[site-url] canonical tag token should not have the trailing slash trimmed

Why are we removing trailing slashes? We should just remove that code.

Oh I think the intent here may have been to prevent tokenized strings like '/node' from getting run through url(). If that's the case we should convert this to a ltrim() rather than the more greedy trim().

yep that seems like the better option

Reviewing this again, we'd probably actually want to skip if the URL is already absolute, as it's pointless running through url() again at that point. Something like:

  if (!empty($content['value'])) {
    $context['value'] = nodewords_replace_tokens($content['value'], $options);
    if (!menu_path_is_external($context['value'])) {
      $context['value'] = trim($context['value'], '/');
      $context['value'] = nodewords_url($context['value'], $options);
    }
    $tags['canonical'] = check_url($content['value']);
  }