author Kyle Huey <khuey@kylehuey.com>
Tue, 11 Aug 2015 06:10:46 -0700
changeset 277167 a13c1f26e351dd6251da641fe7a9eb53790fc2d0
parent 164130 499de7433a6ad869423d718f067403426fde5799
child 301928 05b4771bed8f59c6521be22718f0d7cfd0cd06c4
permissions -rw-r--r--
Bug 1179909: Refactor stable state handling. r=smaug This is motivated by three separate but related problems: 1. Our concept of recursion depth is broken for things that run from AfterProcessNextEvent observers (e.g. Promises). We decrement the recursionDepth counter before firing observers, so a Promise callback running at the lowest event loop depth has a recursion depth of 0 (whereas a regular nsIRunnable would be 1). This is a problem because it's impossible to distinguish a Promise running after a sync XHR's onreadystatechange handler from a top-level event (since the former runs with depth 2 - 1 = 1, and the latter runs with just 1). 2. The nsIThreadObserver mechanism that is used by a lot of code to run "after" the current event is a poor fit for anything that runs script. First, the order the observers fire in is the order they were added, not anything fixed by spec. Additionally, running script can cause the event loop to spin, which is a big source of pain here (bholley has some nasty bug caused by this). 3. We run Promises from different points in the code for workers and main thread. The latter runs from XPConnect's nsIThreadObserver callbacks, while the former runs from a hardcoded call to run Promises in the worker event loop. What workers do is particularly problematic because it means we can't get the right recursion depth no matter what we do to nsThread. The solve this, this patch does the following: 1. Consolidate some handling of microtasks and all handling of stable state from appshell and WorkerPrivate into CycleCollectedJSRuntime. 2. Make the recursionDepth counter only available to CycleCollectedJSRuntime (and its consumers) and remove it from the nsIThreadInternal and nsIThreadObserver APIs. 3. Adjust the recursionDepth counter so that microtasks run with the recursionDepth of the task they are associated with. 4. Introduce the concept of metastable state to replace appshell's RunBeforeNextEvent. Metastable state is reached after every microtask or task is completed. This provides the semantics that bent and I want for IndexedDB, where transactions autocommit at the end of a microtask and do not "spill" from one microtask into a subsequent microtask. This differs from appshell's RunBeforeNextEvent in two ways: a) It fires between microtasks, which was the motivation for starting this. b) It no longer ensures that we're at the same event loop depth in the native event queue. bent decided we don't care about this. 5. Reorder stable state to happen after microtasks such as Promises, per HTML. Right now we call the regular thread observers, including appshell, before the main thread observer (XPConnect), so stable state tasks happen before microtasks.

/* -*- Mode: C++; tab-width: 2; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
/* vim:set ts=2 sw=2 sts=2 et cindent: */
/* This Source Code Form is subject to the terms of the Mozilla Public
 * License, v. 2.0. If a copy of the MPL was not distributed with this
 * file, You can obtain one at http://mozilla.org/MPL/2.0/. */

#include "nsIThread.idl"

interface nsIRunnable;
interface nsIThreadObserver;

 * The XPCOM thread object implements this interface, which allows a consumer
 * to observe dispatch activity on the thread.
[scriptable, uuid(9cc51754-2eb3-4b46-ae99-38a61881c622)]
interface nsIThreadInternal : nsIThread
   * Get/set the current thread observer (may be null).  This attribute may be
   * read from any thread, but must only be set on the thread corresponding to
   * this thread object.  The observer will be released on the thread
   * corresponding to this thread object after all other events have been
   * processed during a call to Shutdown.
  attribute nsIThreadObserver observer;

   * Add an observer that will *only* receive onProcessNextEvent,
   * beforeProcessNextEvent. and afterProcessNextEvent callbacks. Always called
   * on the target thread, and the implementation does not have to be
   * threadsafe. Order of callbacks is not guaranteed (i.e.
   * afterProcessNextEvent may be called first depending on whether or not the
   * observer is added in a nested loop). Holds a strong ref.
  void addObserver(in nsIThreadObserver observer);

   * Remove an observer added via the addObserver call. Once removed the
   * observer will never be called again by the thread.
  void removeObserver(in nsIThreadObserver observer);

   * This method causes any events currently enqueued on the thread to be
   * suppressed until PopEventQueue is called, and any event dispatched to this
   * thread's nsIEventTarget will queue as well. Calls to PushEventQueue may be
   * nested and must each be paired with a call to PopEventQueue in order to
   * restore the original state of the thread. The returned nsIEventTarget may
   * be used to push events onto the nested queue. Dispatching will be disabled
   * once the event queue is popped. The thread will only ever process pending
   * events for the innermost event queue. Must only be called on the target
   * thread.
  [noscript] nsIEventTarget pushEventQueue();

   * Revert a call to PushEventQueue. When an event queue is popped, any events
   * remaining in the queue are appended to the elder queue. This also causes
   * the nsIEventTarget returned from PushEventQueue to stop dispatching events.
   * Must only be called on the target thread, and with the innermost event
   * queue.
  [noscript] void popEventQueue(in nsIEventTarget aInnermostTarget);

 * This interface provides the observer with hooks to implement a layered
 * event queue.  For example, it is possible to overlay processing events
 * for a GUI toolkit on top of the events for a thread:
 *   var NativeQueue;
 *   Observer = {
 *     onDispatchedEvent(thread) {
 *       NativeQueue.signal();
 *     }
 *     onProcessNextEvent(thread, mayWait) {
 *       if (NativeQueue.hasNextEvent())
 *         NativeQueue.processNextEvent();
 *       while (mayWait && !thread.hasPendingEvent()) {
 *         NativeQueue.wait();
 *         NativeQueue.processNextEvent();
 *       }
 *     }
 *   };
 * NOTE: The implementation of this interface must be threadsafe.
 * NOTE: It is valid to change the thread's observer during a call to an
 *       observer method.
 * NOTE: Will be split into two interfaces soon: one for onProcessNextEvent and
 *       afterProcessNextEvent, then another that inherits the first and adds
 *       onDispatchedEvent.
interface nsIThreadObserver : nsISupports
   * This method is called after an event has been dispatched to the thread.
   * This method may be called from any thread. 
   * @param thread
   *   The thread where the event is being dispatched.
  void onDispatchedEvent(in nsIThreadInternal thread);

   * This method is called when nsIThread::ProcessNextEvent is called.  It does
   * not guarantee that an event is actually going to be processed.  This method
   * is only called on the target thread.
   * @param thread
   *   The thread being asked to process another event.
   * @param mayWait
   *   Indicates whether or not the method is allowed to block the calling
   *   thread.  For example, this parameter is false during thread shutdown.
  void onProcessNextEvent(in nsIThreadInternal thread, in boolean mayWait);

   * This method is called (from nsIThread::ProcessNextEvent) after an event
   * is processed.  It does not guarantee that an event was actually processed
   * (depends on the value of |eventWasProcessed|.  This method is only called
   * on the target thread.  DO NOT EVER RUN SCRIPT FROM THIS CALLBACK!!!
   * @param thread
   *   The thread that processed another event.
   * @param eventWasProcessed
   *   Indicates whether an event was actually processed. May be false if the
   *   |mayWait| flag was false when calling nsIThread::ProcessNextEvent().
  void afterProcessNextEvent(in nsIThreadInternal thread,
                             in bool eventWasProcessed);