Actually that workaround referred to *if_msghdr* but recent changes have been to *ifa_msghdr*: https://alioth.debian.org/scm/viewvc.php/trunk/glibc-ports/kfreebsd/if_index.c?root=glibc-bsd&r1=1716&r2=2484
so the problem is not quite what I thought. I think I've been seeing two bugs at once. This history of this is: struct if_msghdr contains a (pointer to a) struct if_data (since forever) which will be either a 32- or 64-bit sized member depending on the architecture. ABI would thus be incompatible between a 32-bit userland and 64-bit kernel. Building glibc in a kfreebsd-i386 sid chroot, on a kfreebsd-amd64 9.0 kernel, fails test_ifindex due to the above issue. Building glibc in a kfreebsd-i386 sid chroot, on a kfreebsd-i386 9.0 kernel like the Debian buildds have, is fine. But then userland tools like ifconfig don't work properly on any 10.x kernel. Building glibc in a kfreebsd-amd64 sid chroot, on kfreebsd-amd64 9.0 or 10.0 kernels is fine, and userland tools work fine on whichever kernel. Making this particularly awkward, however we fix this for kfreebsd-i386, it probably still needs to work on a 9.0 (wheezy) kernel because the buildds must run that until jessie is released. Regards, -- Steven Chamberlain ste...@pyro.eu.org -- To UNSUBSCRIBE, email to debian-bsd-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: https://lists.debian.org/20140921193421.gf31...@squeeze.pyro.eu.org
