Hi Richard,

Thank you for the explanations!

On 29.07.23 at 00:28, Richard Purdie wrote:
On Fri, 2023-07-28 at 20:41 +0200, Michael Opdenacker via
lists.yoctoproject.org wrote:
Greetings,

As far as I understand, the list of distros that the Yocto Project
supports (or should claim it supports) is the intersection of the list
of sanity tested distros (SANITY_TESTED_DISTROS in
meta-poky/conf/distro/poky.conf) and the distribution versions that are
still publicly supported by their vendors (as checked on Wikipedia).
There's no point in claiming to support a distribution for which package
updates are no longer available.

Currently for master, this intersection is only:

   * Ubuntu 20.04 (LTS)
   * Ubuntu 22.04 (LTS)
   * Fedora 37
   * Debian GNU/Linux 11.x (Bullseye)
   * OpenSUSE Leap 15.4

Here are a few obvious absents here:

   * AlmaLinux: version 8.7 and 9.1 are still in SANITY_TESTED_DISTROS
     but they are obsolete. I see an AlmaLinux9 worker on
     https://autobuilder.yoctoproject.org/typhoon/#/workers, but I don't
     know whether it's 9.1 or 9.2 (not obsolete unlike 9.1).
   * Fedora 38: important as Fedora 37 ends in November. I see a Fedora38
     worker, but there are still failed jobs, even though the majority
     are successful. When is a distro version ready to be added to
     SANITY_TESTED_DISTROS?.
   * Debian 12: we also have such a worker, with a mix of successful and
     failed builds. Is this version eligible too?
   * OpenSuse Leap 15.5, as OpenSuse Leap 15.4 expires in November too. I
     see a worker, but with only failed builds.

I'd say that the most urgent one is AlmaLinux 9.2 as otherwise we don't
support AlmaLinux any more. The next ones for the end of the year are
Fedora 38 and OpenSuse Leap 15.5.

Would you have any thoughts?

I was about to send a documentation update to declare that master (and
probably Mickledore and Kirkstone) only support 5 different distribution
versions, but a longer list would look better...
We keep seeming to discuss this but it is really quite simple, we
document what we test on.

We are currently testing on alma8 and alma9. I don't know which version
of alma9 but I'd suspect the most recent one and we can check. We can
list those until the workers are retired.


At least SANITY_TESTED_DISTROS says its AlmaLinux 9.1. I'm not familiar with this distribution, but since 9.1 and 9.2 look different (9.2 being supported, 9.1 not, see https://mirrors.almalinux.org/isos.html), I think we should make the difference.

Do you agree that the documentation should only advertise as supported the intersection of SANITY_TESTED_DISTROS and the ones that are still supported by  their maintainers? If we agree on this and it turns out we neither test AlmaLinux 8.8 nor 9.2, it means that we currently don't support AlmaLinux any more.


With regard to the failed builds, the question is whether the failed
builds are due to the patches under test or some underlying issue with
those systems. They would not be in general use in the pool if they
weren't able to build master so it is fine to document all of those
there for master. I'm sure the failures are due to failures in patches
under test at this point.

The exclusion list for older releases can be and is different as some
workers are not enabled for some releases. You can see the mappings
here:

https://git.yoctoproject.org/yocto-autobuilder2/tree/config.py#n160

i.e. mickledore is matching master so far and kirkstone isn't using
fedora38 yet (for example).


Oops, I don't understand how you deduce this from this file. Mickledore isn't event listed in it.
Cheers,
Michael.

--
Michael Opdenacker, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com

-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.
View/Reply Online (#60694): https://lists.yoctoproject.org/g/yocto/message/60694
Mute This Topic: https://lists.yoctoproject.org/mt/100419636/21656
Group Owner: [email protected]
Unsubscribe: https://lists.yoctoproject.org/g/yocto/unsub 
[[email protected]]
-=-=-=-=-=-=-=-=-=-=-=-

Reply via email to