Description of problem: hal-0.4.5 installed and running upgrade to the new hal-0.4.6 packages just released to updates-released tree for fedora core 3 and hald is not restarted. old packages: hal-devel-0.4.5-1.FC3.i386.rpm hal-0.4.5-1.FC3.i386.rpm hal-gnome-0.4.5-1.FC3.i386.rpm new packages: hal-devel-0.4.6-1.FC3.i386.rpm hal-0.4.6-1.FC3.i386.rpm hal-gnome-0.4.6-1.FC3.i386.rpm Could this be the result of a missing /sbin/ in the package scriptlets? Should that be /sbin/service haldaemon instead of just service haldaemon? How reproducible: Everytime i do the upgrade from 0.4.5 to 0.4.6 rpms. up2date,yum and rpm -Uvh Steps to Reproduce: 1. check to make sure hald is running and hal 0.4.5 rpm are installed 2. upgrade to hal 0.4.6 wait a minute 3. check to make sure hald is running and it isnt. Expected results: I would have thought hald would be restarted after package upgrade. Do you need any logging information?
It should definitely be /sbin/service instead of just service - I haven't seen this before but I have fixed the package; please try http://people.redhat.com/davidz/hal-cvs20050121/ and let me know if it fixes this bug. Thanks, David
i just updated from 0.4.6-1 to the 0.4.6.cvs20050121-1 set still saw the same behavior. hald was stopped but not restarted. Are you seeing similar behavior? Maybe my hald is flaking out for some other reason besides the service commands? Or maybe the lack of /sbin/ in the installed 0.4.6-1 scriplets affected the upgrade still. The /sbin was missing from both the preuninstall and the postuninstall in 0.4.6-1 so it might not work unless the installed AND the upgrade both have /sbin in the scripts. I did an rpm -Uvh --force of the 0.4.6.cvs20050121-1 set while they were the 0.4.6.cvs20050121-1 was installed and hald is running. Just to make sure the missing /sbin from the 0.4.6-1 scripts was the problem. Can you spin up 0.4.6.cvs20050121-2 package set? I want to see if upgrading from 0.4.6.cvs20050121-1 with the /sbin fix works if i upgrade to 0.4.6.cvs20050121-2 with the /sbin fix. -jef
Hi, I've respun and the packages are here http://people.redhat.com/davidz/hal-cvs20050121/2/ Cheers, David
okay just did the -2 update from -1 hald still ended up stopping. I'm going have to put hald in verbose mode i think and see if i can log whats going on later tonite when im home. -jef
Bah.. i dont think i can log this effectively. If i run: /usr/sbin/hald --daemon=no --verbose=yes >/tmp/hal-log.txt 2>&1 the upgrade to the -2 packages doesn't cause hald to shutdown. And it shouldn't since the service call isn't stopping hal since no pid file was created from a service start call. So i simply tried doing an /sbin/service haldaemon condrestart and it failed! I dont think the sleep 3 is a long enough wait for hald to cleanly stop. I replaced sleep 3 with sleep 10 still failed. sleep 20 however worked. Anything you need me to poke at now? I definitely think the problem is with the condrestart. The stop just takes +10 seconds to do a full service shutdown on my system and hence the ssrvice start doesn't actually happen. -jef
okay just double checked with a second system and it does not see the problem. The affected system is an smp machine while the system that works as expected is a single processor machine. Don't know if that matters. Tried to get a few people to reprodce it on irc.. and no one can. I'm willing to chalk this up to a busted system on my end, but it would be nice to understand why its taking hal this long to shutdown. Anything you can think of for me to poke at? -jef
hal-0.4.7-1-FC3 still fails for me. Running /usr/sbin/hald --daemon=no --verbose=yes segfaults. Single processor Athlon/KT133 chipset that was stable under RH9.
This should be fixed in the 0.5.0 series; please reopen this bug if it still fails