Unique origins shouldn't remember their scheme, host, or port
Created attachment 115030 [details] Patch
Comment on attachment 115030 [details] Patch Eric convinced me to do the drag-drop-dead-frame.html changes separately.
Created attachment 115052 [details] Patch
Comment on attachment 115052 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=115052&action=review LGTM. > LayoutTests/fast/frames/resources/sandboxed-iframe-storage-disallowed.html:5 > +<script> > +if (window.layoutTestController) > + layoutTestController.dumpAsText(); Would be nice to have a comment here about why we inline this.
Committed r100213: <http://trac.webkit.org/changeset/100213>
Reopening due to rollback, see bug 72371.
Those Chromium file API tests are blocking this patch, which is blocking a larger security cleanup tracked in Bug 71745.
I'm going to try landing this again with a quirk for the storage identifier. This keeps the filesystem tests (both layout and unit) working as before. It's slightly ugly, but it lets us move forward with the general program of tightening up the SecurityOrigin implementation. I'll need to figure out how this impacts Bug 72342 once this lands.
Created attachment 115644 [details] Patch for landing
Committed r100663: <http://trac.webkit.org/changeset/100663>