Re: reread config

I don't know if anyone else has tried, but I've definately thought about it for 
quite a while.  As you've seen, it's not easy.  What's really a bummer is that 
there's a statically allocated array that holds all the really interesting 
stuff.  So, each child will have to re-read the configuration file or there 
needs to be some way for them to receive "updates" of sorts or told to not 
manage certain consoles or whatever.  See, the real question is, if someone 
adds a console to a .cf file somewhere in the middle, everything gets shifted.  
The processes, if spawned newly, would each manage a new console and lose one 
as well (each get a set in the order of the .cf file).  Those connected to 
consoles have TCP connections with particular child processes.  Do you drop all 
connections and have things re-read and have the children know what group 
they're supposed to manage or do you feed a message to the children 
saying "don't manage this any more" and "manage this now" or do you spawn a new 
child for the newly added consoles and let the new child control all the new 
consoles.  It's kind of a horrid problem.

If you really want to attempt it (and I think it would be wonderful), I'd love 
to know what ideas you have (and maybe the list in general) so that we can all 
punch holes in any ideas...it's something that I think requires a whole lot of 
thought, because of the underlying architecture.  One thought is that 
underlying architecture needs to just change, so that this is easier.  I wish I 
had some basic ideas on how to improve things, but I haven't had a chance to 
think about anything but the pitfalls...hence my paragraph above.  Anyway, good 
luck, and please keep at least me informed...it's a challenge I'd love to see 
conquered (and helping with the idea/planning phase is something I'd love to 
do).

Bryan

Quoting Doug Hughes <doug@gblx.net>:
> I've gone part way through making a patch for rereading the config
> file on a USR1 signal. Then I looked at the readcfg routine.
> Lots of dynamically allocated stuff. So, I thought, before I go
> any further, I'd ask whether anybody else has already done this
> before I start trying to redo dynamic memory, etc.
> 	Doug