Bug 1159378 - Part 4: Add a unit test for periodic updates of service workers on HTTPS origins; r=nsm The reason why this test is interesting here is that it tests the code path added in the part 2 of this bug.

Bug 1159378 - Part 3: Refactor the logic of test_periodic_update.html into a helper script; r=nsm This helper script will allow us to run the same test across both HTTP and HTTPS origins. This patch also uses postMessage instead of a direct function call on the parent window, because in the HTTPS case, that window will be cross-origin. It also moves the code to trigger the update and get the corresponding updatefound event to wait_for_update.html, since getting the service worker registration cross-origin is not possible.

Bug 1159378 - Part 2: When storing the service worker downloaded after a failed network compare into the cache, store its security info in the cache as well; r=nsm

Bug 1159378 - Part 1: Enable nsIStreamLoader to optionally notify an nsIRequestObserver; r=mcmanus

Bug 1154494 - patch 2 - Hold a ref to the sandbox. r=baku

Bug 1154494 - Hit network only once. r=baku,bkelly

Bug 1162243 - add anonymous speculative connections r=hurley

bug 1161903 ensure pending DrainComplete is not run after Flush() r=cpearce Using a Decode task already in the queue had the potential for out of order flushing. This is similar in behavior to the implementation prior to f550eb7809b6, but keeps mDecoder->Flush() on the platform decoder's task queue, in the hope of avoiding any ordering problems from calling in the middle of decoding input, or races with MFTDecoder::mDiscontinuity. The contract is not clear on whether DrainComplete() should be run during Flush(), but I've kept it in this version.

bug 1161903 reset mDrainComplete after Flush() as DrainComplete() may be called before Flush() r=mattwoodrow DrainComplete() can also be called during Flush().

bug 1161892 use separate thread pool for platform decoder task queues r=bholley so that platform decoder tasks will run when their readers wait and block their thread pool.

bug 1161402 assert active proxy if callback is called r=ajones

Bug 1162318 - Rooting hazard in WrapNativeParentHelper, r=peterv

Bug 1162310 - Do not use nonexistent macro when XGILL_PLUGIN is defined, r=bhackett

Bug 1161377 (part 3) - Convert some easy PL_DHashTable{Init,Finish} cases. r=froydnj. This patch converts easy cases, i.e. where the PL_DHashTableInit() call occurs in a constructor and the PL_DHashTableFinish() call occurs in a destructor.

Bug 1161377 (part 2) - Remove PL_NewDHashTable() and PL_DHashTableDestroy(). r=froydnj. They're not needed now that there is an initializing constructor and a destructor.