why does setting stderr=subprocess.STDOUT fix a subprocess.check_output call?
- by ShankarG
I have a python script running on a small server that is called in three different ways - from within another python script, by cron, or by gammu-smsd (an SMS daemon with the wonderful mobile utility [gammu]). The script is for maintenance and contained the following kludge to measure used space on the system (presumably this is possible from within Python, but this was quick and dirty):
reportdict['Used Space'] = subprocess.check_output(["df / | tail -1 | awk '{ print $5; }'"], shell=True)[0:-1]
Oddly enough this line would only fail when the script was called by a shell script running from gammu-smsd. The line would fail with a CalledProcessError exception saying "returned exit status 2", even though the output attribute of the CalledProcessError object contained the correct output. The only command in the sequence of shell commands that would give such an error status would be awk, with status 2 indicating a fatal error.
If the python script with this line was called by cron, by another python script, or from the command line, this line would work fine. I broke my head trying to fix the environment for the script, thinking this must be the problem. Finally though I put in stderr=subprocess.STDOUT, like so:
reportdict['Used Space'] = subprocess.check_output(["df / | tail -1 | awk '{ print $5; }'"], stderr=subprocess.STDOUT, shell=True)[0:-1]
This was a debug measure to help me figure out if some output was coming on stderr. But after this the script started working, even when called from gammu-smsd! Why might this be the case? I ask for future reference when using subprocess...
