Public SourceMage Discussion List

[seth AT swoolley.homeip.net]

On Tue, May 27, 2008 at 07:10:51PM +0200, Treeve Jelbert wrote:
> On Tuesday 27 May 2008 18:56:27 Arwed von Merkatz wrote:
> > On Mon, May 26, 2008 at 08:39:24AM +0200, Thomas Orgis wrote:
> > [...]
> >
> > > PS: When the GCC spells are so tricky with their consistency, would
> > > it be beneficial to have _one_ gcc spell that optionally installs
> > > g++, gfortran and the like; replacing g++ depends by sub-depends on
> > > this one? It would at least ease updates; perhaps even speed up the
> > > overall build as some synergy of common compile work would come into
> > > efect?
> >
> > It was like that originally, but got split so spells could depend on
> > specific compilers (we didn't have sub_depends back then). When
> > subdepends were introduced, the decision was to still keep the gcc
> > spells separate so you don't have to recompile the whole thing just to
> > add one compiler (especially annoying when you include the long-to-build
> > frontends like gcj and ada).
> > Right now I don't have a strong opinion either way, both solutions have
> > advantages and disadvantages.
> 
> I actually have a local version of the spell which integrates things. 
> g++ is then a dummy spell which uses 'depends -sub c++ gcc'
> 
> I find this setup is much faster when rebuilding things. I don't use gcj, 
> fortran or ada.

I think in the whole scheme of things, most people will be rebuilding 
(for upgrades and compiler option changes) more often than adding a 
compiler type, so the subdepends idea is the one that will reduce 
compilation time for most people, particularly in the long run.

The virtual subdepends spell is also a perfect transition plan.

And somebody's already written it up.  Seems like a good time to switch?

Seth