Dan McGee wrote:
On Sat, Jul 25, 2009 at 4:30 PM, Nagy Gabor<ngaba@bibl.u-szeged.hu> wrote:
Maybe instead of trying to remove the directory, it should check all existing packages to see if there is one owning this directory. But this will slow down the -R operation, for each directory inside a package.
-1. This would induce a memory usage boost for fixing a minor issue. Pacman would load _all_ local filelists (the "big" part of local database, in addition, loading is slow with empty disk-cache) in even many simple "-S one_package" operations. I have seen dont.remove.this.file many times so far (for example, in "empty" dirs of some archives), that is much easier in this case, too.
One catch here is the original bug dealt with /var/empty. Judging by the name of that directory, putting anything at all in it would defeat the point of it, even if it is a bogus/hidden file.
Yes, anyway, /var/empty is not defined by FHS. And the manual of sshd says: /var/empty chroot(2) directory used by sshd during privilege separation in the pre-authentication phase. The directory should not contain any files and must be owned by root and not group or world-writable. So I think that this is a problem with other programs that uses /var/empty and doing the workaround putting a hidden file (there are only two packages in all repos that use this dir core/openssh and community/openntpd)
The best thing to encourage here is correct packaging, and use of the !emptydirs option in the PKGBUILD to ensure removing your package doesn't remove some system directory that probably didn't need to be in your package in the first place.
ok -- Gerardo Exequiel Pozzi ( djgera ) http://www.djgera.com.ar KeyID: 0x1B8C330D Key fingerprint = 0CAA D5D4 CD85 4434 A219 76ED 39AB 221B 1B8C 330D