Wikimedia Bugzilla is closed!

Created attachment 13461 [details]
simple tag extension demonstrating the issue.

When using $parser->recursiveTagParse($input), the & character is sometimes escaped and sometimes not:

$input = 'a & b < a | b http://example.com?a&b';
$output = $parser->recursiveTagParse($input);

Here, $output gets
'a & b &lt; a | b <a rel="nofollow" class="external free" href="http://example.com?a&amp;b">http://example.com?a&amp;b">http://example.com?a&amp;b</a>'

The first & is not escaped, and the & in the URL is escaped. < and > are escaped. The source documentation of Parser::recursiveTagParse does not mention the expected behavior. However it should be either escaping the & characters or not.

[As I have to parse the output of recursiveTagParse into a DOMDocument in my extension, this behavior makes my life real hard as I cannot just escape all & before parsing]

I attached a mini-tag extension that can be used to further analyze the problem. Note that in a tag extension, all unescaped & in the returned string are converted to &amp; by MediaWiki at a later stage so that correct HTML is returned even if someone would return the output of $parser->recursiveTagParse($some_wikitext) directly. This "hides" the bug but doesn't solve the underlying problem.


Thanks for developing MediaWiki by the way. It's great!
Philipp