cvs commit: ports/mail/postfix-gps Makefile

[Edwin Groothuis]

On Sun, Jan 02, 2005 at 02:39:41PM -0800, Kris Kennaway wrote:
> On Mon, Jan 03, 2005 at 08:32:06AM +1100, Edwin Groothuis wrote:
> > On Sun, Jan 02, 2005 at 12:18:18PM -0800, Kris Kennaway wrote:
> > > On Sun, Jan 02, 2005 at 10:20:37PM +1100, Edwin Groothuis wrote:
> > > 
> > > > > The problem is that INDEX builds also need to set this or the
> > > > > dependency list will be all wrong.  Doing this the obvious way gives a
> > > > > huge slowdown.
> > > > 
> > > > What are the obvious ways? This is what I made in a galaxy far far
> > > > away when I still had the hope I could change the world. What is
> > > > does is when PACKAGE_BUILDING is defined it sets the default WITH_
> > > > and WITHOUT_ values are being set.
> > > 
> > > And that gives a huge slowdown, because of all the extra make/shell
> > > gunk that is processed 12000 times :)
> > 
> > I couldn't find a single shell command in this.
> 
> OK (I might have been confusing it with the slowdown that happened
> along with a previously-committed OPTIONS-related change) but did you
> benchmark it?

I would like to, but at this moment I'm not sure what and how:

- It doesn't have any impact on ports when PACKAGE_BUILDING isn't set.

- It doesn't have any impact on ports which don't use OPTIONS.

- When running it without PACKAGE_BUILDING set, it takes half a
  second for databases/libdbi-drivers. When running it with
  PACKAGE_BUILDING set, it takes 2.5 seconds. Why? Because it has
  to go through the databases/mysql40-client and databases/postgresql7
  ports and find their dependencies.
  So yes with this patch it's slower, but that is because it does
  do all the intended work instead of bailing out halfway.

-- 
Edwin Groothuis      |            Personal website: http://www.mavetju.org
edwin at mavetju.org    |          Weblog: http://weblog.barnet.com.au/edwin/