Bojan Smojver wrote: > On Tue, 2009-06-09 at 17:29 +1000, Bojan Smojver wrote: >> PS. We do have a workaround for ldap along these lines, where we >> report ldap in --libs, unless --avoid-ldap is passed to the config >> script. Maybe we should do this here, to avoid breaking compilation >> compatibility. Opinions? > > Quick and dirty...
I'd accept that patch on a technical basis +1, although I still argue that it's totally unnecessary. The LDAP variant was necessary because the apr_ldap API is bogus. The apr_dbm and apr_dbd API's were much better thought out. Build systems shouldn't be relying on obtaining dependencies 'accidentally' from other dependencies, which is what happened here w/ Subversion. We provided a contract for a generic interface "to a database", they leveraged this by coincidence. My vote is -1 for the necessity of the patch, for a net vote of 0. let's see what the folks here believe with respect to db providers? As a packager, I can see how you might view 1.3.4-> as a minor version rev, but from the perspective of any user deploying a package linked against 1.3.4 and migrating to 1.3.7, no packages need to be recompiled. That is binary compatibility.
