phil ringnalda

First of a series of quick fixes for early adopters of MT 2.6:

If you use (the newly redesigned) encode_xml filter on a tag that is automatically run through the (new) sanitize filter (as you probably do if you have a comment feed), or if you combine encode_xml with remove_html (as you quite likely do), then you’ve got a problem with your RSS feed: encode_xml now wraps anything that looks to it like HTML in a CDATA section (starts with the less-than character, then ![CDATA[, and ends with ]] and the greater-than character, and can cause surprising results if misinterpreted so I’m talking around an example). That’s mostly okay, since it makes for faster parsing: if you have the old-style entity-encoded HTML then an XML parser has to go through it character by character, looking for XML tags and entities that it needs to parse, while a CDATA section tells the parser that it doesn’t need to look, and you are a little less likely to break a parser with a CDATA section than with parsed text including entities.

However, because of the order that the filters appear in {your MT directory}/lib/MT/Template/Context.pm, encode_xml runs first, adding the CDATA section if it sees HTML, and then either sanitize or remove_html runs after that, and they will both remove the opening CDATA tag, leaving you with nothing but the closing tag and possibly naked invalid XML. For a quick fix, edit Context.pm, find the section in sub post_process_handler that reads:

if ($local_args{'encode_xml'}) {
    $str = encode_xml($str);
}

cut it out of where it is, and paste it below

if (my $spec = $local_args{'sanitize'}) {
    require MT::Sanitize;
    if ($spec eq '1') {
        $spec = $ctx->stash('blog')->sanitize_spec ||
            MT::ConfigMgr->instance->GlobalSanitizeSpec;
        }
    $str = MT::Sanitize->sanitize($str, $spec);
}

so that both remove_html and sanitize get to run before encode_xml starts sticking in CDATA that will confuse them.