alias stanza wip patch

[James Hunt]

Hi Marc,

On 21/06/11 16:50, Marc - A. Dahlhaus wrote:
> We simply remove the JobClass of the removed config form the alias list
> inside of the known_job and if  ->job and ->alias are empty after that
> we remove the entire entry from known_jobs in that case...
Having re-read your mail here, I think I have been "thinking at crossed purposes" here - I was
fixating somewhat on JobAlias rather than NamedJob.

I see what you are saying now. This is a good idea, although we would still be storing job *names*
in 2 locations (maybe not such a big deal): in JobClass->name in job_classes and also in NamedJob->name.

Nominally, the NamedJob struct would also disallow aliases with the same names as a job (assuming
nih_hash_string_new()). This is probably the behaviour we want anyway though.

> We don't have to change that. As the ->job member above is exactly that.
> We have an array behind ->aliases that would be filled with jobs that
> provide a given alias. We just need to extend the lookup-logic for that
> list that is used by the signals initctl sends to check which job should
> handle the signal.
Right.

I'll try and grab you on irc this week or next so we can discuss this further.

Speak soon.

Kind regards,

James Hunt
____________________________________
http://upstart.ubuntu.com/cookbook
http://upstart.ubuntu.com/cookbook/upstart_cookbook.pdf