Comment 16 for bug 1124384

Revision history for this message
Dimitri John Ledkov (xnox) wrote :

ppa:xnox/backports has the updated upstart package that has the following fixes:

* reload configuration always works and all events are preserved
* the postinst script was modified to not execute stateful re-exec if runlevel 2 has not been reached yet

The reason for the latter, is that we still loose events that are blocking jobs that are not running yet across the stateful re-exec, because EventJobOperators (aka start_on & stop_on conditions) are not serialised to include progress information w.r.t. their states (e.g. TRUE or FALSE for each condition to tell apart if it has fired or not yet).

The above fixes allow to always run reload-configuration, including early boot.
And safeguard against automatic re-execution if upstart package is upgraded during early boot.

To fix stateful re-execution in early boot, we need to serialise all objects used by upstart which upstream has avoided to do so far. Also it's questionable if stateful re-execution in early boot is a desired requirement.