Closed TimothyThomas closed 6 years ago
Thanks Timothy for the awesome bug report. I think this is a python2/3 difference:
[rory:~/tmp/foo] $ python3 -c 'from __future__ import print_function; import subprocess; print(subprocess.check_output(["ls"]));'
b'EXAMPLE\n'
[rory:~/tmp/foo] $ python2 -c 'from __future__ import print_function; import subprocess; print(subprocess.check_output(["ls"]));'
EXAMPLE
I pushed a fix for this that passes the universal_newlines option. Thanks so much!
Aha! I didn't consider possible differences between Python 2/3. Thanks so much for the fix and your work on this project as a whole.
When I use
scheduler='pbspro'
the engines start up fine and do their work, but the ipcluster dies and fails to shutdown the engines (see ipcluster log output below). It seems to be an issue with parsing thejob_id
returned bysubprocess.check_output()
inside ofBcbioPBSPROEngineSetLauncher
(lines 737-739). I believesubprocess.check_output()
returns bytes by default, so theoutput.strip()
operation raises an exception since it expects astr
. Passing the argumentuniversal_newlines=True
will makesubprocess.check_output()
return astr
and should prevent this error.Anyone else experience this issue? I'm wondering if it could be something else on my end, since it's surprising no one else has reported this.
The ipcluster log file has the following: