Record Stability.iOS.Experimental.Counts metric

Recorded when iOS app has started after a crash caused by system signal
or hang. Not recorded when app has started after UTE or XTE. See
go/bling-stability-101 for imformation about signal crashes, hangs, UTEs
and XTEs. This metric intends to compliment Stability.Counts.
Stability.Counts does include UTEs / XTEs and thus is a noisy metric
which significantly overcounts user-visible crashes.
Stability.IOS.Experimental.Counts is not recorded on OOM crashes, so it
does not represent all use-visible crashes, but OOMs are so rare in
Chrome for iOS that Stability.IOS.Experimental.Counts is still a good
proxy for user-visible crashes. IOS.MetricKit.ForegroundExitData
includes OOM crashes and can be used in conjunction with this metric.

Bug: 1166922
Change-Id: I390a2da2b8b9ed81d215ce58375911019c2228dc
Reviewed-on: https://chromium-review.googlesource.com/c/chromium/src/+/2631287
Commit-Queue: Eugene But <eugenebut@chromium.org>
Reviewed-by: Robert Kaplow <rkaplow@chromium.org>
Reviewed-by: Olivier Robin <olivierrobin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#844349}

ios/chrome/browser/metrics/mobile_session_shutdown_metrics_provider.mm

@@ -108,6 +108,18 @@ enum class MobileSessionAppState {
   kMaxValue = BackgroundXte
 };
 
+// Values of the Stability.iOS.Experimental.Counts histogram.
+// These values are persisted to logs. Entries should not be renumbered and
+// numeric values should never be reused.
+enum class IOSStabilityUserVisibleCrashType {
+  // Termination caused by system signal (f.e. EXC_BAD_ACCESS).
+  TerminationCausedBySystemSignal = 0,
+  // Termination caused by Hang / UI Thread freeze (ui thread was locked for 9+
+  // seconds and the app was quit by OS or the user).
+  TerminationCausedByHang = 1,
+  kMaxValue = TerminationCausedByHang
+};
+
 // Returns value to log for Stability.iOS.UTE.MobileSessionOOMShutdownHint
 // histogram.
 MobileSessionOomShutdownHint GetMobileSessionOomShutdownHint(
@@ -157,6 +169,33 @@ GetMobileSessionAppWillTerminateWasReceived(bool has_possible_explanation) {
              : MobileSessionAppWillTerminateWasReceived::WasReceivedForUte;
 }
 
+// Records Stability.iOS.Experimental.Counts if necessary.
+void LogStabilityIOSExperimentalCounts(
+    MobileSessionShutdownType shutdown_type) {
+  IOSStabilityUserVisibleCrashType type =
+      IOSStabilityUserVisibleCrashType::kMaxValue;
+  switch (shutdown_type) {
+    case SHUTDOWN_IN_FOREGROUND_WITH_CRASH_LOG_WITH_MEMORY_WARNING:
+    case SHUTDOWN_IN_FOREGROUND_WITH_CRASH_LOG_NO_MEMORY_WARNING:
+      type = IOSStabilityUserVisibleCrashType::TerminationCausedBySystemSignal;
+      break;
+    case SHUTDOWN_IN_FOREGROUND_WITH_MAIN_THREAD_FROZEN:
+      type = IOSStabilityUserVisibleCrashType::TerminationCausedByHang;
+      break;
+    case SHUTDOWN_IN_BACKGROUND:
+    case SHUTDOWN_IN_FOREGROUND_NO_CRASH_LOG_NO_MEMORY_WARNING:
+    case SHUTDOWN_IN_FOREGROUND_NO_CRASH_LOG_WITH_MEMORY_WARNING:
+    case FIRST_LAUNCH_AFTER_UPGRADE:
+    case MOBILE_SESSION_SHUTDOWN_TYPE_COUNT:
+      // Nothing to record.
+      return;
+  };
+
+  UMA_STABILITY_HISTOGRAM_ENUMERATION(
+      "Stability.iOS.Experimental.Counts", type,
+      IOSStabilityUserVisibleCrashType::kMaxValue);
+}
+
 // Logs |type| in the shutdown type histogram.
 void LogShutdownType(MobileSessionShutdownType type) {
   UMA_STABILITY_HISTOGRAM_ENUMERATION("Stability.MobileSessionShutdownType",
@@ -297,6 +336,8 @@ void MobileSessionShutdownMetricsProvider::ProvidePreviousSessionData(
   UMA_STABILITY_HISTOGRAM_COUNTS_100("Stability.iOS.TabCountBeforeCrash",
                                      allTabCount);
 
+  LogStabilityIOSExperimentalCounts(shutdown_type);
+
   // Log metrics to improve categorization of crashes.
   LogApplicationBackgroundedTime(session_info.sessionEndTime);
 

tools/metrics/histograms/enums.xml

@@ -39791,6 +39791,19 @@ Called by update_gpu_driver_bug_workaround_entries.py.-->
   <int value="3" label="Actions"/>
 </enum>
 
+<enum name="IOSStabilityUserVisibleTerminationType">
+  <summary>
+    One of several separate stability metrics. The values are non-contiguous as
+    they are a subset of values matching fields of the Stability proto (from the
+    SystemProfile).
+  </summary>
+  <int value="0"
+      label="Termination caused by system signal (f.e. EXC_BAD_ACCESS)"/>
+  <int value="1"
+      label="Termination caused by Hang / UI Thread freeze (ui thread was
+             locked for 9+ seconds and the app was quit by OS or the user)."/>
+</enum>
+
 <enum name="IOSTabSwitcherPageChangeInteraction">
   <int value="0" label="Unknown Interaction"/>
   <int value="1" label="Scroll View Drag"/>

tools/metrics/histograms/histograms_xml/stability/histograms.xml

@@ -345,6 +345,25 @@ reviews. Googlers can read more about this at go/gwsq-gerrit.
   </summary>
 </histogram>
 
+<histogram name="Stability.IOS.Experimental.Counts"
+    enum="IOSStabilityUserVisibleTerminationType" expires_after="2021-06-15">
+  <owner>eugenebut@chromium.org</owner>
+  <owner>olivierrobin@chromium.org</owner>
+  <summary>
+    Recorded when iOS app has started after a crash caused by system signal or
+    hang. Not recorded when app has started after UTE / XTE or started after
+    Chrome update. See go/bling-stability-101 for information about signal
+    crashes, hangs, UTEs and XTEs. This metric intends to complement
+    Stability.Counts. Stability.Counts does include UTEs / XTEs and thus is a
+    noisy metric which significantly overcounts user-visible crashes.
+    Stability.IOS.Experimental.Counts is not recorded on OOM crashes, so it does
+    not represent all user-visible crashes, but OOMs are so rare in Chrome for
+    iOS that Stability.IOS.Experimental.Counts is still a good proxy for
+    user-visible crashes. IOS.MetricKit.ForegroundExitData includes OOM crashes
+    (iOS 14 and higher) and can be used in conjunction with this metric.
+  </summary>
+</histogram>
+
 <histogram name="Stability.iOS.TabCountBefore{Event}" units="tabs"
     expires_after="2021-05-05">
   <owner>djean@chromium.org</owner>