Created attachment 77142 [details] Project showing the bug <rdar://problem/8668230> To reproduce this bug, build and run the attached project. With r71883, the program reports that all requests will be sent, and finishes after logging “Frame load finished”. With r71884, it only reports the first seven requests and never finishes.
r71884 was wrong in moving willSendRequest to ResourceLoader::start(). It’s wrong to do all the scheduling, which is host-dependent, before calling willSendRequest, because the latter can change the host.
Unfortunately, it is not easy to revert r71884 because of later changes to the code.
Created attachment 77176 [details] Call willSendRequest() before scheduling the request
Comment on attachment 77176 [details] Call willSendRequest() before scheduling the request Clearing flags on attachment: 77176 Committed r74453: <http://trac.webkit.org/changeset/74453>
All reviewed patches have been landed. Closing bug.
http://trac.webkit.org/changeset/74453 might have broken Leopard Intel Release (Build), Leopard Intel Debug (Build), Qt Linux Release minimal, Qt Linux ARMv5 Release, Qt Linux ARMv7 Release, Qt Windows 32-bit Release, and Qt Windows 32-bit Debug
Committed r74459: <http://trac.webkit.org/changeset/74459>
(In reply to comment #7) > Committed r74459: <http://trac.webkit.org/changeset/74459> I don’t think this is related to this bug.
(In reply to comment #8) > (In reply to comment #7) > > Committed r74459: <http://trac.webkit.org/changeset/74459> > > I don’t think this is related to this bug. No, I'm sorry. It belonged to https://bugs.webkit.org/show_bug.cgi?id=51445, but a ChangeLog merge error at commit time left my entry below the top :-(