On 26.11.2010, at 14:50, Diego Biurrun wrote: > On Wed, Nov 24, 2010 at 08:40:56AM +0200, Miika Komu wrote: >> >> On 23/11/10 22:18, Diego Biurrun wrote: >>> I think we need to decide what to do about the libconfig dependency in >>> the firewall. I'm not sure why only the firewall needs it to parse >>> configuration files, but hipd and hipfw should be unified in this >>> regard. >>> >>> The dependency on libconfig is also not completely enforced. We claim >>> it to be necessary in INSTALL, but not in the spec files. So the packages >>> that get built (and we distribute) behave differently depending on whether >>> or not libconfig was available on the system at build time or not. >>> >>> Since the dependency is not recorded, this might result in a package that >>> contains binaries that are linked against libconfig without declaring a >>> dependency on libconfig in its metadata. The result will be breakage. >>> >>> Thoughts? >> >> as a simple solution, what about just mandating libconfig in the spec files? > > I'm fine with that, but then we should fail in configure if libconfig is > not available and get rid of the #ifdefs that suggest that a configuration > without libconfig is supported. libconfig is currently only required for esp_token config. I would therefore like to keep things as they are (-> optional). > Also, IIUC libconfig is not available on CentOS 5.5, at least not without > configuring some extra package repositories (I'd be happy to know how). > Thus requiring libconfig basically means dropping support for CentOS 5.5. -- Dipl.-Inform. Rene Hummen, Ph.D. Student Chair of Communication and Distributed Systems RWTH Aachen University, Germany tel: +49 241 80 20772 web: http://ds.rwth-aachen.de/members/hummen