The multipart code in SubresourceLoader has some subtleties I misunderstood when I tried to change them in http://trac.webkit.org/changeset/104756. That patch is causing a multipart test to assert in chromium win, and I suspect there are other issues I haven't found yet. I'm going to revert all the multipart changes (except the changes to CachedImage and when we call clear()) and give them more thought before trying again.
Created attachment 122485 [details] patch
Comment on attachment 122485 [details] patch rs=me
Comment on attachment 122485 [details] patch Attachment 122485 [details] did not pass qt-ews (qt): Output: http://queues.webkit.org/results/11152701
Created attachment 122505 [details] Patch for landing
Comment on attachment 122505 [details] Patch for landing Clearing flags on attachment: 122505 Committed r105000: <http://trac.webkit.org/changeset/105000>
All reviewed patches have been landed. Closing bug.