> Michael Tautschnig wrote: >> The attached patch is supposed to fix this, but should be carefully reviewed. >> > > Most of them seem reasonable. > But I am not sure with this: > > --- start --- > Index: examples/simple/scripts/GRUB/10-setup > =================================================================== > --- examples/simple/scripts/GRUB/10-setup (revision 4793) > +++ examples/simple/scripts/GRUB/10-setup (working copy) > @@ -2,7 +2,7 @@ > > error=0 ; trap "error=$((error|1))" ERR > > -fcopy -Uv boot/grub/menu.lst > +fcopy -v /boot/grub/menu.lst > > exit $error > --- end --- > > Hmm, I gues sthne idea oif using -U here is, that in softupdate mode the > scripts that shpould fill in some variables intoi the template grub file, > cannot work as it doesn't know about the boot poartition of the system. And > so you get a broken menu.lst. > > I still fell into that problem because I used fcopy recursively, and > temporarily came up with having useable menu.lst files for all host in > files. > > If you copy the file here even in softupdate, you have to make sure it > get's filled in correctly! > >
This is actually strange, I think. AFAIK disk_var.sh is also available during softupdates (it is kept in the log directory, I think), so updating grub menu.lst should be fine. But maybe there is some bug in the postinst, or other weird things going on. In any case: Properly filled templates is indeed an issue and the FAI simple examples should be made very clean in this direction. Thanks for your careful review, Henning! Best, Michael
pgpMyIsx0JP4C.pgp
Description: PGP signature
