Package: busybox, partman This bug is detected on sarge-netinst.iso for ia64 20040806 (RC1). The system is an HP rx2600 dual Itanium II. This error occurs on both a pristine iso and an iso with my new boot.img and an updated elilo-installer udeb installed.
The busybox shell running elilo-installer's postinst terminates pre-maturely, generating an incorrect error. The upper levels interpret this error as a failure of postinst rather than postinst catching the error and reporting its own (more exact) error. This is a bit of an unusual bug in that it is triggered by another regression bug in either partman or the kernel packaging for ia64. The following code fragment from elilo-installer.postinst from my working copy with debugging lines illustrates the problem: --- snip --- db_progress INFO elilo-installer/progress_running echo "about to elilo bootpart=$bootpart, rootfs=$rootfs" \ >> /tmp/elilo-installer chroot /target /usr/sbin/elilo --autoconf --boot $bootpart \ --root $rootfs --efiboot > /tmp/errmsg 2>&1 ERRCODE=$? echo "done elilo" >> /tmp/elilo-installer echo "error code=$ERRCODE." >> /tmp/elilo-installer --- snip --- The trigger of this event is the failure of partman or ??? to install fat-modules causing elilo to error return because it cannot mount the vfat f/s in $bootpart (/dev/sda1). The trace file /tmp/elilo-installer contains "about to ..." as its last line with the shell vars properly expanded, in this case, bootpart=/dev/sda1 and rootfs=/dev/sda2. The echo "done elilo" is never reached. I have tried a number of alternatives including "err=$(....)". The file /tmp/errmsg records the error message from elilo. In addition, the command run from the install shell returns $? of 0. The same command run from multi-user (bash) returns $? of 1. a snippit of install syslog follows with inline comments: Aug 9 22:58:05 main-menu[238]: DEBUG: resolver (libblkid1-udeb): mark, dependency from e2fsprogs-udeb Aug 9 22:58:05 main-menu[238]: DEBUG: resolver (e2fsprogs-udeb): mark, dependency from partman-basicfilesystems Aug 9 22:58:05 main-menu[238]: DEBUG: resolver (ext2-modules): search, dependency from partman-basicfilesystems Aug 9 22:58:05 main-menu[238]: DEBUG: resolver (kernel-image-2.4.26-itanium-smp-di): mark, dependency from partman-basicfilesystems Aug 9 22:58:05 main-menu[238]: DEBUG: resolver (fat-modules): package doesn't exist (ignored) --- this occurs earlier for fat, xfs, ext3, and reiserfs as well... --- the udebs appear to be on the netinst iso however... Aug 9 22:58:05 main-menu[238]: DEBUG: resolver (partman-basicfilesystems): mark, dependency from partman-target Aug 9 22:58:05 main-menu[238]: DEBUG: resolver (bugreporter-udeb): mark Aug 9 22:58:05 main-menu[238]: DEBUG: resolver (cdrom-checker): mark Aug 9 22:58:05 main-menu[238]: DEBUG: resolver (di-utils-shell): mark Aug 9 22:58:05 main-menu[238]: DEBUG: resolver (di-utils-reboot): mark Aug 9 22:58:05 main-menu[238]: DEBUG: Menu item 'elilo-installer' selected Aug 9 22:58:05 main-menu[238]: DEBUG: configure elilo-installer, status: 2 Aug 9 22:58:12 main-menu[238]: (process:8790): mount: fs type vfat not supported by kernel Aug 9 22:58:12 main-menu[238]: (process:8790): elilo: An error occured mounting /dev/sda1 -- This is where elilo dies. This run may not have had the redirect. Aug 9 22:58:12 main-menu[238]: WARNING **: Configuring 'elilo-installer' failed with error code 1 -- postinst should have handled this. Aug 9 22:58:12 main-menu[238]: WARNING **: Menu item 'elilo-installer' failed. Aug 9 22:58:58 main-menu[238]: INFO: Modifying debconf priority limit from 'high' to 'medium' There are additional issues caused by the module load failure(s) that render the RC1 iso unusable. I am currently investigating those as well. Regards, Jim -- ************* Jim Lieb Wild Open Source Inc. [EMAIL PROTECTED] Cell: 831.295.9317 Office: 831.421.0883 Fax: 831.421.0885 -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
