On Wed, Oct 21, 2009 at 11:18:22AM +0200, Thomas Jansen wrote: > On Tue, Oct 20, 2009 at 01:26:15PM +0200, Diego Biurrun wrote: > > On Mon, Oct 19, 2009 at 10:47:11PM +0200, Jan Marten wrote: > > > --- trunk/community-operator/Makefile.am Mon Oct 19 21:51:52 2009 > > > (r1239) > > > +++ trunk/community-operator/Makefile.am Mon Oct 19 22:47:11 2009 > > > (r1240) > > > @@ -20,6 +20,7 @@ > > > INCLUDES += -I@PISA_HIPL_SRCDIR@/libhiptool > > > INCLUDES += -I@PISA_HIPL_SRCDIR@/libdht > > > INCLUDES += -I@PISA_HIPL_SRCDIR@/opendht > > > +INCLUDES += -I@PISA_HIPL_SRCDIR@/firewall > > > INCLUDES += -I@PISA_HIPL_SRCDIR@/hipd > > > INCLUDES += -I@PISA_HIPL_SRCDIR@/i3/i3_client > > > INCLUDES += -I@PISA_HIPL_SRCDIR@/pjproject/pjnath/include > > > > All of this is very ugly. > > > > Why don't we just use the full relative path in the #include statements? > > Even if _we_ do, HIPL doesn't IIRC. The problem is that we include one or two > HIPL headers, which in turn include further HIPL headers that expect that each > required subdirectory is available as -I$PATH. Well, "because they suck" is never a good answer to the question "Why do we suck?" :) For some reason overusing -I is very popular. I guess once autotools have made your brain rot for long enough people just don't notice ugliness any longer... > The most elegant solution would > be getting rid of HIPL dependencies altogether. Yes, that should be the goal. Otherwise PISA should just be another HIP branch. > Unfortunately I think the hipconf calls we use in our applications are > not part of the RFC but HIPL specific. :-/ Diego