coreutils's rm isn't able to recoursively remove directories (rm -R), tells "Function not implemented". That causes many ebuilds to fail. I'm running an Gentoo chroot on an RHEL (2.6.9-42.EL) instance on S390 mainframe. This might well be an upstream bug, as coreutils changed it's directory traversal to use procfs when it thinks certain functions (fchdir(), etc) were not available. Using /proc obviously breaks chroot's. http://thread.gmane.org/gmane.comp.lib.gnulib.bugs/6839 http://lists.gnu.org/archive/html/bug-coreutils/2008-08/msg00149.html https://bugs.launchpad.net/ubuntu/hardy/+source/coreutils/+bug/192239 http://bugs.gentoo.org/show_bug.cgi?id=224483 Reproducible: Always
Which version of coreutils do you use (`emerge -pv coreutils`)? Michael
Can you use `mount --bind /proc /<chroot>/proc` in the outer system as a work-around? ps, wouldn't work w/o proc either.
(In reply to comment #1) > Which version of coreutils do you use (`emerge -pv coreutils`)? Michael > [ebuild R ] sys-apps/coreutils-8.5
(In reply to comment #2) > Can you use `mount --bind /proc /<chroot>/proc` in the outer system as a > work-around? ps, wouldn't work w/o proc either. No, that makes it even worse: it tries to use /proc/<pid>/fd/*, which obviously gives wrong results in chroot environments.
running Gentoo without /proc mounted is not supported
(In reply to comment #5) > running Gentoo without /proc mounted is not supported With /proc it doesn't work either. In fact, it's even worse then. So, you have to options now: a) reopen the bug and accept it as valid b) leave it and so confess that you don't care about critical bugs in base packages. cu
you have provided 0 details as to how it breaks when /proc is not mounted. you've only said "Using /proc obviously breaks chroot's" which is ridiculously incorrect. no details -> bug still invalid.
(In reply to comment #7) > you have provided 0 details as to how it breaks when /proc is not mounted. > you've only said "Using /proc obviously breaks chroot's" which is ridiculously > incorrect. no details -> bug still invalid. Perhaps you should read postings (including references) before judging on anything. coreutils tries to traverse the tree by taking the cwd and fds from /proc. This breaks chroot environment, as procfs has no special filename masking for chroots.
all the bugs you refer to are out of date and do not apply to coreutils-8.5, nor are they necessarily ones we care about. if you're using such an old kernel that the *at funcs are not available, and you refuse to mount /proc, then it isnt our problem. mount /proc or upgrade your kernel.
