On Fri, Jan 05, 2007 at 12:49:59AM +0000, Stuart Henderson wrote:
> > > OK, I have a fair amount of experience with FreeBSD kernels, but none
> > > using OpenBSD. My platform is a teeny little Zaurus, and I am trying
> > > to see if I could use some tools such as ccache to speed compilation.
> >
> > Yes - there is one big 'why' in all this. ccache is really useful if you
> > repeatedly rebuild the same thing, and cannot afford to actually rely on
> > make/the makefiles doing the right thing. However, unless you are going
> > to do some hefty development work on that Zaurus, which I really
> > wouldn't recommend, you are unlikely to need to build much of anything
> > on it.
>
> Quite a few things need building from ports on the ARM arch's (zaurus
> and armish share packages; building on an N2100 is a reasonably easy
> and not horrendously expensive way to speed up builds of software to
> run on a Z).
You are, of course, right - and I shouldn't have written the above, as I
was well aware of this. Still, while one would need to build some ports
on zaurus, one wouldn't need to *rebuild* a lot of ports on a zaurus;
and only in the latter case is ccache useful.
> It's a good job some people did some hefty development work on them,
> or we wouldn't have OpenBSD/Zaurus which is a pretty useful thing.
Certainly, it's a good thing that we have OpenBSD/Zaurus; and yes, this
took some hefty development work, I am sure. However, given the
paragraph by the original poster quoted above, I don't think it likely
that he will be hacking OpenBSD/Zaurus soon; and general development
work is likely to be far more pleasant over an ssh link. (Of course, one
does not always have internet handy; but the whole `mount src and obj
directories from a big server' mentioned in the first post isn't going
to work in that case, either.)
I'm just trying to understand why ccache, which was written to solve a
very specific issue (recompiling the same code lots of times, without
being able to rely on make doing the right thing), would be useful in
this case; if it doesn't give a largish benefit, it's probably not worth
doing.
Joachim
