Re: [log4perl-devel] Config file reload on signal

[Mike S. <msc...@ao...>]

Wiggins d'Anconia wrote:

 > I like the idea but I think this goes back to the notion of
 > init_and_watch calling die on failure. So two quick questions,

You've got a lot of good points. There's a bunch open questions which we 
need to find answers for before we proceed (partially just summarizing 
your points):

[1] Are we ok with the way Log::Log4perl handles re-initializing the 
configuration by throwing out everything and starting from scratch or 
should we rather go with the log4j model of a somewhat more complicated 
reload?

[2] What's supposed to happen if a config reload fails?
a) die
b) keep the old one (and if: how can we notify the user? If it's a daemon?)

[3] How does that fit in if we upgrade Log::Log4perl to handle logger 
repositories?

Couple of comments on your questions:

 > 2) Is it possible to compartmentalize the config reading+updating, etc.
 > code to the extent of rather than passing init_and_watch a signal to
 > catch, that a particular class method could be called instead

Sounds good, we should offer that. And, as you said, a callback to 
init_and_watch which calls back into the application when the re-init 
happens.

 > Currently I am using Log4perl extensively in a POE framework app (by the
 > way it works beautifully, well with the exception of FileRotate

FileRotate has been updated a couple of times recently, as far as I'm 
aware (FileRotate isn't part of Log::Log4perl) it should be stable now. 
Grab the latest from CPAN.

Lots of stuff ahead - and we're always welcoming helping hands :)

-- 
-- Mike
Mike Schilli
m...@pe...