-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On 8/22/14, 10:30 AM, Jeff Mahoney wrote:
On Fri Aug 22 10:29:23 2014, Takashi Iwai wrote:
At Fri, 22 Aug 2014 09:33:41 -0400, Jeff Mahoney wrote:
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Hi all -
I'm doing the integration work for 3.17-rc1 and, as usual, there are a ton of drivers that are typically only used for embedded devices. Generally, I'd prefer to just disable them, but there are always a few people who want to run openSUSE on things like tiny Atoms.
Do we want to consider moving uncommon drivers into something like kernel-default-uncommon and save disk space on most every other system? Of course, then there's also the concept of kernel-default-base again, where it would probably contain typical file systems (xfs/ext4/btrfs) and drivers for devices most commonly offered by qemu-kvm. I understand it doesn't cover cases like device assignment, but those are more advanced use cases in which the full kernel-default package would be used.
The problem I'm looking to solve is "what would work well for several major use cases?" I'd define those as typical desktop/server/notebook hardware and KVM instances without device assignment. (Xen is a different conversation since dom0 and domU have different requirements WRT drivers.) The fallback for not fitting into one of the nice boxes is just installing another package which uses as much disk space as the old kernel package used to, so it's not a huge inconvenience.
OTOH, having multiple kernel sub-packages has been somewhat of a PITA WRT updates and dependencies, so I thought I'd solicit opinions.
I find it OK to split if the size matters. It's a bit painful but we know it works from the experience with SLE.
But, if we really split, how would the sizes of both packages be? Is kernel-common significantly large?
I don't actually know. I didn't want to invest the time into going through the driver list to create the split before there was interest.
Followup:
While this might end up being a good idea generally, the SPI part of
it is a false alarm.
This upstream commit auto-selected SPI:
e4462ffc1602d9df21c00a0381dca9080474e27a
Author: Antti Palosaari