Hi Ken,
----- Original Message -----
> ...
> ps: I currently have a clean QA run with these changes, so I'm
> also keen to hear of any fallout in that department - thanks!
>
This turned out to be optimistic - a number of tests were passing
because of existing pmlogger.default configuration files (due to
package upgrades, and due to not the packages not installing some
files anymore).
So, at least backward compatibility is maintained! However, once
the old configs are moved aside, and the default pmie/pmlogger's
start up (with their freshly generated config files), there is a
bit of resultant QA fallout.
I've been working through the issues, but there's one I'd like to
pick your brain over. Test qa/107 is failing now and its because
the pmlogconf generated config uses "log advisory on" and not any
"log mandatory on" clauses (as the old default config did). This
seems to cause test 107 to come away thinking nothing is being
logged at all (pmlc reports "log advisory off" for some reason?)
even though the original config specified "advisory on" and I have
verified that data *is* actually being logged.
Is this a pmlc issue? Or should pmlogconf generate configurations
that specify "mandatory on"? Or should the test be changed so it
accepts "adv off nl" as "something is being logged after all"?
Also, when pmlogconf generates a configuration file initially (so,
when no pre-existing one found where specified) I've changed it to
add a default [access] section - could you review that too please?
Have attempted to specify no-remote-access-allowed-by-default, and
also no-local-changes-allowed-by-default.
thanks!
--
Nathan
|