...as well as content policy handling, which is started within responseReceived().
Created attachment 193605 [details] patch
Comment on attachment 193605 [details] patch Attachment 193605 [details] did not pass chromium-ews (chromium-xvfb): Output: http://webkit-commit-queue.appspot.com/results/17205560 New failing tests: platform/chromium/http/tests/security/mixedContent/insecure-iframe-in-main-frame-blocked.html http/tests/security/contentSecurityPolicy/1.1/form-action-src-blocked.html http/tests/security/contentSecurityPolicy/1.1/form-action-src-get-blocked.html
Created attachment 193634 [details] attempted fix to test failure
Comment on attachment 193634 [details] attempted fix to test failure View in context: https://bugs.webkit.org/attachment.cgi?id=193634&action=review > Source/WebCore/loader/MainResourceLoader.cpp:-327 > - ref(); // balanced by deref in continueAfterContentPolicy and cancel Why don't you need these manual ref counts anymore?
(In reply to comment #4) > (From update of attachment 193634 [details]) > View in context: https://bugs.webkit.org/attachment.cgi?id=193634&action=review > > > Source/WebCore/loader/MainResourceLoader.cpp:-327 > > - ref(); // balanced by deref in continueAfterContentPolicy and cancel > > Why don't you need these manual ref counts anymore? I don't fully understand why they were there in the first place (they date back to the initial "make the loader platform-independent" effort). I can't find any other place we have raw ref/deref calls to protect an object during policy callbacks. Also, as far as I can tell, DocumentLoader is a little tougher to assassinate than MainResourceLoader. My best guess is that these are already vestigial, but it's hard to prove.
Comment on attachment 193634 [details] attempted fix to test failure Ok. That's the part of this change that seems the scariest, but maybe I've got my rose colored glasses on today. ;)
Comment on attachment 193634 [details] attempted fix to test failure Clearing flags on attachment: 193634 Committed r146216: <http://trac.webkit.org/changeset/146216>
All reviewed patches have been landed. Closing bug.