Re: x86 release builds are slow

["Steven M. Bellovin" <smb%cs.columbia.edu@localhost>]

On Tue, 6 May 2008 12:36:06 +0200
Manuel Bouyer <bouyer%antioche.eu.org@localhost> wrote:

> On Tue, May 06, 2008 at 01:23:33AM -0700, Tom Spindler wrote:
> > > Because if you're like me, you'd like more people to do full
> > > builds before checking in changes :)
> > 
> > To chime in to what he said: around 30% of the changes I commit are
> > due to full buids not compiling. (About 2/3rds are for breakage in
> > the less-popular architectures.)
> > 
> > So, yeah, I too would like fewer kernels to build for the release.
> 
> removing kernel would make some breakages dissapear from the default
> builds, not avoid the breakages themselves.
> 
That's true if the way we speed things up is by avoiding compiling
certain files or options.  I agree that that would be a bad approach.
OTOH, if we can speed things up structurally -- if, for example, we
could share .o files among many kernels, either by using LKMs more or
by a different build strategy per my suggestion (and yes, both are
difficult) -- there wouldn't be a problem.  We'd still be compiling
everything once; we simply wouldn't be recompiling the same things over
and over again.


                --Steve Bellovin, http://www.cs.columbia.edu/~smb