sparse arrays and a few more fixes

[anbe42]

properly support (or rather error out on) sparse module arrays
use cp --reflink=auto
revert dkms.service installation on Debian

[anbe42]

@scaronni: I have one more change in the Debian package: adding back documentation for the deprecated CLEAN and the new special case CLEAN=true. Should we upstream that as well? Better wording welcome.

[scaronni]

Hi @anbe42, this pull request looks great to me, let's merge it when you think you're ready, I would like to cut out a new release.

I've seen the patches for adding back the documentation for CLEAN and CLEAN=true. If it's ok for you, I would prefer to keep things as they are and not include them. The thing is deprecated, and that looks sort of like rolling it back.

I also made everyone switch to the new format here in NVIDIA for the dozen of DKMS packages, and would like to avoid having things lingering in various dkms.conf files forever.

Thanks.

[anbe42]

Hi @anbe42, this pull request looks great to me, let's merge it when you think you're ready, I would like to cut out a new release.

Perhaps you could make a 3.2.2 bugfix release with the changes we have accumulated, maybe temporarily reverting the incomplete per-module-conditionals commits. Then I'd have something for the next Debian stable point release and can bring the new per-module-conditionals feature in 3.3.0. And maybe I get around doing the module install manifest implementation in time, too.

[scaronni]

Sounds good. Let me do it. I've actually reworked also the entire per module conditionals in a separate tree.