Liangent had the idea to make LanguageConverter check for lang attributes in the page content, and disable conversion for those pieces of text, similar to how -{}- disables conversion.
This would avoid e.g. <span lang="ja">-{...}-</span> which is double work, certainly when extensions have to disable conversion.
Version: unspecified
Severity: normal
In LanguageConverter::autoConvert, I see code snippet:
// disable convert to variants between <code></code> tags $codefix = '<code>.+?<\/code>|'; // disable convertsion of <script type="text/javascript"> ... </script> $scriptfix = '<script.*?>.*?<\/script>|'; // disable conversion of <pre xxxx> ... </pre> $prefix = '<pre.*?>.*?<\/pre>|';
Maybe we want to replace these with a real parser, for easier implementation of the wanted feature in this bug.
Another way: I believe there're some HTML parsers inside wikitext parser, to remove harmful attribs. Maybe we can add some markNoConversion calls when it sees elements with other lang="".
Sure, I'll take this for now. If/when I write the parsoid language converter, it would be straightforward to fix this. Fixing it in PHP is harder...
@tstarling -- do you think Remex would be appropriate for this ( https://phabricator.wikimedia.org/T39617#414904 )?
@cscott: Hi, I'm resetting the task assignee due to inactivity. Please feel free to reclaim this task if you plan to work on this - it would be welcome! Also see https://www.mediawiki.org/wiki/Bug_management/Assignee_cleanup for more information - thanks!