Related to Bug 137300 but I suspect not the same in cause (although the mentioned re-write of code may be the solution). Disk contained 5 partitions. XML from install log - sorry, was far too annoyed when this occurred to cut and paste the entire log so I have just this fragment. --- <partitions> <device devnode="/dev/hda" disklabel="msdos"> <partition format="False" mb="494" minor="1" mkfsopts="" mountopts="" mountpoint="" origminor="1" resized="False" type="unknown"/> <partition format="False" mb="40013" minor="2" mkfsopts="" mountopts="defaults" mountpoint="/" origminor="2" resized="False" type="unknown"/> <partition format="False" mb="39997" minor="3" mkfsopts="" mountopts="" mountpoint="" origminor="3" resized="False" type="ntfs"/> <partition format="False" mb="33934" minor="4" mkfsopts="" mountopts="" mountpoint="" origminor="4" resized="False" type="extended"/> <partition format="False" mb="33934" minor="5" mkfsopts="" mountopts="" mountpoint="" origminor="5" resized="False" type="ntfs"/> </device> </partitions> --- 1 was valid swap 2 was valid ext2fs, recently formatted, blank 3 was valid NTFS, mounted at /mnt/windows (stage and portage snapshot were here). This was a working W2K system. 4/5 also valid NTFS The install log file... --- GLI: July 31 2006 19:40:02 - Setting root password. GLI: July 31 2006 19:40:02 - Livecd root password set. GLI: July 31 2006 19:40:02 - Portmap started. GLI: July 31 2006 19:40:02 - Completed pre_install steps GLI: July 31 2006 19:59:36 - partitioning: Processing /dev/hda... GLI: July 31 2006 19:59:36 - Partition 1 has origminor 1...saving start sector 63 and end sector 1012094 GLI: July 31 2006 19:59:36 - Partition 2 has origminor 2...saving start sector 1012095 and end sector 82959659 GLI: July 31 2006 19:59:36 - Partition 3 has origminor 3...saving start sector 82959660 and end sector 164875094 GLI: July 31 2006 19:59:36 - Partition 4 has origminor 4...saving start sector 164875095 and end sector 234372284 GLI: July 31 2006 19:59:36 - Partition 5 has origminor 5...saving start sector 164875158 and end sector 234372284 GLI: July 31 2006 19:59:36 - Deleting old minor 1 to be recreated later GLI: July 31 2006 19:59:36 - Deleting old minor 2 to be recreated later GLI: July 31 2006 19:59:36 - Deleting old minor 3 to be recreated later GLI: July 31 2006 19:59:36 - Deleting old minor 5 to be recreated later GLI: July 31 2006 19:59:36 - Deleting extended partition with minor 4 GLI: July 31 2006 19:59:37 - Partitioning: Second pass... GLI: July 31 2006 19:59:37 - Partitioning: Third pass....creating partitions GLI: July 31 2006 19:59:37 - Drive has 234375000 sectors GLI: July 31 2006 19:59:37 - Partition 1 has 494MB GLI: July 31 2006 19:59:37 - Old start sector 63 retrieved GLI: July 31 2006 19:59:37 - Retrieved start sector is not the same as the calculated next start sector GLI: July 31 2006 19:59:37 - Old end sector 1012094 retrieved GLI: July 31 2006 19:59:37 - Adding primary partition 1 from 63 to 1012094 GLI: July 31 2006 19:59:37 - This is a bad thing. An exception occured outside of the normal install errors. The error was: ''unknown'' GLI: July 31 2006 19:59:37 - Traceback (most recent call last): GLI: July 31 2006 19:59:37 - File "/opt/installer/GLIClientController.py", line 165, in run self._install_steps[self._install_step]['function']() GLI: July 31 2006 19:59:37 - File "/opt/installer/templates/x86ArchitectureTemplate.py", line 336, in partition self._add_partition(parted_disk, start, end, "primary", newpart['type']) GLI: July 31 2006 19:59:37 - File "/opt/installer/templates/x86ArchitectureTemplate.py", line 69, in _add_partition if fs: fstype = fsTypes[fs] GLI: July 31 2006 19:59:37 - KeyError: 'unknown' --- After installer failed the partition table was left blank, I had to recreate it by hand.
This won't happen with the newly rolled 0.4 release, which will be part of the 2006.1 release. This issue was fixed many months ago in CVS. The installer will now bail before touching the disk at all if there are any partition types that it doesn't know how to handle, which is what happened here.
Cheers, only realised after hitting the commit button that I'd omitted to include resolved bugs when I searched to see if this had been raised before.