Re: new "sv status" flags and exit-tracking patch, and misc.

[Rafal Bisingier <ravbc@man.poznan.pl>]

[-- Attachment #1: Type: text/plain, Size: 1599 bytes --]

On Fri, Dec 09, 2005 at 03:29:36PM +0100, Gerrit Pape wrote:
> On Thu, Dec 08, 2005 at 01:37:51PM -0600, Charles Duffy wrote:
> > Gerrit Pape wrote:
> > >I guess ./run and ./finish almost always will be scripts, more precisely
> > >shell scripts.  runsv as of now maintains two files supervise/stat and
> > >supervise/pid.  How about removing these files, and add a single one
> > >containing the informations discussed here?:
> > >
> > >./supervise/info:
> > ><state> <want> <pid> <uptime> <runrc> <finishrc>
> > 
> > Sounds reasonable. Are you willing to impliment this yourself, or do you 
> > want me to?
> 
> Contribution alwys is nice, and I'm rather busy these days.  So I would
> be happy if you can put work into it.  Thanks!
> 
> <uptime> will not work though in a file that's only written when some
> status changes, this should be <starttime> instead I guess.  And you
> suggested to differentiate the run/finish return codes by 'exited
> cleanly', and 'killed by signal'.  To make numeric tests (`test "$runrc"
> -lt 0`) work, how about 0-255 for 'exited cleanly', and -SIGNO for
> 'killed by signal'?  What if ./run or ./finish did not exit yet?

If <state> will be different for ./run and ./finish time, then one can
simply check:
- if <state> is run, then runrc is for the previous execution of ./run
  and analogical for finish
- if runrc/finishrc is '-' then the appropriate script didn't exit yet

That's why <state> should differentiate executing ./run from executing
./finish, and runrc/finishrc fields shouldn't be empty

-- 
Rafal Bisingier

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]