sensors_temperatures() return {} on my CentOS system

[nicolargo]

On my CentOS 7.1 system with PsUtil 5.1.2.

I have the following sensors:

$ ll /sys/class/hwmon/hwmon0/device/temp*
-r--r--r-- 1 root root 4096 23 janv. 15:20 /sys/class/hwmon/hwmon0/device/temp1_crit
-r--r--r-- 1 root root 4096 23 janv. 15:20 /sys/class/hwmon/hwmon0/device/temp1_crit_alarm
-r--r--r-- 1 root root 4096 23 janv. 15:20 /sys/class/hwmon/hwmon0/device/temp1_input
-r--r--r-- 1 root root 4096 24 janv. 13:27 /sys/class/hwmon/hwmon0/device/temp1_label
-r--r--r-- 1 root root 4096 23 janv. 15:20 /sys/class/hwmon/hwmon0/device/temp1_max
-r--r--r-- 1 root root 4096 23 janv. 15:20 /sys/class/hwmon/hwmon0/device/temp2_crit
-r--r--r-- 1 root root 4096 23 janv. 15:20 /sys/class/hwmon/hwmon0/device/temp2_crit_alarm
-r--r--r-- 1 root root 4096 23 janv. 15:20 /sys/class/hwmon/hwmon0/device/temp2_input
-r--r--r-- 1 root root 4096 24 janv. 13:27 /sys/class/hwmon/hwmon0/device/temp2_label
-r--r--r-- 1 root root 4096 23 janv. 15:20 /sys/class/hwmon/hwmon0/device/temp2_max
-r--r--r-- 1 root root 4096 23 janv. 15:20 /sys/class/hwmon/hwmon0/device/temp3_crit
-r--r--r-- 1 root root 4096 23 janv. 15:20 /sys/class/hwmon/hwmon0/device/temp3_crit_alarm
-r--r--r-- 1 root root 4096 23 janv. 15:20 /sys/class/hwmon/hwmon0/device/temp3_input
-r--r--r-- 1 root root 4096 24 janv. 13:27 /sys/class/hwmon/hwmon0/device/temp3_label
-r--r--r-- 1 root root 4096 23 janv. 15:20 /sys/class/hwmon/hwmon0/device/temp3_max
-r--r--r-- 1 root root 4096 23 janv. 15:20 /sys/class/hwmon/hwmon0/device/temp4_crit
-r--r--r-- 1 root root 4096 23 janv. 15:20 /sys/class/hwmon/hwmon0/device/temp4_crit_alarm
-r--r--r-- 1 root root 4096 23 janv. 15:20 /sys/class/hwmon/hwmon0/device/temp4_input
-r--r--r-- 1 root root 4096 24 janv. 13:27 /sys/class/hwmon/hwmon0/device/temp4_label
-r--r--r-- 1 root root 4096 23 janv. 15:20 /sys/class/hwmon/hwmon0/device/temp4_max
-r--r--r-- 1 root root 4096 23 janv. 15:20 /sys/class/hwmon/hwmon0/device/temp5_crit
-r--r--r-- 1 root root 4096 23 janv. 15:20 /sys/class/hwmon/hwmon0/device/temp5_crit_alarm
-r--r--r-- 1 root root 4096 23 janv. 15:20 /sys/class/hwmon/hwmon0/device/temp5_input
-r--r--r-- 1 root root 4096 24 janv. 13:27 /sys/class/hwmon/hwmon0/device/temp5_label
-r--r--r-- 1 root root 4096 23 janv. 15:20 /sys/class/hwmon/hwmon0/device/temp5_max

But sensors_temperatures() return {}:

$ ipython
Python 2.7.5 (default, Nov  6 2016, 00:28:07) 
Type "copyright", "credits" or "license" for more information.

IPython 4.2.0 -- An enhanced Interactive Python.
?         -> Introduction and overview of IPython's features.
%quickref -> Quick reference.
help      -> Python's own help system.
object?   -> Details about 'object', use 'object??' for extra details.

In [1]: import psutil

In [2]: psutil.sensors_temperatures()
Out[2]: {}

[sethmlarson]

It's because of this line in psutil/_pslinux.py:

    basenames = sorted(set(
        [x.split('_')[0] for x in
         glob.glob('/sys/class/hwmon/hwmon*/temp*_*')]))

I'll submit a PR to fix this issue.

[giampaolo]

This makes me sad. I wonder why /sys/class/hwmon/ differs like this.
@SethMichaelLarson your PR breaks on my Ubuntu 16.04 which has no "device" intermediate directory.

[sethmlarson]

Doesn't /**/ in a glob pattern include no directory?

[giampaolo]

No. :( It's probably best to use os.walk and filter for file names starting with "temp_*" so that the tree depth doesn't matter.

[sethmlarson]

Dang! I could probably whip something up using os.walk. Probably would be a lot more efficient than just globbing with a star pattern anyways.

[giampaolo]

Performance isn't really a concern in this case. The whole tree should be way less than 100 files/dirs in total.

[giampaolo]

Can you try 561f32a?
I don't really like it as a solution but I can't think of anything better.

I ruled out os.walk because there are symlinks making the whole tree too big to parse. After all you were right: os.walk is too slow.

[sethmlarson]

@giampaolo I figured, os.walk is a os.listdir call for each directory and an os.stat call for each file and directory. They made os.scandir for a reason! :P Glad you found something that works faster. :)

[giampaolo]

So did you test this? Does it work?

[sethmlarson]

@giampaolo I don't have a CentOS system. User who opened the issue does.

[nicolargo]

@giampaolo Works like a wharm on my CentOS 7.1 system:

In [1]: import psutil

In [2]: psutil.sensors_temperatures()
Out[2]: 
{'coretemp': [shwtemp(label='Physical id 0', current=34.0, high=80.0, critical=98.0),
  shwtemp(label='Core 0', current=31.0, high=80.0, critical=98.0),
  shwtemp(label='Core 1', current=34.0, high=80.0, critical=98.0),
  shwtemp(label='Core 2', current=33.0, high=80.0, critical=98.0),
  shwtemp(label='Core 3', current=32.0, high=80.0, critical=98.0)]}

[giampaolo]

Cool. I will package a 5.1.3 version soon.

[giampaolo]

OK, new version is out.