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 f879a89c876967194b6b5b4b2422d07978306e67..5d478bce02f99d9a09f2688ae425ec38e272a28b 100644
--- a/tools/metrics/histograms/histograms.xml
+++ b/tools/metrics/histograms/histograms.xml
@@ -24621,6 +24621,57 @@ Therefore, the affected-histogram name has to have at least one dot in it.
   <summary>Temperature reading at sensor 9 (PECI) taken every 30s.</summary>
 </histogram>
 
+<histogram name="Platform.Thermal.Temperature.Cpu.0" units="Celsius">
+  <owner>mka@chromium.org</owner>
+  <summary>Temperature reading at sensor 0 of the CPU.</summary>

[vadimb1, 2015/01/21 18:48:39]

Please add information about how often this is logged, otherwise it's unclear
how to interpret that (i.e. logged when the temperature is changed, or every
second, or once a day, or when the system is restarted - it's all different).
Same for other Temperature histograms.

[mka, 2015/01/21 19:19:33]

my idea is to log it every second, however it could happen that the actual
sensor reading is performed at a slightly lower rate (every few seconds). the
component reporting the metrics doesn't actually read the sensor but subscribes
to the sensor to be notified on changes. the polling rate of the sensor might
vary between different sensors and devices.

not sure whether to express this here in detail or say "every few seconds"

[vadimb1, 2015/01/21 22:32:56]

Well the idea of the summary (which is shown on the dashboards) is that anyone
looking at it should get as much information as reasonably possible to fit in a
few lines of text so they don't need to dig into the sources or e-mail you every
time. Not every person looking at this is familiar with your project; we had
lots of scenarios when people find a histogram (in many cases created a few
years ago) by text search and want to quickly check if it is close to what they
were looking for.

In this case I would say "collected every few seconds (may vary between
devices)".

See for example "Platform.Temperature.Sensor01" that says "Temperature reading
at sensor 1 (I2C_CPU-Object) taken every 30s."

BTW what is the difference, can you use Platform.Temperature.* or is your
histogram completely unrelated to that group?

Also on this subject, most of the other Platform.* histograms start description
with "Chrome OS" - now sure if that is your case but you may follow the example
for consistency if it is (but up to you).

[mka, 2015/01/21 23:25:19]

ack, I'll update the description
yes, my histogram is completely unrelated. I saw the Platform.Temperature.*
histograms and considered to add the new histograms to this group, but as there
are also the thermal zone histograms it seemed more appropriate to group them
together under thermal
Jetstream runs Chrome OS (Core), I'll add it to the description

+</histogram>
+
+<histogram name="Platform.Thermal.Temperature.Wifi0" units="Celsius">
+  <owner>mka@chromium.org</owner>
+  <summary>Temperature reading at wireless interface 0.</summary>
+</histogram>
+
+<histogram name="Platform.Thermal.Temperature.Wifi1" units="Celsius">
+  <owner>mka@chromium.org</owner>
+  <summary>Temperature reading at wireless interface 1.</summary>
+</histogram>
+
+<histogram name="Platform.Thermal.Temperature.Wifi2" units="Celsius">
+  <owner>mka@chromium.org</owner>
+  <summary>Temperature reading at wireless interface 2.</summary>
+</histogram>
+
+<histogram name="Platform.Thermal.Zone.Cpu.States">
+  <owner>mka@chromium.org</owner>
+  <summary>
+    Distribution of time spent in the states of the thermal zone of the CPU.

[vadimb1, 2015/01/21 18:48:39]

1. Same question about how often this is logged.
2. What are the states? Since it's not enum it will be an integer, how to
interpret that integer?
3. So what is logged here? Time or state? E.g., say I have the histogram:
0-1: 50
1-2: 60
How one would understand that? 50ms spent in state 0? Or 50s? Or we've seen the
system 50 times in state 0?

Same for the other states.

[mka, 2015/01/21 19:19:33]

every second
I left it vague on purpose as the interpretation of the values might vary across
different devices. What is true for all devices is that a higher state
corresponds to a higher temperature
the state is logged, so it would be 50s spent in state 0
 

[vadimb1, 2015/01/21 22:32:56]

Well, make it clearer from the summary. It's very similar to the previous
histograms group but the summary reads differently. I would say something like
"State of the termal zone, collected every second. State depends on device,
higher state corresponds to a higher temperature" or "Time spent in the states
of termal zone, one count collected every second...".

Also you probably could use units="State" here. Then the histograms table will
be 
  State Quantity PDF CFD
while without units it will be 
  Range Quantity PDF CFD

[mka, 2015/01/21 23:25:19]

ok
I'll update the histograms to use units

+  </summary>
+</histogram>
+
+<histogram name="Platform.Thermal.Zone.Wifi0.States">
+  <owner>mka@chromium.org</owner>
+  <summary>
+    Distribution of time spent in the states of the thermal zone of wireless.
+    interface 0.
+  </summary>
+</histogram>
+
+<histogram name="Platform.Thermal.Zone.Wifi1.States">
+  <owner>mka@chromium.org</owner>
+  <summary>
+    Distribution of time spent in the states of the thermal zone of wireless.
+    interface 1.
+  </summary>
+</histogram>
+
+<histogram name="Platform.Thermal.Zone.Wifi2.States">
+  <owner>mka@chromium.org</owner>
+  <summary>
+    Distribution of time spent in the states of the thermal zone of wireless
+    interface 2.
+  </summary>
+</histogram>
+
 <histogram name="Platform.TPM.DictionaryAttackCounter">
   <owner>dkrahn@chromium.org</owner>
   <summary>