Re: Revision of GNU Parallel's processing of SIGTERM

[Martin d'Anjou]

Ole,

I have thought about this more, and run some experiments, and I have 
determined that it is better for me to enhance GNU parallel than to 
write my own scripts around it or around other competing tools. I have 
started working on some of these features, but I would be much more 
enthusiastic if I had some kind of assurance from you that you would 
accept then in a future release of GNU parallel.

I volunteer for doing all this work, documentation and tests included.

Here is a summary:
- Ability to change the termination sequence and its delays when the 
hardcoded one does not work for all cases
- Ability to kill the top parent process before killing the entire 
process tree
- Ability to kill the running jobs when SIGTERM is used (right now they 
keep running)
- Ability to kill the running jobs when --halt value% determines that no 
more jobs should be run
- If no option is selected, the old behaviour remains

Cheers,
Martin