I filed a (wishlist) bug against sysklogd as when I install it, I do: dpkg -i sysklogd*.deb
>From a directory on a Zip drive. Later on I was puzzled why I couldn't "umount" the Zip drive - it was because sysklogd had been run from there. This doesn't particularly bug me, but what do other people think - should all postinst scripts "cd" somewhere safe before running /etc/init.d/...? I've attached the bug report and Joey's reply below. ---cut-here--- [EMAIL PROTECTED]: Re: Bug#17278: sysklogd: restart from root directory] On Mon, Jan 19, 1998 at 05:33:10PM +0000, Adrian Bridgett wrote: > Package: sysklogd > Version: 1.3-21 > Priority: wishlist > > I often install sysklogd from a zip drive like this: > > cd /zip > sudo dpkg -i sysklogd* > cd > > However afterwards I cannot umount the /zip drive as sysklogd is running. I > have to do "sudo /etc/init.d sysklogd restart" when $PWD is not in /zip and > then I can umount the drive. > > Perhaps you could do a "cd /" before restarting sysklogd, or maybe since I > am not using dselect this isn't supported? Only very few packages do this - to be honest, I only found msqld and roxen doing this. Roxen can only be started from a well-defined directory and msqld? I don't know why there is a "cd /tmp" when _stopping_ the program. I'm closing this bugreport. If you still want to change this you should go to debian-policy and start a proposal that every /etc/init.d scrips that support start/stop/restart etc. must have a defined "cd /some/whe/re" before starting the daemon. /some/whe/re should be /tmp if the package doesn't depend on a particular start path. ---cut-here--- email: [EMAIL PROTECTED] | Debian Linux - www.debian.org http://www.poboxes.com/adrian.bridgett | Because bloated, unstable PGP key available on public key servers | operating systems are from MS
