On Fri, 2015-03-27 at 17:31 +0100, Atom2 wrote: > Am 27.03.15 um 10:28 schrieb Ian Campbell: > > Unfortunately this has unexpectedly made things worse. See: > > http://www.chiark.greenend.org.uk/~xensrcts/logs/36757/ > > > > The issue is that readlink -f returns the full absolute path, so given > > $ ls /boot/xen* > > xen -> xen-X.Y > > xen-X.Y > > Then: > > $ readlink /boot/xen > > xen-X.Y > > $ readlink -f /boot/xen > > /boot/xen-X.Y > > > > This breaks because on these systems /boot is a separate partition, so > > from the bootloaders point of view we need to be accessing /xen-X.Y > > not /boot/xen-X.Y. > > > > I don't know of an easy way to get the path of a file relative to the > > filesystem root which contains it. So I would propose to drop the second > > hunk and the second paragraph of the commit message and to make the > > first hunk simply "flaskpolicy=$flaskpolicy" (rather than unescaping the > > usages, this will make it easier to change in the future). > > > > I think that resubmission will end up being to the new colo at this > > point. > > > Hi Ian, > wouldn't the following snippet of code actually achieve what you want: > > mylink="$(readlink -f /boot/xen)" > relpath="${mylink#$(stat -c "%m" "$mylink")}"
Modulo the quoting of the shell-within-Perl we are dealing with here I think it might, yes. Thanks. > P.S. I know I still owe you the 64bit stat counter changes. > I haven't forgotten about it, but have not been able to dig into it yet. No worries. _______________________________________________ Xen-devel mailing list Xen-devel@lists.xen.org http://lists.xen.org/xen-devel
