libnetfilter_acct requires kernel's netfilter accounting support which was introduced in 3.3 and can be enabled with NETFILTER_NETLINK_ACCT config option. So we need to check for it in pkg_setup() in a similar way it is done in libnetfilter_conntrack ebuild, smth like: CONFIG_CHECK="~NETFILTER_NETLINK_ACCT" pkg_setup() { linux-info_pkg_setup kernel_is lt 3 3 && die "requires at least 3.3 kernel version" } Reproducible: Always
+ 13 Dec 2012; Sergey Popov <pinkbyte@gentoo.org> + libnetfilter_acct-1.0.0.ebuild, libnetfilter_acct-1.0.1.ebuild: + Add check for NETFILTER_NETLINK_ACCT kernel option, wrt bug #447092 Is there any reason you want 'die' in pkg_setup? Package builds fine with older kernels, so warning will be enough, i think...
(In reply to comment #1) > + 13 Dec 2012; Sergey Popov <pinkbyte@gentoo.org> > + libnetfilter_acct-1.0.0.ebuild, libnetfilter_acct-1.0.1.ebuild: > + Add check for NETFILTER_NETLINK_ACCT kernel option, wrt bug #447092 > > Is there any reason you want 'die' in pkg_setup? Package builds fine with > older kernels, so warning will be enough, i think... If it builds fine then it's completely ok for me to print a warning instead of `die`. I merely copied this code snippet from libnetfilter_conntrack ebuild and I'd never tested libnetfilter_acct build process on kernels prior to 3.3. Should've done that, but it seems you've already tested it:) So, it's up to you whether change that `die` or not. Should I try and build it on smth like 3.2?
> (In reply to comment #2) > Should I try and build it on smth like 3.2? No, you do not have to - i have already done it with 3.0.35, so warning will be enough. + 13 Dec 2012; Sergey Popov <pinkbyte@gentoo.org> + libnetfilter_acct-1.0.0.ebuild, libnetfilter_acct-1.0.1.ebuild: + Add check for kernel version, wrt bug #447092
Thank you very much! This one went through really quick:)
(In reply to comment #2) > > Is there any reason you want 'die' in pkg_setup? Package builds fine with > > older kernels, so warning will be enough, i think... > > If it builds fine then it's completely ok for me to print a warning instead > of `die`. I merely copied this code snippet from libnetfilter_conntrack > ebuild and I'd never tested libnetfilter_acct build process on kernels prior > to 3.3. Should've done that, but it seems you've already tested it:) Maybe we should also remove that die() in libnetfilter_conntrack, then? And check other similar ebuilds, too?
(In reply to comment #5) > Maybe we should also remove that die() in libnetfilter_conntrack, then? And > check other similar ebuilds, too? Well, i though that a policy about it is clear - if it builds with old kernel or headers - then we should only warn user(cause it will not work though)