On Wed, Nov 13, 2013 at 12:11:27PM +0100, Adam Borowski wrote: > So you save a small number of inodes, and get problems if the filesystem's > layout is unconventional. Such savings don't seem to be worth the trouble > to me.
I was questioning the existence of said trouble. I still do that. If the number of affected users is small, then ignoring those few is a sensible thing to do. For instance the publican package saved 3/4 of its binary package size. If this were a problem, then maybe we should have seen a bug report. > You don't know what directory resides on what filesystem. While splitting > up /usr tends to be trouble, it is not unusual. For example Maemo has: > > /opt/pymaemo/usr/lib/python2.5 on /usr/lib/python2.5 type bind (bind,rbind) > /opt/pymaemo/usr/share/pyshared on /usr/share/pyshared type bind (bind,rbind) > /opt/pymaemo/usr/lib/pyshared on /usr/lib/pyshared type bind (bind,rbind) > /opt/pymaemo/usr/share/python-support on /usr/share/python-support type bind > (bind,rbind) > /opt/pymaemo/usr/lib/python-support on /usr/lib/python-support type bind > (bind,rbind) These cases would not be problematic for my proposal. I was suggesting to explicitly allow cross-directory hard links within any of these locations, but not crossing each other. > So I'd keep this requirement as is. There is no requirement besides conffiles not being hard links. Having hard links that cross /usr/share/pyshared and /usr/lib/pyshared is not a policy violation. Hence I was suggesting to clarify the policy on this aspect. Helmut -- To UNSUBSCRIBE, email to debian-devel-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/20131115101619.ga2...@alf.mars
