Add UMA histogram to count hard revokes of user certificate error decisions.

chrome/browser/ssl/chrome_ssl_host_state_delegate.cc

Index: chrome/browser/ssl/chrome_ssl_host_state_delegate.cc
diff --git a/chrome/browser/ssl/chrome_ssl_host_state_delegate.cc b/chrome/browser/ssl/chrome_ssl_host_state_delegate.cc
index 64f622e2877e106b87ea6e4c71815869b4aa93ee..abd34d95bacda35d91eb264c990dc99e86454e3f 100644
--- a/chrome/browser/ssl/chrome_ssl_host_state_delegate.cc
+++ b/chrome/browser/ssl/chrome_ssl_host_state_delegate.cc
@@ -9,6 +9,7 @@
 #include "base/command_line.h"
 #include "base/logging.h"
 #include "base/metrics/field_trial.h"
+#include "base/metrics/histogram.h"
 #include "base/strings/string_number_conversions.h"
 #include "base/time/clock.h"
 #include "base/time/default_clock.h"
@@ -339,6 +340,7 @@ void ChromeSSLHostStateDelegate::RevokeUserDecisions(const std::string& host) {
 // idle sockets.
 void ChromeSSLHostStateDelegate::RevokeUserDecisionsHard(
     const std::string& host) {
+  UMA_HISTOGRAM_COUNTS("interstitial.ssl.revoke_user_decisions_hard", 1);

[jar (doing other things), 2014/08/14 22:10:37]

Several issues/suggestions:

Yeah, this would in a strange way generate a count... but it would be
meaningless, as there would be no way to normalize it.  Suppose you find that
you get 200,000 times that a user walked through this code... what would that
tell you?  How would that show you a different answer from it saying 400,000?? 
Ans: Nothing really.

You would much rather (for example) look at how many times a user saw an
interstitial, and count clicked through, vs back away (cancel).  Hence you would
(in that case) sample *either* a counter for "CLICK_THROUGH" or for
"CANCEL_REQUEST".  The histogram to generally use for that would be
UMA_HISTOGRAM_ENUMERATION("some name", sample_enum,
LARGER_THAN_ANY_SAMPLED_ENUMS).

You might also want to normalize relative to the number of times folks did an
SSL validation (and hence didn't even get the need to click through), as well as
how many times they hit a problem where they couldn't click through (i.e.,
perhaps there would be four valid enum values??).

IF you wanted to only do a count, you wouldn't use this macro
(UMA_HISTOGRAM_COUNTS), as it expects samples in the range of about 1 to 10,000,
and puts it into 1 of 50 distinct buckets.  That macro is more useful in trying
to (for example) record how many frames were on a page (that might be the
sample), or how many KB of data was loaded during a page load, etc.  It wouldn't
be used for a singleton count.

...and if you really just wanted a random counter, showing (without
normalization!) how many times a feature was used, you *might* use an Event
counter rather than a histogram... but that is a different story... and less
commonly useful.

   RevokeUserDecisions(host);
   scoped_refptr<net::URLRequestContextGetter> getter(
       profile_->GetRequestContext());