCRIPTION" value="${DEFAULT_APPLICATION_DESCRIPTION}" -->
Du &auml;r nu utloggad ur <!--#echo var="APPLICATION_DESCRIPTION" -->.

...yields (in browser):

Du är nu utloggad ur den CAS-tjänst du besökte.

...instead of (in browser):

Du är nu utloggad ur den CAS-tjänst du besökte.

...and in HTML:

Du &auml;r nu utloggad ur den CAS-tj&#195;&#164;nst du bes&#195;&#182;kte.

...instead of in HTML:

Du &auml;r nu utloggad ur den CAS-tjänst du besökte.


This was not so in Apache 2.2.11.

The .shtml page is in UTF-8 (with BOM) and has:

<meta http-equiv="Content-Type" content="text/html; charset=utf-8" />

In httpd.conf, AddDefaultCharset is NOT set (so uses default value).
Comment 1 Ruediger Pluem 2009-08-12 11:51:56 UTC 
I guess this is caused by the fix for PR25202 (r730296, r732583). This might be tricky to fix.
Comment 2 Joe Orton 2009-08-13 03:34:59 UTC 
The mod_include docs do clearly state that, for echo,

a) encoding=entity is the default, and
b) encoding=entity will not work correctly if a character encoding other than ISO-8859-1 is in use

For the example given here, at least, using 

<!--#echo encoding="none" var="..."

would be the correct use, surely?

But the potential for breaking backwards compat in bug 25202 does seem to have been clearly called.  Iwan Stanley's approach in bug 25202 comment 7 seems vastly preferable to the change which got committed.
Comment 3 Anders Kaseorg 2009-08-17 07:48:05 UTC 
(In reply to comment #2)
> For the example given here, at least, using 
> <!--#echo encoding="none" var="..."
> would be the correct use, surely?

That doesn’t work, though, when the variable may contain <>&".
Comment 4 Rafael Gattringer 2009-11-02 11:34:05 UTC 
Created attachment 24464 [details]
Testcase that shows the problem, SSI has to be enabled!
Comment 5 Rafael Gattringer 2009-11-02 11:41:24 UTC 
I would like to confirm this problem. See my attached SSI file for an example.
Comment 6 Graham Leggett 2010-09-19 09:03:34 UTC 
Fixed on trunk in r998651.