Package: udev Version: 0.056-3 Followup-For: Bug #395889 Hi,
> on a 2.6.18 powerpc/pegasos system, which is mostly etch with sid > kernels, i > tested both proposed fixes : > > if ! [ -r /proc/1/root -a /proc/1/root/ -ef /proc/self/root/ ] > > returned FALSE in both cases (Well, i think i forgot the !, so it > probably did return TRUE in both cases). So, this solution is not > working. You're right that both tests return exactly the opposite: ! [ -r /proc/1/root -a /proc/1/root/ -ef /proc/self/root/ ] [ "$(stat -c "%d/%i" /)" = "$(stat -Lc "%d/%i" /proc/1/root)" ] And that's because of the '!'. I put the '!' in my patch because it fitted there: We want to test if we are in a chroot, not if we are outside. As Mark pointed out, my test will only work as root. But the stat(1) test has the same problem (I've tested this), since it uses the same information. Oh, and by the way, mind the trailing slashes in the second and third path name, they're important. Both tests won't work if the chroot is bind-mounted to the old root. I'm not sure if this is a problem. Anyway the old test did work in that case. > Given the critic Christoph had (warnings on pre-2.6.18, altough we hardly care > for this in etch, which will have 2.6.18), mayeb something like this : > > if [ "`uname -r`" -ge 2.6.18 ] && > [ "$(stat -c "%d/%i" /)" = "$(stat -Lc "%d/%i" /proc/1/root)" ]; then > # WE ARE IN THE REAL SYSTEM > else > # WE ARE IN THE CHROOT > fi If you want to go with the stat(1) test, another way to avoid the warnings would be [ "$(stat -c "%d/%i" /)" = "$(stat -Lc "%d/%i" /proc/1/root)" ] &>/dev/null Regards, Christoph -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
