Control: retitile -1 dkms module FTBFS for amd64 kernel + i386 user land Control: severity -1 important
On Thu, Jan 31, 2019 at 5:09 AM Andreas Beckmann <a...@debian.org> wrote: > > On 2019-01-30 20:12, Petter Reinholdtsen wrote: > > > > [Andreas Beckmann] > >> zfs-dkms does not build a kernel module for e.g. the 4.19.0-2-686 i386 > >> kernel. This happened in a i386 chroot on a amd64 host, but this should > >> not really be a problem, since this setup works fine for other > >> packages. > > > > Are you sure this isn't intentional? It is strongly recommended not to > > use ZFS on 32 bit architectures, if I recall correctly. > > "intentional" would have been to fail with a clear error message during > the configure step (optionally recommending an override option), not > after successfully compiling tons of source files to fail in some crypto > code with a dubious error message: > #error Assembler code is only available for x86 and AMD64 systems > I didn't even try it on a non-x86 platform :-) > > This looks rather like completely "untested". > I think it's "untested" for a setup like 64bit kernel + 32bit user land, so that I'm downgrading the severity. These setups are tested to be working on x86 systems: 1) amd64 kernel + amd64 userland (recommended) 2) i686-pae kernel + i386 userland (not recommended and howls a scary debconf warning when being configured, but it works) Will take a look on the mixed options but it's definitely not the recommended way to go so far because the lack of testing. Also I think this could potentially affect x32 as well because they use an amd64 kernel. Cheers, Aron
