Re: [Gnu-arch-users] unmatched files

[Tom Lord]

    > From: Colin Walters <address@hidden>

    > I think that the build system is generally a better place to
    > solve 'make dist'.


Why? 

Why do we want two different mechanisms both of which purport to
identify "the subset of this tree which is source?"

Both `make dist' and revision control need to know that subset: why
should I have to specify it _twice_?

Not just `make dist' -- the compilation rules can make pretty good use
of that too (though, for obvious bootstrapping reasons, don't happen
to in tla distributions).   So can `make clean'.   

Why do I have to keep identifying this subset of a tree over and over
again, in N different languages, for N different purposes?   Isn't it
a concept worth capturing in a "first class" way?


-t