Re: gcc+gcj build-times?

[Andreas Tobler <toa at pop dot agri dot ch>]

Mark Wielaard wrote:
> Hi,
>
> On Wed, 2006-02-22 at 17:58 +0000, Andrew Haley wrote:
> > OK.  It's been a problem for most of my working life that some
> > developers have fast boxes heavily laden with RAM, and need reminding
> > about what other people experience.  :-(
>
> For builder.classpath.org which also only has 500MB available we use the
> following gmake patch/workaround to get the build to complete (otherwise
> it would just trash the box and never finish):
> http://savannah.gnu.org/bugs/?func=detailitem&item_id=15182

For the record, I patched the gnumake on my sparc64-freebsd current 
machine and the step where gmake was stalling for a long time went down 
from 31' to around 6'. Great.
But, I do still not understand why we do this step again and again. 
Means, the first time is ok :) but when I do a (g)make check, the 
compile_classpath is executed again and the target compile-classes is 
built again. Although I didn't change anything.
The same happens if I do an (g)make install or another (g)make after 
changing a non dependant file. This is regardless of the os I try, it 
happens on linux, darwin, solaris or as mentioned on freebsd. Freebsd 
shows as most annoying since it is develwork in progress and it is the 
slowest machine.

I'll need to sit down and debug a bit ....

Andreas