Snap for 9410359 from 36f488b8 to tm-qpr2-release

import android.view.Window;

import android.os.Process;

import androidx.annotation.Nullable;

import androidx.annotation.UiThread;

import androidx.annotation.VisibleForTesting;

import androidx.annotation.WorkerThread;

                .findFirst();

    }

    /**

     * Once this window listener is attached to a window's root view, it traverses the entire

     * view tree on the main thread every time onDraw is called. It then saves the state of the view

     * tree traversed in a local list of nodes, so that this list of nodes can be processed on a

     * background thread, and prepared for being dumped into a bugreport.

     *

     * Since some of the work needs to be done on the main thread after every draw, this piece of

     * code needs to be hyper optimized. That is why we are recycling ViewRef and ViewPropertyRef

     * objects and storing the list of nodes as a flat LinkedList, rather than as a tree. This data

     * structure allows recycling to happen in O(1) time via pointer assignment. Without this

     * optimization, a lot of time is wasted creating ViewRef objects, or finding ViewRef objects to

     * recycle.

     *

     * Another optimization is to only traverse view nodes on the main thread that have potentially

     * changed since the last frame was drawn. This can be determined via a combination of private

     * flags inside the View class.

     *

     * Another optimization is to not store or manipulate any string objects on the main thread.

     * While this might seem trivial, using Strings in any form causes the ViewCapture to hog the

     * main thread for up to an additional 6-7ms. It must be avoided at all costs.

     *

     * Another optimization is to only store the class names of the Views in the view hierarchy one

     * time. They are then referenced via a classNameIndex value stored in each ViewPropertyRef.

     *

     * TODO: b/262585897: If further memory optimization is required, an effective one would be to

     * only store the changes between frames, rather than the entire node tree for each frame.

     * The go/web-hv UX already does this, and has reaped significant memory improves because of it.

     *

     * TODO: b/262585897: Another memory optimization could be to store all integer, float, and

     * boolean information via single integer values via the Chinese remainder theorem, or a similar

     * algorithm, which enables multiple numerical values to be stored inside 1 number. Doing this

     * would allow each ViewProperty / ViewRef to slim down its memory footprint significantly.

     *

     * One important thing to remember is that bugs related to recycling will usually only appear

     * after at least 2000 frames have been rendered. If that code is changed, the tester can

     * use hard-coded logs to verify that recycling is happening, and test view capturing at least

     * ~8000 frames or so to verify the recycling functionality is working properly.

     */

    private class WindowListener implements ViewTreeObserver.OnDrawListener {

        public final View mRoot;

            this.name = name;

        }

        /**

         * Every time onDraw is called, it does the minimal set of work required on the main thread,

         * i.e. capturing potentially dirty / invalidated views, and then immediately offloads the

         * rest of the processing work (extracting the captured view properties) to a background

         * thread via mExecutor.

         */

        @Override

        public void onDraw() {

            Trace.beginSection("view_capture");

            mNodesBg[mFrameIndexBg] = resultStart;

        }

        private ViewPropertyRef findInLastFrame(int hashCode) {

        private @Nullable ViewPropertyRef findInLastFrame(int hashCode) {

            int lastFrameIndex = (mFrameIndexBg == 0) ? mMemorySize - 1 : mFrameIndexBg - 1;

            ViewPropertyRef viewPropertyRef = mNodesBg[lastFrameIndex];

            while (viewPropertyRef != null && viewPropertyRef.hashCode != hashCode) {