On Tue, Nov 25, 2003 at 08:05:57AM -0800, Greywolf wrote:
  | Thus spake Luke Mewburn ("LM> ") sometime Today...
  | 
  | LM> Our "defacto" mechanism is that variables are tested against "no",
  | LM> even "yes/no" variables, because there's no simple way to enforce
  | LM> that such variables may only contain those two values, and it's easier
  | LM> to have a single consistent check (== "no", or != "no") than cater for
  | LM> all the various alternatives.
  | 
  | Minor observation:
  | 
  | MOST variables are thus tested, except for the few which are tested
  | against (re_type) .if[n]*def, something which has caused me great
  | consternation and confusion on occasion ("Hey, I just defined this
  | to be 'no', why's it still defining/including this stuff?").

Most of the "user tweakable" variables ("MKxxx", etc) are tested
against "no".

Some of the "user tweakable" variables ("NOxxx") are tested for
existance, and I've been migrating these to the former category
over time, because of the various problems with these, including
the inability to undefine from the make(1) commandline or
otherwise manipulate them.


  | Perhaps it would be better to have one file in /usr/share/mk
  | (bsd.defaults.mk? bsd.init.mk?) which sets up all the default values
  | before /etc/mk.conf gets included, with no conditional checking (leave
  | that to the "real" .mk files to fix up)?  Just a thought...

That is what <bsd.own.mk> is for.