Karl Fogel <firstname.lastname@example.org> writes:
> Bob Miller <email@example.com> writes:
> > Here's a third issue.
> > 3. If configure can't find Berkeley DB, it should die horribly
> > with a good error message. Make will die horribly later on,
> > anyway...
> Well, it should just warn the user that it won't be able to build the
> filesystem (or the server side, more generally) and move on. Or
> something like that... Since the client doesn't depend on Berkeley DB,
> it shouldn't pay for db's absence.
I agree with Karl that it should just disable the server-side stuff.
At the moment, as an intermediate step, the configure script is
supposed to die if it can't find Berkeley DB. That's what the call to
SVN_CHECK_BERKELEY_DB is there for, along with ac-helpers/berkeley-db.m4.
If configure succeeds, even when GCC can't find <db.h>, then there's a
mistake in that autoconf macro. It does build and run a program that
links against Berkeley DB, so I'm confused as to how it could possibly
succeed, and then have libsvn_fs fail.
Received on Sat Oct 21 14:36:08 2006