If a form submission via GET is cancelled, FrameLoader::receivedMainResourceError fails to clear m_submittedFormURL, as the form's action doesn't match the requested URL.
Created attachment 158005 [details] Patch
WDYT, Jochen?
I'd defer to Adam or Nate on this
Cool. Ideas for a test would be helpful. :)
Makes sense to me, but I'd like to hear from Nate before landing.
(In reply to comment #5) > Makes sense to me, but I'd like to hear from Nate before landing. For context, http://trac.webkit.org/browser/trunk/Source/WebCore/loader/FrameLoader.cpp#L2571 is the line where it breaks down now. The URLs don't match, as the latter has the whole query string.
Comment on attachment 158005 [details] Patch LGTM, but no test?
(In reply to comment #7) > (From update of attachment 158005 [details]) > LGTM, but no test? Yes, it should have a test before landing it. I talked with Jochen about it, and neither of us could quickly come up with a test that would trigger the conditions. He noted there was a method on testRunner to make the next request fail, but didn't think it would work in the context of a form submit. I'd like to have a test here, I'd appreciate your input as to how it might be structured.
Presumably we'll get a test as part of Bug 93777?
(In reply to comment #9) > Presumably we'll get a test as part of Bug 93777? Sure. You'll get two, in fact! But that's somewhat of a cop out, isn't it? "Dude, I totally have a test in the next patch. I swear!" :)
Comment on attachment 158005 [details] Patch Yeah, but given that the next patch is already written and the tests are there, I don't think there's that much risk that you won't come through. :)
Comment on attachment 158005 [details] Patch In that case, CQ?
Comment on attachment 158005 [details] Patch Clearing flags on attachment: 158005 Committed r125436: <http://trac.webkit.org/changeset/125436>
All reviewed patches have been landed. Closing bug.