On Sat, Mar 04, 2006 at 12:40:46AM -0800, Steve Langasek wrote: > severity 355108 normal > thanks > > On Fri, Mar 03, 2006 at 11:42:52AM +0100, Sven Luther wrote: > > > Well, it seems that in some rare case (two example below) either libparted > > will issue an error or not know about a given partition table. > > > In these cases, the installer will think there is no partition table, and > > propose the creation of a new one, erasing the old partition table, and thus > > causing the total loss of all data previously found on the harddisk. > > If it *proposes* the creation of a new partition table, then obviously this > is still under control of the user. Overwriting data when the user tells > you to is not "data loss" for the purpose of bug severities.
I disagree, less technical savy users could be surprised by this, and not realize they are doing something wrong, but you are the RM, and will be responsible for the consequences :) > > A proposal to fix this would be to create a partman-kernel or whatever, > > which would be called if libparted fails, and then go looking at > > /proc/partitions, to see if there are partitions there, and propose a > > read-only partition setup, which would allow the user to install in > > existing partition, and more importanly don't let him to silently > > over-write existing partition tables. > > I think this is an interesting idea, at least. However, consider that one > of the reasons why parted will reject a current partition table as invalid > is due to overlapping partitions... Well, sure, but the idea is to look at /proc/partitions, and we have to go from the assumptions that the kernels knows how to do things right, since it is him who will be mounting the partitions anyway. At the very least, partman should export the libparted exceptions to the user in these case, or have a exception handling wrapper or whatever, which will not bring this 'let's overwrite all data' case for every minor libparted exception. Let's hope that Xavier's work on a graphical partitioning tool will bring some responses to these problems. Friendly, Sven Luther > > -- > Steve Langasek Give me a lever long enough and a Free OS > Debian Developer to set it on, and I can move the world. > [EMAIL PROTECTED] http://www.debian.org/ -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
