der Mouse --> tech-userlevel (2006-05-25 16:21:55 -0400):
> > The named rc script from the base system and the one from
> > pkgsrc/net/bind9 have the same rcorder keywords, so there's no
> > difference in how they are handled by rcorder.
> 
> To the extent that that's true, it's an accident of the current
> implementation.  There is nothing in rcorder's interface contract[%]
> that bears on how it orders scripts that do not have any constraints
> compelling their order - basically, how it turns the partial order
> deduced from the keywords into a total order.  It's perfectly free to

Sure, but the named scripts in question both have `REQUIRE: SERVERS'
while SERVERS has `REQUIRE: mountcritremote'. That should be enough
to make named start correctly for the OP (assuming his named chroot
dir is at least in `critical_filesystems_remote').


> sort them by MS5 hashes of the script contents and time-of-day, if it
> feels like it, which would mean that any script change can make it run
> at a different place at startup - and indeed it could run at a
> different place next startup without any changes to files.  (This
> unpredictability is one of the reasons I don't like rcorder.)

What about adding `PROVIDE: n' and `REQUIRE: n-1' to each rc script?
;-)


Cheers, Jukka

> [%] Based on the 3.0 rcorder manpage, at least.

-- 
bashian roulette:
$ ((RANDOM%6)) || rm -rf ~