proper binmode settings so that with -CSD, ikiwiki will support unicode

however, due to robustness, that's not enabled by default yet

3 files changed, 37 insertions, 5 deletions

diff --git a/IkiWiki/Render.pm b/IkiWiki/Render.pm
index 23f6b1c72..de35d24e1 100644
--- a/IkiWiki/Render.pm
+++ b/IkiWiki/Render.pm
@@ -341,8 +341,8 @@ sub render ($) { #{{{
 	
 	my $type=pagetype($file);
 	my $srcfile=srcfile($file);
-	my $content=readfile($srcfile);
 	if ($type ne 'unknown') {
+		my $content=readfile($srcfile);
 		my $page=pagename($file);
 		
 		$links{$page}=[findlinks($content, $page)];
@@ -366,9 +366,10 @@ sub render ($) { #{{{
 		}
 	}
 	else {
+		my $content=readfile($srcfile, 1);
 		$links{$file}=[];
 		check_overwrite("$config{destdir}/$file", $file);
-		writefile($file, $config{destdir}, $content);
+		writefile($file, $config{destdir}, $content, 1);
 		$oldpagemtime{$file}=time;
 		$renderedfiles{$file}=$file;
 	}
diff --git a/doc/todo/utf8.mdwn b/doc/todo/utf8.mdwn
new file mode 100644
index 000000000..536ec75b2
--- /dev/null
+++ b/doc/todo/utf8.mdwn
@@ -0,0 +1,27 @@
+ikiwiki should support utf-8 pages, both input and output
+
+Currently ikiwiki is belived to be utf-8 clean itself; it tells perl to use
+binmode when reading possibly binary files (such as images) and it uses
+utf-8 compatable regexps etc.
+
+utf-8 IO is not enabled by default though. While you can probably embed
+utf-8 in pages anyway, ikiwiki will not treat it right in the cases where
+it deals with things on a per-character basis (mostly when escaping and
+de-escaping special characters in filenames).
+
+To enable utf-8, edit ikiwiki and add -CSD to the perl hashbang line.
+(This should probably be configurable via a --utf8 or better --encoding=
+switch.)
+
+The following problems have been observed when running ikiwiki this way:
+
+* If invalid utf-8 creeps into a file, ikiwiki will crash rendering it as
+  follows:
+
+	Malformed UTF-8 character (unexpected continuation byte 0x97, with no preceding start byte) in substitution iterator at /usr/bin/markdown line 1317.
+	Malformed UTF-8 character (fatal) at /usr/bin/markdown line 1317.
+
+  In this example, a literal 0x97 character had gotten into a markdown
+  file. 
+  
+  Here, let's put one in this file: "—"
diff --git a/ikiwiki b/ikiwiki
index 97afa853f..6bf58017d 100755
--- a/ikiwiki
+++ b/ikiwiki
@@ -193,24 +193,27 @@ sub srcfile ($) { #{{{
 	error("internal error: $file cannot be found");
 } #}}}
 
-sub readfile ($) { #{{{
+sub readfile ($;$) { #{{{
 	my $file=shift;
+	my $binary=shift;
 
 	if (-l $file) {
 		error("cannot read a symlink ($file)");
 	}
 	
 	local $/=undef;
-	open (IN, "$file") || error("failed to read $file: $!");
+	open (IN, $file) || error("failed to read $file: $!");
+	binmode(IN) if $binary;
 	my $ret=<IN>;
 	close IN;
 	return $ret;
 } #}}}
 
-sub writefile ($$$) { #{{{
+sub writefile ($$$;$) { #{{{
 	my $file=shift; # can include subdirs
 	my $destdir=shift; # directory to put file in
 	my $content=shift;
+	my $binary=shift;
 	
 	my $test=$file;
 	while (length $test) {
@@ -232,6 +235,7 @@ sub writefile ($$$) { #{{{
 	}
 	
 	open (OUT, ">$destdir/$file") || error("failed to write $destdir/$file: $!");
+	binmode(OUT) if $binary;
 	print OUT $content;
 	close OUT;
 } #}}}