Flexible include (was: Re: adding functions to function.c

[Alessandro Vesely]

Boris Kolpackov wrote:
>
> [...chicken and egg example...]
>
> If I can't load bootstrap.make I can't load rules to build DSO with
> $(lastword). You would probably say that the solution would be to
> have a "simple" build system (written on standard GNU make) to build
> DSO. But then you might as well build custom GNU make and reexecure
> your real build system using this custom make. As you can see the
> difference is not qualitative. Just some thoughts.

Included makefile can have prerequisites that in turn need to be built
That may also happen in non-bootstrap situations, e.g. if one generates
dependencies in some complicated way that requires a specific order.

Sometimes $(eval) can be used to include dependencies after they have
been generated. Some other times it is either too hard or plainly
unfeasible to do so and one has to resort to a recursive call. I think
the latter is what Boris calls a "qualitative difference".

Would it be a feasible and useful to have a delayed include? I mean
to consider the include-file as a regular target but only include it
after it has been either rebuilt or diagnosed up to date.