martin f krafft <[email protected]> writes: > tags 563727 patch > thanks
> Indeed: > diff -u /tmp/console-screen.sh /etc/init.d/console-screen.sh > --- /tmp/console-screen.sh 2010-01-05 11:42:38.000000000 +1300 > +++ /etc/init.d/console-screen.sh 2010-01-05 11:42:58.000000000 +1300 > @@ -82,11 +82,10 @@ > CONSOLE_TYPE=`fgconsole 2>/dev/null` || return 0 > > if [ ! $CONSOLE_TYPE = "serial" ] ; then > - readlink /proc/self/fd/0 | grep -q -e /dev/vc -e '/dev/tty[^p]' -e > /dev/console > - if [ $? -eq 0 ] ; then > - VT="yes" > - reset_vga_palette > - fi > + if readlink /proc/self/fd/0 | grep -q -e /dev/vc -e '/dev/tty[^p]' -e > /dev/console; then > + VT="yes" > + reset_vga_palette > + fi > fi > > [ $VT = "no" ] && return 0 I believe change in the previous version to add set -e should also be undone. Init scripts should in general not use set -e. I opened a bug on Policy a while back to make that clearer. See, for instance, the comment in /etc/init.d/skeleton. Many common operations that one wants to do in an init script, such as starting a daemon and checking its return status, or things like the above are not set -e clean. Neither are the LSB functions to output status of init scripts starting, nor is there really any reason to go to the effort of making them so. For init scripts, explicitly checking the return status where it matters is more robust. set -e often results in strange and difficult-to-debug failures when used in init scripts. -- Russ Allbery ([email protected]) <http://www.eyrie.org/~eagle/> -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected]
