On Sat, 2018-05-05 at 09:44 +0900, Masanori Goto wrote: > Recently I encountered this amd64 kernel issue on my i386 architecture > machine. I think this is useful to be mentioned. Why don't we want to > apply this proposed text? > > > You're quite right, this should have been documented. It might be > > worth mentioning linux-headers-amd64 as well. Also, module-assistant > > doesn't support foreign architectures but DKMS is fine. > > +1 to linux-headers-amd64. One tricky thing is, this is not officially > supported so module-assistant doesn't work, so that should be clearly > mentioned. I also had some issues to install some kernel related > packages (e.g. virtualbox-dkms) so I'm not sure DKMS is fine or not - > I rather think DMKS also doesn't work but I'd like to hear your > feedback. Should we also mention that? [...]
Some module packages do their own architecture detection and may fail when the primary architecture is i386 and the kernel flavour is amd64. This can happen regardless of what architecture the kernel package is labelled as! DKMS itself doesn't seem to have a problem with it. At some point I tested the available DKMS-based packages with this configuration and reported bugs, but there may have been regressions since then. Where module-assistant goes wrong is that it always sets the package architecture to be the primary architecture. I looked at fixing this, but there's no one place to fix it as every dependent package duplicates logic and templates. DKMS doesn't normally build or install packages, so it doesn't have this problem. Ben. -- Ben Hutchings If more than one person is responsible for a bug, no one is at fault.
signature.asc
Description: This is a digitally signed message part
