Android: Clean up histogram buckets range

chrome/android/java/src/org/chromium/chrome/browser/tab/TabWebContentsObserver.java

 // Copyright 2015 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 package org.chromium.chrome.browser.tab;
 
 import android.os.SystemClock;
 import android.support.annotation.IntDef;
 import android.view.View;
 
@@ skipping 182 matching lines @@
             boolean isMainFrame, String validatedUrl, boolean isErrorPage) {
         if (isMainFrame) mTab.didStartPageLoad(validatedUrl, isErrorPage);
 
         mTab.handleDidStartProvisionalLoadForFrame(isMainFrame, validatedUrl);
     }
 
     @Override
     public void didCommitProvisionalLoadForFrame(long frameId, boolean isMainFrame, String url,
             int transitionType) {
         if (isMainFrame && UmaUtils.isRunningApplicationStart()) {
-            // Currently it takes about 2000ms to commit a navigation if the measurement
-            // begins very early in the browser start. How many buckets (b) are needed to
-            // explore the _typical_ values with granularity 100ms and a maximum duration
-            // of 1 minute?
-            //   s^{n+1} / s^{n} = 2100 / 2000
-            //   s = 1.05
-            //   s^b = 60000
-            //   b = ln(60000) / ln(1.05) ~= 225
-            RecordHistogram.recordCustomTimesHistogram("Startup.FirstCommitNavigationTime2",
+            // Current median is 550ms, and long tail is very long. ZoomedIn gives good view of the

[pasko, 2016/12/15 13:36:19]

wow, I did not realize that so drastically! yay!

Seems like it is mostly coming from AMP (See old timeline for
CustomTabs.IntentToFirstCommitNavigationTime), which seems to have coincided
with the the rollout of AMP in GSA.

[Peter Wen, 2016/12/15 13:39:11]

Yay! I didn't have context to know whether that was good or bad. :)

+            // median and ZoomedOut gives a good overview.
+            RecordHistogram.recordCustomTimesHistogram(
+                    "Startup.FirstCommitNavigationTime2.ZoomedIn",
                     SystemClock.uptimeMillis() - UmaUtils.getForegroundStartTime(),
-                    1, 60000 /* 1 minute */, TimeUnit.MILLISECONDS, 225);
+                    200, 1000, TimeUnit.MILLISECONDS, 100);
+            RecordHistogram.recordCustomTimesHistogram(
+                    "Startup.FirstCommitNavigationTime2.ZoomedOut",
+                    SystemClock.uptimeMillis() - UmaUtils.getForegroundStartTime(),
+                    50, TimeUnit.MINUTES.toMillis(10), TimeUnit.MILLISECONDS, 50);

[Alexei Svitkine (slow), 2016/12/15 00:34:37]

Is the 50 for the min value intentional? If so, add a comment.

[pasko, 2016/12/15 13:36:19]

I think the comment just above sufficiently explains it, up to you guys.

[Peter Wen, 2016/12/15 13:39:11]

Done adding a quick comment and reference to other histogram.

             UmaUtils.setRunningApplicationStart(false);
         }
 
         if (isMainFrame) {
             mTab.setIsTabStateDirty(true);
             mTab.updateTitle();
         }
 
         RewindableIterator<TabObserver> observers = mTab.getTabObservers();
         while (observers.hasNext()) {
@@ skipping 88 matching lines @@
         }
     }
 
     @Override
     public void destroy() {
         MediaCaptureNotificationService.updateMediaNotificationForTab(
                 mTab.getApplicationContext(), mTab.getId(), 0, mTab.getUrl());
         super.destroy();
     }
 }