Re: Google summer of code for GNU make: introducing myself.

[Tom Tromey]

>>>>> "David" == David Boyce <address@hidden> writes:

David> I'm not sure if it's not already too late to make alternate
David> suggestions, but I'd prefer a design which does not require
David> modifying the makefile. After all, the method of determining
David> out-of-dateness (hereafter OOD) is orthogonal from what the
David> makefile encodes, which is simply (a) what files depend on what
David> other files and (b) what commands to run when an OOD situation
David> is discovered.

Unfortunately that isn't as true as we might like.  Many, many
Makefiles work around the current lack of content signatures using the
stamp file technique.  (By "many" I mean, to first approximation, all
Makefiles generated by automake.)

I think the stamp file approach would break with a content-based OOD
algorithm, because the stamp file contents do not change.

Naturally, this can be fixed.  (And perhaps automake rewritten as a
set of GNU make macros ... but that's for another summer :-) Still, it
does indicate that changing algorithms is not transparent.

Tom