Re: build system rules & algorithms

[Mike Shal]

On 6/10/09, John Calcote <address@hidden> wrote:
> One more comment Mike - I think it would be most instructive (not to mention
> most interesting to the GNU make community) to modify GNU make to use "Beta
> build system" algorithms. I don't really see any conflict (off the top of my
> head) between beta algorithms and makefile syntax, so I may just take a stab
> at it.

Thanks for the comments John! The main problem I think you would run
into with make is that the Makefiles and included dependency files are
fairly free-form. While this is certainly nice when you're writing
them, I think it is difficult to come up with a way to only load the
parts of the graph that are necessary even if you have a list of file
changes up front. If you take a C program and provide the input
"foo.h" has changed, which Makefiles need to be parsed in order to
make sure all links to foo.h are satisfied? If you're using an
autodependency generation method (like with 'gcc -MMD'), any of those
generated .d files could list foo.h as a dependency.

I'd love to be proven wrong though! Perhaps you can come up with a
good way to restructure the information so make can only load what it
needs.

Also, in response to your other email, if you have any questions about
tup specifically feel free to ask that on the tup-users mailing list
linked on the webpage. I just made that like a week ago though, and
it's currently only me and a few of my friends who I forced to sign
up.

-Mike