I cannot run /sbin/chkconfig on an EV5 system (AlphaServer 1000A 5/300) - all I get is "Illegal instruction". I recompiled it and it works fine, so it may be that it was compiled for EV6 only. This caused the install to leave the /etc/rc.d/rc?.d directories empty as well.
I don't *think* it's compiled wrong. If you rebuild it with -mcpu=ev5, does it work?
Yes, it works if I recompile it. All I did to recompile it was "rpm -i chkconfig-1.2.16-1.src.rpm" and "rpm -bc chkconfig.spec" and run the resulting binary. It looks like it was compiled for EV5: $ rpm -qvp --qf '%{optflags}\n' chkconfig-1.2.16-1.alpha.rpm -O2 -mieee -mcpu=ev5 $ I have several programs that don't work (I can't even boot into the normal multi-user runlevel). I guess it is possible that something is screwy on my system; this is an older Alpha that I've cobbled together from parts from several. I do have a hard drive with a bad spot (but I'm using RAID5 so it _shouldn't_ be a problem except that of course the partition with the bad spot is not used and that partition of the RAID is running degraded). I think my RAM is okay; is there any kind of memory tester (like "memtest86" on x86)? I've got Digital Unix 5.1 CDs sitting on my desk; I guess I could try installing it and see what it sees. I'm willing to try anything you can suggest.
I just did a fresh install after removing the questionable disk. I didn't do anything special; I did a text install on a serial console, did a "server" install, letting the installer partition the disks. I booted once with "init=/bin/sh" so I could turn off console gettys, install getty_ps (I think this should be installed by default) and set up a getty on ttyS0. I also fixed the aboot.conf (I had to add "console=ttyS0" and fix the kernel path: the installer put "0:1/vmlinux...." in there, but the "1/" should just be "/"; I'll bugzilla that). Here is the tail of the boot and a login attempt: INIT: Entering runlevel: 3 egrep: /proc/stat: No such file or directory /etc/rc3.d/S99local: [: : integer expression expected /etc/rc3.d/S99local: /etc/issue: Read-only file system /etc/rc3.d/S99local: /etc/issue: Read-only file system /etc/rc3.d/S99local: /etc/issue: Read-only file system cp: cannot stat `/etc/issue': No such file or directory /etc/rc3.d/S99local: /etc/issue: Read-only file system (none) login: root Login incorrect login: It has the same problems as before: chkconfig didn't work so /etc/rc3.d is empty except for S99local. Then there are other programs not working either so the whole thing is screwed up. I wouldn't expect to get the exact same problem from flakey RAM or bad disk. I thought perhaps I had a corrupted CD, but I just booted up with "init=/bin/sh", brought up the network with "ifup eth0" (which worked okay), and FTPed the chkconfig package from ftp.beta.redhat.com. No difference, except I noticed that passive FTP (which apparently is the default now in "ftp") did not work either. I don't know if I should bugzilla that or not, since I don't know if the problem is just me or what.
Are both the rebuilt and original source RPMs built with the same optflags if you query them?
Now that I look at it, my recompiled chkconfig does NOT have "-mcpu=ev5", while the distributed version does. However, I just forced it to use "-mcpu=ev5" and recompiled (so the flags match the distributed version) and my recompiled version still works fine. I just found that the other main package I'm having this trouble with is initscripts. I just recompiled and reinstalled it and chkconfig (and then re-ran chkconfig to add all the init scripts), and now the system boots okay into multiuser mode (I'm able to log in remotely via ssh now).
If you try the chkconfig-1.2.17-1 RPM from people.redhat.com/notting/chkconfig/ does that work?
Oh, never mind. This is the RPM problem. Everything that linked in popt statically (chkconfig and initlog) would die horribly. RPM got fixed, so this should be fixed with the new chkconfig build.
verified fix in internal testing ...