Thomas Klausner <wiz%NetBSD.org@localhost> writes: > On Sat, Jun 09, 2012 at 03:40:42PM -0400, Greg Troxel wrote: >> So what should we do about this? It seems like inet6 should be on by >> default, and turned off on deficient platforms, rather than the other >> way around. > > I think we should change packages that have this option to have it on > by default, and get convert the users of the deprecated variable to > the option. That makes sense, but I think it isn't people that are using the variable. As I understand things: There is a USE_FEATURES/MISSING_FEATURES scheme, and inet6 is put in MISSING_FEATURES on platforms without IPv6 compile-time support. There is a mapping from USE_INET6 to option inet6. USE_INET6 is defined in NetBSD in /usr/share/mk/bsd.own.mk, and the deprecated option processing is how inet6 is enabled. So it seems like the thing to do is to decouple base system options from pkgsrc and: put inet6 in PKG_SUGGESTED_OPTIONS for all packages with an inet6 option perhaps, do something so that the inet6 option implies USE_FEATURES=inet6, so we get a compile-time failure instead of a package with a broken PLIST. optional, can be done later remove the legacy mapping for USE_INET6 This will mean people with NetBSD (and maybe OpenBSD, FreeBSD, DragonFly) systems that are trying not to use IPv6 in the base system (via MKINET6=no or USE_INET6=no) will have to add PKG_DEFAULT_OPTIONS+= -inet6 but that seems pretty straightforward and less mysterious. Counting with simple egrep, I see 121 packages with inet6 options in options.mk, and 36 that have inet6 in PKG_SUGGESTED_OPTIONS (wip and pkgsrc). It's 105/34 in just pkgsrc. In Makefile*, the count is 35/3. So this is fairly easy to do, at 156 packages.
Description: PGP signature