Use the pure perl Data::Dumper when generating setup files to ensure that utf-8 characters are written out as such, and not as the encoded perl strings the C Data::Dumper produces.

Note that the text produced by the C version was interpreted fine
when ikiwiki loaded the setup file. But it was not user-friendly.

IkiWiki/Setup/Standard.pm

@@ -26,6 +26,8 @@ sub dumpline ($$$$) { #{{{
 	local $Data::Dumper::Pad="\t";
 	local $Data::Dumper::Sortkeys=1;
 	local $Data::Dumper::Quotekeys=0;
+	# only the perl version preserves utf-8 in output
+	local $Data::Dumper::Useperl=1;
 	
 	my $dumpedvalue;
 	if (($type eq 'boolean' || $type eq 'integer') && $value=~/^[0-9]+$/) {

debian/changelog

@@ -14,6 +14,9 @@ ikiwiki (2.68) UNRELEASED; urgency=low
     formatted time.
   * Fix issue with utf-8 in wikiname breaking session cookies, by
     entity-encoding the wikiname in the session cookie.
+  * Use the pure perl Data::Dumper when generating setup files to ensure that
+    utf-8 characters are written out as such, and not as the encoded perl
+    strings the C Data::Dumper produces.
 
  -- Joey Hess <joeyh@debian.org>  Fri, 17 Oct 2008 20:11:02 -0400
 

doc/bugs/unicode_chars_in_wikiname_break_auth.mdwn

@@ -9,6 +9,8 @@ Replacing "·" with "-" in `wikiname` fixed this login issue.
 > the wikiname. But it doesn't seem to work, somehow the encoded utf-8
 > value still doesn't make it through. (CGI::Session seems to have underermined utf-8
 > issues too.) Seems like I will have to possibly break some sessions and
-> entity-encode the wikiname in the cookie.. done. --[[Joey]]
+> entity-encode the wikiname in the cookie.. [[done]]. --[[Joey]]
 
 (BTW, such a char was replaced by -I don't remember what encoding thingie- in my setup file, when running `ikiwiki-transition setupformat`.)
+
+> Thanks for the heads up, fixed that too. --[[Joey]]