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 da7ec7f45794773ed93077035ce0a25a71356f86..aa8652f120c24284de016430dbf8986fb0b32faa 100644
--- a/tools/metrics/histograms/histograms.xml
+++ b/tools/metrics/histograms/histograms.xml
@@ -5534,6 +5534,16 @@ http://cs/file:chrome/histograms.xml - but prefer this file for new entries.
   </summary>
 </histogram>
 
+<histogram name="ChildProcessSecurityPolicy.FileSystemCanCommitOrigin"

[Ilya Sherman, 2016/10/05 00:50:40]

It looks like you're declaring a new top-level name,
"ChildProcessSecurityPolicy".  Could you please check whether there's any
existing top-level group names that would be appropriate to nest this under?

[ncarter (slow), 2016/10/05 19:03:06]

Done. This particular case was a nice fit into the existing "FileSystem."
category. Though actually, it seems there are several
"ChildProcessSecurityPolicy." UMA histograms declared, just none appear to be in
histograms.xml?! I would fix it in this patch if I weren't targeting a merge. Do
we have anything server-side that will warn us about non-exported histograms
that are being uploaded in UMA reports?

[Ilya Sherman, 2016/10/05 22:37:09]

Ah, interesting -- some histograms with this category do appear in the internal
histograms.xml file, but don't seem to have any data shown on the dashboard.  I
wonder if they're not generating any data, or what's going on...

(It's hard to have a server-side warning, because we only see the hashes, and we
quite intentionally can't unhash unmapped histograms.  We do have
find_unmapped_histograms.py though.)

+    units="boolean">

[Ilya Sherman, 2016/10/05 00:50:40]

Please specify an enum here with semantically appropriate labels.

[ncarter (slow), 2016/10/05 19:03:06]

Thanks for suggesting this. Looking around I found BooleanHit which looks like a
good match -- is that the preferred pattern for a hit counter histogram, where
there's no value to sample, you just want to record that an event happened? I've
considered UMA_HISTOGRAM_COUNTS(..., 1) before, but it just looked so wrong --
there's no non-histogram macro like UMA_EVENT_COUNTER() right?

[Ilya Sherman, 2016/10/05 22:37:09]

Yep, BooleanHit sounds like a good fit.  It's relatively rarely used, so there's
no macro defined for it; UMA_HISTOGRAM_BOOLEAN is a good choice for now.

+  <owner>nick@chromium.org</owner>
+  <summary>
+    Logs the result of the origin check from of ChildProcessSecurityPolicy's
+    HasPermissionsForFileSystemFile. True is success and is common; false is
+    failure and indicates a bug.
+  </summary>
+</histogram>
+
 <histogram name="Chrome.Android.Activity.CrashCounts" enum="AndroidActivityId">
   <obsolete>
     Deprecated 04/2016 as doesn't have data nor owner.