Peter Eisentraut <peter_e@gmx.net> writes:
> Tom Lane writes:
>> Counterexample: src/test/regress is cleaned by 'make clean', but not
>> built by 'make all'. I think contrib should behave the same.
> That's because 'make check' is an exception since you're supposed to run
> that as part of the normal installation procedure.
Yes, and "make distclean" is part of the normal tarball building
procedure. I want a guarantee that there will be no cruft in the tarball.
If contrib isn't cleaned out then we don't have that guarantee.
> However, the point is that contrib should be separated from the "real"
> source. Perhaps the makefiles aren't even functioning. Are we going to
> guarantee that? Maybe we should, but it wasn't the case in the past.
If "make clean" invokes the same in contrib, we'll quickly find out if
there are any broken "clean" targets in contrib ;-). This is a lot
different from depending on the contrib stuff to *build*. You'll note
I did not propose that toplevel "make all" should recurse into contrib...
regards, tom lane