Re: non recursive makefiles

[Helge Hess]

On Jan 12, 2005, at 16:47, Nicola Pero wrote:
> For me, GCC and autogsdoc are the performance killers
> during compilation, make is fast enough even on a large tree.  I can't
> turn off GCC :-)

I don't think that recursion was "attacked" because it is too slow but 
because it prohibits -j 2.
And the latter will roughly double the performance of your GCC and 
autogsdoc on a two CPU system, which is very common (since this can be 
distributed very well on two (or more!) processors).

Recently I joined some KDE developer meetings and they are using the -j 
X feature aggressively. The have a build distributor called "Icecream" 
which manages build tasks on tens to hundreds of machines.
Useful KDE development is basically impossible without the -j X option 
(and several compilation slave machines) since g++ is even slower than 
cc1obj.

Personally I have no immediate need for -j X, compilation with 
gstep-make is "fast enough" for me. The compilation time is more an 
issue on my MacOSX machines (which provide the -j feature ;-), the 
cheapest Linux/ix86 machines with gstep-make are almost as fast as the 
most expensive Macs with Xcode/pbxbuild.


Having said all that I'm 100% with Nicola, usability of the GNUstep 
makefiles is GREAT. It should not be broken to gain some compilation 
speed.

Personally I would be interested how long a straight forward port of 
Ant to ObjC would take (just rewrite the Java classes into 
Objective-C). Nicola is right that inventing a new build system is a 
lot of work, but Ant is supposed to have a pretty good engine and 
modelling.
The only "issue" is that it uses XML as the build description which can 
be awkward (but also can be fixed by using a different frontend).

Greets,
  Helge
--
http://docs.opengroupware.org/Members/helge/
OpenGroupware.org