set exit code 0 at the end

[javierggt]

Description

This PR sets the exit code to 0 at the end of execution.

I can confirm that the current master version return code is not zero. This causes test_regress.sh to fail. Someone who is more familiar with perl should check this.

Interface impacts

None

Testing

Unit tests

• No unit tests

Functional tests

I checked that the program runs and returns the given value using the same command as in test_regress.sh:

starcheck -dir $SKA/data/ska_testr/test_loads/2019/MAY2019/oflsa

I did not check an error elsewhere to make sure the error is properly set. I would need to know what error to check for.

[jeanconn]

I think what we want is non-zero exit code if the perl exits via any of the "die" statements and a zero exit code otherwise. Not sure if setting the exit code at the end does that and/or if the END block should preserve and re-spit-out the exit code if it gets there via die. Will look at this some more.

[taldcroft]

I think this should be at the end of the END {} block above the docs.

[jeanconn]

The problem with that, I think, is that END still runs after an unexpected "die", so you get zero exit status even on a failure.

[taldcroft]

Ah good point. Sorry, never mind.

[jeanconn]

I think maybe

END {
   my $exit_status = $?;
   ...
   exit($exit_status);
}

[jeanconn]

Also this code does work to output zero but it didn't start with master with the merges that are the problem.

[javierggt]

Also this code does work to output zero but it didn't start with master with the merges that are the problem.

That's right. It started at another commit. I rebased it.

[javierggt]

Now looking better at the code it is obvious which statement is the one that sets the RC code (to 9):

	kill 9, $pid;                    # must it be 9 (SIGKILL)?
        my $gone_pid = waitpid $pid, 0;  # then check that it's gone

[jeanconn]

I saw that nine but I wasn't 100% sure it was setting the exit code. Either way, my suggestion is in #404 . It seems to work for the success case and for the simplest "die" case of "-dir" not being a products directory.

[jeanconn]

I was reading up more on appropriate handling of errors around END but maybe we don't care about best practices.

[jeanconn]

Superseded by #404