Re: [RESEND PATCH] s390: fix dead defaults for S390_MODULES_SANITY_TEST and S390_UNWIND_SELFTEST

[Alexander Gordeev]

On Wed, May 06, 2026 at 05:51:19PM +0100, Julian Braha wrote:
> On 5/6/26 13:37, Alexander Gordeev wrote:
> > In the end of the day I see no impact on the end result (aka the
> > generated config) - please correct me if I am wrong.
> 
> Yes, there is actually an impact on the end result / the generated
> config:

That is what is currently generated (likely because of "depends on KUNIT"):

CONFIG_S390_UNWIND_SELFTEST=m
CONFIG_S390_MODULES_SANITY_TEST=m
# CONFIG_KUNIT_ALL_TESTS is not set

> When a user of the kernel build system (e.g. 'make menuconfig') enables
> the config option KUNIT_ALL_TESTS, the user expects *all of the tests*
> to be enabled. Which is what commit
> 25d36a85c61b ("s390/test_unwind: convert to KUnit")
> was attempting to do with these tests, but it appears they made a
> copy-paste mistake. This is the commit referenced in the Fixes tag.
> 
> The point is, these tests were intended to be included with
> KUNIT_ALL_TESTS, but aren't. This is unintended behavior, and the
> definition of a bug in the kernel configuration spec.

I could not recreate it. Could you please provide the exact steps?

> That being said, I do agree with you that this is pretty low in
> severity, as far as bugs go. It is kunit tests we're talking about,
> after all. I'll check with Greg to see if he thinks this is worthy of
> a Fixes tag. Another, similar patch I sent out includes the Fixes tag,
> and has not received any objections from its 3 reviewers and testers
> so far:
> https://lore.kernel.org/all/20260405161545.161006-1-julianbraha@xxxxxxxxx/

Well, this is not about severity, it is about whether an actual problem
did exist and was fixed. The patch itself is fine 

> > I think once the tool is officially included you could refer to it
> > in the commit message,
> 
> I will also check on the best practices for tool-assisted development.
> In the past, I have received encouragement for acknowledging the tool:
> https://lore.kernel.org/all/bfdfdb05-77e2-455d-b68d-9da3fd9d1c0d@lucifer.local/
> and the official kernel docs require an acknowledgement for AI tool
> assistance (although this tool is not AI and doesn't generate code...)

I do appreciate your effort to bring a new tool! But unless it is part of
the toolchain it shold not be referred to in a commit message as far as
I am concerned. IOW, once it is in the tree, it is fine to mention it:

https://lore.kernel.org/all/20260405122749.4990dcb538d457769a3276e0@xxxxxxxxxxxxxxxxxxxx/
officially 

> - Julian Braha

Thanks!