Add histograms for deletion of session cookies on startup.

content/browser/net/sqlite_persistent_cookie_store.cc

Index: content/browser/net/sqlite_persistent_cookie_store.cc
diff --git a/content/browser/net/sqlite_persistent_cookie_store.cc b/content/browser/net/sqlite_persistent_cookie_store.cc
index 6b2ff26ec3703f05d83b3d892ac16b6468a35700..86268cb9be7a426c88ddc6635eaf296e69b196ad 100644
--- a/content/browser/net/sqlite_persistent_cookie_store.cc
+++ b/content/browser/net/sqlite_persistent_cookie_store.cc
@@ -701,8 +701,16 @@ bool SQLitePersistentCookieStore::Backend::InitializeDatabase() {
 
   initialized_ = true;
 
-  if (!restore_old_session_cookies_)
+  if (!restore_old_session_cookies_) {
+    base::Time start_time = base::Time::Now();
     DeleteSessionCookiesOnStartup();
+
+    UMA_HISTOGRAM_TIMES("Cookie.Startup.TimeSpentDeletingCookies",
+                        base::Time::Now() - start_time);
+    UMA_HISTOGRAM_COUNTS("Cookie.Startup.NumberOfCookiesDeleted",
+                         db_->GetLastChangeCount());

[erikwright (departed), 2015/05/12 19:39:50]

Either:

(1) Have DeleteSessionCookiesOnStartup return this count
(2) Put the metric reporting inside DeleteSessionCookiesOnStartup
(3) Inline the contents of DeleteSessionCookiesOnStartup here

Reason: It's not obvious from what we see here that
DeleteSessionCookiesOnStartup only executes a single statement. But this code
assumes that. If DeleteSessionCookiesOnStartup were later changed to (for
example) delete on a domain-by-domain basis, the assumption would be broken.

[erikchen, 2015/05/12 20:33:49]

Oh yes, good point. This logic originally started in
DeleteSessionCookiesOnStartup(), but got moved around while I was adding
different metrics. Thanks!

+  }
+
   return true;
 }