Killing-Process woes

[Ronald Fischer]

> > The background processes are actually (zsh-) scripts, which do some
> > setup (basically setting various environment variables), and then invoke
> > a (Cygwin-)Ruby program which does the "real work". The program is
> > executed by something like
> > 
> >     ruby myprog.rb
> > 
> > (Note that this Ruby program is NOT invoked in background).
> > 
> > When my SIGINT trap is entered, I can see from ps indeed the
> > relationship between the processes involved, for instance
> > 
> >     10852    9296    6224      10536  cons3    3672028 08:05:10
> >     /usr/bin/ruby
> >      9296    6224    6224      11236  cons3    3672028 08:05:10
> >      /usr/bin/zsh
> > 
> > The PID of my background process - the zsh wrapper - in this concrete
> > case is 9296, and we can see that this is the parent of the Ruby
> > process, 10852. The problem is that if I just kill 9296, the Ruby
> > process keeps running, orphaned:
> > 
> >     10852       1    6224      10536  cons3    3672028 08:05:10
> >     /usr/bin/ruby
> > 
> > I've found on Stackoverflow the suggestion to treat this as a process
> > group and use negative PIDs. I tried this too, but it didn't work. Here
> > is a similar example:
> > 
> 
> Not implemented as you found out below.  But I don't know that the
> negative process number is in use anywhere.  Are you sure it wasn't a
> signal number as a option to kill?

No, the article refered to a process group (and this indeed would be
done by negative PIDs), but as I said, this didn't work anyway.

> Perhaps use the -f --force switch might help.

No, doesn't help either.

For the time being, I have reverted to analyzing the output of ps. It is
pretty tedious:

# Get the PID of the shell script 
local wrapper_proc=$!
# Give the wrapper some time to start the Ruby process below. Without
this, the
# Ruby process would not be visible yet.
sleep 3
# Find out the PID of the child process of the wrapper
local sub_pid=$(ps |grep -oE "^ *[0-9]+ *$wrapper_proc "|awk ' {print
$1}')
# Sanity check ....
if [[ $sub_pid =~ ^[0-9]+$ ]]
then
  # Add this to the array of these child processes
     additional_pids+=$sub_pid
 else
     echo "Info: Could not extract VP pid from '$sub_pid'"
 fi

Inside my SIGINT trap, I do not only kill the processes found via
$jobstates, but also the processes collected in $additional_pids. An
awful solution, and one which is not easy to maintain and may break!

Ronald

--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple