Is it normal that my running kernel is kernel-preempt instead of kernel-default?

[arvidjaar]

solver doesn't recognise that the requirement satisfied by a package (kernel-preempt-devel) can be solved by another package (kernel-devel)

kernel-devel and kernel-flavor-devel are two different packages that complement each other. kernel-devel does not replace kernel-flavor-devel, nor does it satisfy requirement for kernel-flavor​-devel.

[jetojedno]

kernel-devel and kernel-flavor-devel are two different packages that complement each other. kernel-devel does not replace kernel-flavor-devel, nor does it satisfy requirement for kernel-flavor​-devel.

I suspect they both provide the same capability which is required in another package, and the solver "believes" the capability is going away when the kernel-flavor-devel package is removed rather than recognising that it's still being provided by an alternative package (kernel-devel).

But I could be wrong: I've not inspected the code - I'm just surmising from the solver's behaviour and several years of debugging systems' unintended or unwanted behaviours.

p.s. this behaviour was from using yast2 sw_single rather than ziper remove, so I may be assigning blame to the wrong piece of code.

[OrsoBruno]

These two lines from the attached log are interesting:

2020-12-28 23:13:56|install|kernel-preempt-devel|5.3.18-lp152.57.1|x86_64||repo-update|84e0925335602d9ef693e175378ba58835bdbcf40360cfe01c6dba56408573d0|
2020-12-28 23:13:59|install|kernel-default-devel|5.3.18-lp152.57.1|x86_64||repo-update|da9daf0e56d2211a3fc4b402aff67ee3aeac1e8342e1aadf0da0dccddcb95962|

AFAIK kernel-preempt-devel and kernel-default-devel should provide the same capabilities (while kernel-devel is a different story).
But those packages are only involved in building kernels or kernel modules from source (you have kernel-source installed...) but don't provide "running kernel" code.
If, as stated in the title, your running kernel is -preempt, you should have also a kernel-preempt-5.3.18xxx installed but there is no such evidence in the log you posted; moreover, without VirtualBox it is not clear what triggered the install of such kernel-preempt-5.3.18xxx (well _IF_ it is installed at all, kernel-preempt-devel should do no harm by itself...)

[OrsoBruno]

One more thought, there is no kernel-preempt in Leap 15.1, so is it possible that during a "dup" from 15.1 to 15.2 the solver cannot decide which upgrade from kernel-default-4.12x to choose and then installs _both_ kernel-default-5.3 _and_ kernel-preempt-5.3 ??
If that is the case, we may see a flood of such cases in the Forums now that 15.1 is approaching EOL...

[jetojedno]

we may see a flood of such cases in the Forums now that 15.1 is approaching EOL...

Thanks @OrsoBruno. Yes, I made a mistake in my post - it was kernel-default-devel I meant not kernel-devel, which is also installed.

The dependency is dkms requires kernel-syms, which requires kernel-devel and kernel-default-devel or kernel-preempt-devel (I assume there's some kind of one-of-many logic here I can't see using rpm -q). kernel-devel requires kernel-macros provided by kernel-default-devel or kernel-preempt-devel.

These two defects (one in the solver when removing kernel-preempt-devel and one when doing the dup) could both be the same defect: the solver is either doing things out of order (i.e. solving the kernel-syms > kernel-*-devel requirement before kernel-devel > kernel-*-devel pulls in the right flavour); or it's removing the provided capability from it's list without checking if that capability is provided by another package which is or is planned to be installed..

Yes, I'd expect quite a few of these, although most people won't notice as it doesn't affect functionality.

Statistics on the installation volumes of kernel-devel, kernel-default, kernel-default-devel, kernel-preempt, kernel-preempt-devel downloads or installs (if available) would tell you how widespread the problem is. Specifically the change in the ratios between kernel-devel & kernel-*-devel and kernel-* & kernel-*-devel are indicators.

[arvidjaar]

dkms requires kernel-syms, which requires kernel-devel and kernel-default-devel or kernel-preempt-devel (I assume there's some kind of one-of-many logic here I can't see using rpm -q).

It requires both.

bor@10:~> zypper info --requires kernel-syms
Information for package kernel-syms:
------------------------------------
Repository     : Main Update Repository
Name           : kernel-syms
Version        : 5.3.18-lp152.60.1
...
Requires       : [4]
    pesign-obs-integration
    kernel-devel = 5.3.18-lp152.60
    kernel-default-devel = 5.3.18-lp152.60
    kernel-preempt-devel = 5.3.18-lp152.60


bor@10:~>

DKMS is the only package (in standard repositories) that requires kernel-syms and kernel-syms is not really needed at all - to compile external modules for your kernel you only need kernel-flavor-devel which automatically pulls in other required dependencies (and kernel-syms is not among them). I guess it is legacy package which pulls in all flavors to facilitate building kernel modules for every possible kernel variant. I do not have kernel-syms at all and have no problems compiling modules (I do not use DKMS).

In any case, all of this has absolutely nothing to do with the title of your post, because installing kernel-preempt-devel certainly does not make your system run kernel-preempt and in the whole thread there is no single evidence that you are actually running it.