Re: [ANNOUNCE] 5.10.162-rt79

[Florian Bezdeka]

On Thu, 2023-02-23 at 15:33 +0100, Daniel Wagner wrote:
> Hi Florian,
> 
> On Thu, Jan 26, 2023 at 06:41:27PM +0100, Florian Bezdeka wrote:
> > From the CIP projects perspective we would like to improve the
> > situation.
> > 
> > From my perspective the following could be done:
> > 
> >   - Instead of (or in addition to) building and testing released -rt
> >     branches enable testing of -rt release candidates
> >   - Make sure the build results get back to the maintainers
> > 
> > I'm not sure if every -rt branch has a -rc branch. I'm not familiar
> > with the -rt release process yet.
> 
> The process so far is, that every stable maintainer updates his tree (merges the
> stable tree) and does a local build and local tests. Usually when merging latest
> upstream stable release there are no or little fallouts. When the maintainer is
> happy he does the release by pushing the changes to the usptream branch. The
> release candiates come only into play when there is something the maintainer is
> not sure how to handle or -rt patches are backported which need some more eyes
> to look on. That means Sebastian's approval :)

Ok, so there are no automated test for each release.

> 
> IIRC, I did give a presentation on the workflow some time ago...
> 
> https://lpc.events/event/4/contributions/293/attachments/237/416/maintaining-out-of-tree-patches-over-the-long-term.pdf
> https://www.youtube.com/watch?v=2ab4Knwlmo4

Thanks for the hint!

> 
> When we started with this process kernelci didn't build these branches but that
> is long time ago.

At least (some) stable release branches are being build now but it
seems that nobody is caring about the result. That's one of the things
I would like to address.

> 
> Personally, I don't mind doing an official -rc for every release and getting
> some additional builds and tests run by kernelci.

Ok, but that would be something that "all" -rt-stable maintainers
should agree on, no? At least to some degree to make the testing part
consistent.

> 
> Though just piping the results back is the easy thing, the time consuming task
> is to fix those problems. Do you plan to help out here?

Well, having the pipelines might be easy, but we don't have them yet, 
right? Reacting on the pipeline results would be the next step that
possibly could happen in cooperation with the CIP project.

But yes, generally speaking we want to help. But we don't have infinite
resources (as usual), so we'd like to know where most impact could be
achieved.

Are there any plans to build and test the stable -rt branches with the
help of automation, for example by using the kernelci infrastructure?
If no: Would that be something that the RT project is interested in?

I'm asking for kernelci because the CIP project is already using that.
They defined a clear set of supported architectures, kernel configs,
test suites and concrete hardware (to some degree). This allows them to
make sure each release candidate of each branch is in good shape before
doing the actual release.

Regards,
Florian

> 
> Thanks,
> Daniel