tools/metrics/histograms/histograms.xml

Index: tools/metrics/histograms/histograms.xml
diff --git a/tools/metrics/histograms/histograms.xml b/tools/metrics/histograms/histograms.xml
index 6c89c0a9380ff3b2fc00705b4a0a6b48ca42e2df..149f0de1a7021cb2c3b533465c9379d06271284e 100644
--- a/tools/metrics/histograms/histograms.xml
+++ b/tools/metrics/histograms/histograms.xml
@@ -52679,6 +52679,18 @@ http://cs/file:chrome/histograms.xml - but prefer this file for new entries.
   </summary>
 </histogram>
 
+<histogram name="Process.Sandbox.JobAvoidedCorrectly" enum="Boolean">
+  <owner>pastarmovj@chromium.org</owner>
+  <summary>
+    Measure how often the automatic recognition for terminal services
+    environment would have incorrectly decided that the job object should be
+    applied, when it shouldn't have been as dictated by the flag
+    --allow-no-sanbox-job. This histogram will guide the final removal of the
+    flag, which will be possible when the number of false values becomes
+    insignificant.

[jwd, 2017/03/02 20:55:24]

Ah, ok, so you care about the false case most. Based the name and summary, it
still seems strange that the true case also includes "the job object was
applied, but the flag agreed with that".

What about something like FlagOverrodeLocalSessionCheck, and reverse the values
recorded to the histogram? 

I also wonder why you need the histogram on 297? It seems to muddy the meaning
of what's being captured. If you do need it, I'd suggest switching this to a 3
value enum histogram with remote session, local session and flag agrees, local
session and flag disagrees or something.

[pastarmovj, 2017/03/02 21:11:09]

Agreed and renamed :)
This one was actually a mistake - it reappeared when I reverted Patchest 6 to
PS5 as part of the ongoing discussion. It is indeed only the binary state from
the first count.

+  </summary>
+</histogram>
+
 <histogram name="Process.Sandbox.Launch.Error" enum="WinGetLastError">
   <owner>shrikant@chromium.org</owner>
   <summary>