Subject: Re: make includes optimization
To: None <erh@netbsd.org>
From: Chris G. Demetriou <cgd@netbsd.org>
List: tech-install
Date: 02/13/2000 13:51:19
erh@nimenees.com writes:
> > 	Objections?  Improvents?  Where should I put the scripts?

(1) using '.if ... make( ...' for anything, ever, is broken.  yes, i
know that some things already use it, but they should be considered
broken.  (what happens if you say 'make includes all install' or
similar?)

(2) the existing implementation is kind of nice in several ways:

	* simplicity (the fact that you want a script like the one
	  you've suggested is an example of complexity).

	* ease of maintenance.  Need an include file, add it,
	  and the right thing happens.  No need to troll through
	  the source tree to mark things in the right way.
	  (I'll bet that if a change like this is made, there
	  will be at least a few screw-ups with things that should
	  be installed not being installed because somebody forgot
	  to frob the right makefiles.)

(3) it'd be interesting to see how the time wasted by extra work in
'make includes' relates to other chunks of the build process.  i'd
rather waste a relatively small amount of time doing the build to get
a known-good-and-maintainable build process.  This only impacts
developers (i.e. people who hack the system or run -current), and so,
since i'm wagering it's not a gigantic expense relative to the rest of
the build, it's not really worth optimizing in my eyes.  If you're
interested in making builds faster on ancient hardware, or something,
make cross-builds work.  it'll help more, and will go further as
faster hardware becomes available, even.





cgd
-- 
Chris Demetriou - cgd@netbsd.org - http://www.netbsd.org/People/Pages/cgd.html
Disclaimer: Not speaking for NetBSD, just expressing my own opinion.