[Swan-dev] should final.sh shut down pluto?

[Antony Antony]

On Fri, Jun 24, 2016 at 10:30:13AM -0400, Andrew Cagney wrote:
> On 5 February 2016 at 16:31, Andrew Cagney <andrew.cagney at gmail.com> wrote:
> > On 5 February 2016 at 15:56, Paul Wouters <paul at nohats.ca> wrote:
> >> On Fri, 5 Feb 2016, Andrew Cagney wrote:
> >>
> >>> While this question is kind of philosophical, there's a real problem
> >>> behind it.
> >>
> >>
> >> "It depends". I like to leave it running so i can ssh  in after the
> >> test and look at the running system. Some tests requirea shutdown,
> >> for instance for leak detective.
> >
> > Don't run final.sh?  The script runkvm.py seems to even have a feature
> > that does this; I don't know if it works - I just run the scripts by
> > hand.
> 
> Based on more recent discussion about improving the test infrastructure:
> 
> - because we're constrained by the scripts that can be run we've been
> adding stuff to final.sh (I'm guilty of this :-) which really belongs
> in a second run script, for instance: run1west.sh run2east.sh (I've a
> strong preference for names that let "LANG=C ls -1 *.sh" define the
> order).

additional run scripts would be nice to have.

Some minor comments about script naming.
I think it would be nicer if the name was <nn>-<host>run.sh 
the current scripts are <host>(init|run).sh (eastinit.sh). Having eastining.sh and run1east.sh could be confusing.

Also if we do this ideally rename the current (west|road|north)run.sh to 01-westrun.sh

I wonder if it is better to add sequnce number as prefix or in the middle. 
There are already ideas about having multiple hosts named road1 road2,.. 
Where there more hosts, say road1 there will be
01-road1run.sh  or wonder is it better clearly seperate host name. 01-road1-run.sh

> - after final.sh has been run, there should be a better defined tear
> down / extract step to kill off pluto and then extract its logs (I
> suspect this fits in better with docker).

It is important not to shutdown when you want to manually debug single test.

The current setup is easier to leave the host running and manually login debug further.
I have a bias to keep doing this. Also currently when runng with TESTLIST it is alway shutdown.
What is missing in this model?