Re: adding functions to function.c

[Alessandro Vesely]

Boris Kolpackov wrote:
> 
> Alessandro Vesely <address@hidden> writes:
> 
> > Boris Kolpackov wrote:
> > > [...]
> > > The idea is much simpler than that: GNU make needs scripting facility.
> > > Instead of reinventing the wheel (by implementing our own) we can just
> > > reuse something suitable. Guile seem like the best choice so far.
> >
> > Make could also load a shared object that contains user defined functions,
> > a la PHP.
> 
> For most tasks it is too low-level; it is much easier to write your function
> on lisp than to write it on C, build DSO, load it into make, etc.

I don't think so. Actually, writing a short C prog that does what's needed
is a common technique. The prog has access to selected make's variables
(those that you expand on its command line) and returns values via $(shell).

The more exotic functionality of calling specific entry points of a DSO
is similar in principle. 

I don't know how you're going to implement guile support in make. Perhaps,
if DSOs loading already existed, then that could be achieved by simply
defining guile as a preloaded DSO (arranging for the not quite standard
behaviour of gh_enter().)

> Also I believe Guile supports loading DSOs and calling functions from them.

Yes, it does.
http://www.gnu.org/software/guile/docs/guile-ref/Dynamic-Libraries.html

However, that will not check if the DSO is up to date.

> 
> >The so may build on the fly, as it happens with included files.
> 
> I was thinking about this too but realized that it often creates a chicken
> and egg problem [...]

OTOH libguile.so will still need to be built with make...