http://bugs.winehq.org/show_bug.cgi?id=30653
--- Comment #3 from Dan Kegel dank@kegel.com 2012-05-15 11:13:07 CDT --- Created attachment 40185 --> http://bugs.winehq.org/attachment.cgi?id=40185 More careful patch
Dylan suggested taking advantage of the fact that there should only be one bunch of marked paragraphs. While implementing that, I noticed a possible old bug: the original loop used item->member.para.next_para even on non-paragraphs, which is probably a no-no. This patch also gets rid of the spurious invaidates, and if you ask it to with +richedit_check, verifies its assumption that there is only one bunch of marked paragraphs.