Hi On Samstag, 19. Juli 2008, Robert Millan wrote: > On Wed, Jul 16, 2008 at 12:56:29PM +0200, david manyé wrote: > > d131:~# grub-install /dev/hda5 > > Searching for GRUB installation directory ... found: /boot/grub > > /usr/sbin/grub-install: line 384: 4981 Excepció de coma flotant > > > > [...] > > > > i've tried unsuccessfully the default lenny version 0.97-41 and 0.97-30 > > but older versions than -30 work perfectly: 0.97-27, 0.97-27etch1, > > 0.97-28 and 0.97-29. i've got -28, -29 and -30 versions from > > snapshots.debian.net. > > > > all the system is installed in /dev/hda5 with ext2 filesystem. > > Are you sure -29 worked but -30 didn't? This would indicate > ext3_256byte_inode.diff introduced this regression.
This would indeed hint at ext3_256byte_inode.diff, as it is the only change affecting the actual source code. Looking at the diff again, the only potential reason might be s_inode_size being 0, although I have no idea if that is possible with ext2 partitions or if this indicates unrelated filesystem corruption. > Which version of e2fsprogs did you use to create that ext2fs? (and with > which flags) Personally I didn't observe any breakage (admittedly, 99% of my tests used ext3 and only few ext2, but involved almost daily install tests) with the, at the time, current version of e2fsprogs in sid since february, and given that this patch is present in Fedora 9, Ubuntu 8.04 Hardy Heron, OpenSuSE 11.0 and Mandriva 2008.01 spring, I would have expected expect more reports about potential breakage, even though I assume that a fair share of those installs either preferred ext3 over ext2 or reformatted the partition while installing. Neither installing etch 4.0r3, nor sarge 3.1r5 in virtualbox-ose onto a zeroed disk, partitioning it as ext2, followed by dist-upgrading to current sid and reinstalling grub 0.97-41 (grub-install --recheck --no-floppy /dev/hda) revealed any problems in my tests. Regards Stefan Lippers-Hollmann
signature.asc
Description: This is a digitally signed message part.