openSUSE Release Engineering minutes February 23

All meeting minutes can be found here:
https://etherpad.opensuse.org/p/ReleaseEngineering-meeting
Meeting is hosted here
https://meet.opensuse.org/ReleaseEngineeringMeeting

## Attendees
guillaumeg, DimStar, ddemaio, Sarah, DocB, lkocman,deneb_alpha, m4u,
Gerald,maxlin


## Leap

pkglistgen is failing on network issues. Issue was escalated to IT
Infra, 
WARNING SUSE-INTERNAL-LINK:
https://sd.suse.com/servicedesk/customer/portal/1/SD-77220
Resolved by using ipv6  exclusively

Delayed Beta availability. Please expect it by end of month.
What about the backlog of SR? No change since last week
  
Bugzilla policy Leap related update prior Beta 
https://lists.opensuse.org/archives/list/factory@lists.op...
So as of now it's effective for Beta (will remove it from next
meeting).


When will Qt 5.15 hit Leap 15.4?
https://code.opensuse.org/leap/features/issue/7 let's keep QT related
discussion 
Should be there:
https://build.opensuse.org/package/show/openSUSE:Leap:15....
but not yet in repositories? (not published yet)


Leap Micro 5.2 - No update yet, aside from setup being done.
https://build.opensuse.org/project/show/openSUSE:Leap:Mic...

Discussion with Alex Osthof regarding volunteer interested in Cloud. We
came up with idea that he could start a 
Public Cloud testsuite for Leap, there seems to a blank space where we
could benefit from additional testing.
Dirk: wouldn't that be redundant after CtLG? Pehaps then test it only
in Leap
Gerald: What about some targetted smoke test, to ensure it works,
branding,...? Dirk agrees.
Tasks were given, now it's on Zack.

Doug: https://github.com/cisco/openh264/issues/3480
Just needed clarification on the PR as needing to be hosted by cisco. 
Looks like 3rd party is what is needed, correct? 
It will be streamlined in 15.4, just not sure if it will be a cisco
repository or another, correct?
it will be a cisco repo for openh264, rest will be streamlined.
Waiting for feedback.

## openSUSE Tumbleweed

* Kernel 5.16.10 in Tumbleweed with simple drm disabled again
* systemd-sysvinit is deprecated; there is a limited functionality
systemd-sysvcompat available (upgraders receive it automatically), but
most users probably won't really need it.
* Python 3.6 being EOL, removal of python36-FOO modules ongoing. Last
week: 73 left, now 51
* glibc 2.35 is shipped since snapshot 0216; not as much trouble as
previous updates, but Microsoft Teams fails with it. Use --no-sandbox
(twitter post warning users about this should be already out)
* Plasma 5.24.1 shipped in snapshot 0217
* KDE Frameworks 5.91 in staging; LibreOffice 7.3 in the same place,
fixes found build issues. QA pending
* GCC 12 pre-integration tests progressing (Staging:Gcc7); Scheduled
for actual merge in April/May

## Richard (Kubic/TW MicroOS) 

Not available

Working on MicroOS images for WSL, using a unique approach of bundling
the image inside the TW WSL image, which ends up acting like our usual
patched bootloader for MicroOS.
ongoing experiments with with kiwi-less, container-first VM image
building/adding nerdctl support to Toolbox/buildkit and its limits as a
container builder/TW-to-MicroOS converter

## Max

* Debugging autologin fail on KDE/sddm - wrong config file path [fixed]
* Fixing orphaned package issue
* Found uncleaned package still exists in SUSE:SLE-15-SP4:GA on OBS,
these package has been deleted in SLE15 SP4 on IBS and the binaries on
OBS was empty after syncing, however source package still exists, this
misleading skippkg-finder, the packages including boost, boost.175,
icu, yast2-schema, etc.
* Fix a unconditional matched bug in skippkg-finder for
15.3:Update:Respin

Leap Micro 5.2 - we're looking into it

On the GNUHealth in Leap 15.4 staging please check the python related
issues in the staging project. 

## Guillaume - Arm

Tumbleweed:
  * No blocker
  * glibc 2.35 includes restartable sequences (rseq) in particular an
optimised sched_getcpu that no longer needs to make a kernel call.
Performance improvements are expected, especially with MySQL.
  * Upcoming GCC 12 tested for aarch64:
https://build.opensuse.org/project/show/home:dirkmueller:...
  * armv6: rust fails to build for a while now.
https://bugzilla.opensuse.org/show_bug.cgi?id=1196328

Leap 15.3:
  * Issue with installation media (net installation) for devices w/o
Real-Time-Clock. Workaround: Trigger rebuild of installation media.
https://bugzilla.opensuse.org/show_bug.cgi?id=1194660
lkocman: to update the bug with installmedia refresh.
lkocman: respin done waiting for final s390x task to finish and then
let's republish it
https://build.opensuse.org/project/show/openSUSE:Leap:15....

Leap 15.4:
   * aarch64: 
        * Go compiler from SLE still lacks ld.gold which is still
required. This make some packages to fail to build (e.g.
openSUSE:Backports:SLE-15-SP4/rclone). Bug opened since Leap 15.3:
https://bugzilla.opensuse.org/show_bug.cgi?id=1183043
   * armv7: 
        * armv7 images are now tested in openQA with an armv7 kernel
overlay with a fix for the bug with btrfs and lzo compression:
https://bugzilla.opensuse.org/show_bug.cgi?id=1193852 


## Sarah - s390x
Tumbleweed:
* Reboot does not work:   
https://bugzilla.opensuse.org/show_bug.cgi?id=1193889
(forwarded to IBM for all Linux distributions on s390x - escalated by
Distinguished Engineer because of a well known issue related to new gcc
versions in combination with the kernel: IBM has created a patch, which
has to be tested now)

*Bug in s390-tools:
https://bugzilla.opensuse.org/show_bug.cgi?id=1195914 fixed with a
workaround yesterday
-> Waiting for next Tumbleweed release with tests in openQA
DimStar: that snapshot is already in openQA - looks to be broken now
for a different reason
    
    Builds for Tumbleweed are working again
    Progress in kernel debugging:
https://bugzilla.suse.com/show_bug.cgi?id=1188896

## Doug
*openSUSE Summit at OSCAL
* CfP started
* Google Summer of Code
* Application Review until March 6
* Leap 15.4
* Progressing with release annoucement
* Events
* CLT Sent info on schedule and program
* Open Source Festival (OSCA)
* Sponsoring Linux Application Summit
* Leap Micro talk submitted
* Keynote spot for dors/cluc

## Dirk

chasing a file read data corruption in 15.4 kernel due to transparent
hugepages, kernel team
developed a workaround
A few further fixes for /bin/sh == dash, remaining are kernel,
kubernetes and python macros

## Gerald

[no update]

## Wolfgang (Package Hub), Scott Bahling

Question from Fabian regarding openSUSE related branding in Backports:
Installcheck issue:   can't install sddm-branding-openSUSE-0.19.0-
bp154.1.1.x86_64:
nothing provides sddm-theme-openSUSE needed by sddm-branding-openSUSE-
0.19.0.x86_64

lkocman: To my understanding we couldn't have any bits built against
the openSUSE Branding in Backports itself as it was supposed to be
built the SLES way. More information and perhaps documentation is
needed.

In general at least we should avoid generating installcheck issues, if
there would be other way around it.

sbahling: We have reports (generated daily) at
https://packagehub.suse.com/reports/ of non-installable packages. I
would love that the dependencies get fixed or non-installable packages
get removed. This might be important for Leap as well (though the
current scripts do not take into account any packages provided by the
Leap specific repos). Core script is in SUSE internal gitlab
https://gitlab.suse.de/PackageHub/tools/packagehub_deps - uses SUSE
internal repo servers.
max: we do have a basic installcheck for Backports, the result has
uploaded to
https://build.opensuse.org/package/view_file/openSUSE:Bac...
, it is for Backports project build status only though
    
fvogt: Allowing openSUSE branding in Backports should be possible, it
should not be picked by default anyway.
max: yes, but problem for sddm-branding-openSUSE in Backports is: it
requires sddm-theme-openSUSE, which was generated from plasma5-
openSUSE, however we can not have plasma5-openSUSE in Backports,
because it BuildRequires wallpaper-branding-openSUSE, there is these
substitues be defined in Backports project config(I think Ismail did
that before), therefore plasma5-openSUSE just not able to be build in
Backports since wallpaper-branding-openSUSE can not be provided due to
the substitues below.

# openSUSE -> SLE magic so many environments can work
Substitute: desktop-data-openSUSE-extra desktop-data-SLE-extra
Substitute: desktop-data-openSUSE desktop-data-SLE
Substitute: openSUSE-release sles-release
Substitute: wallpaper-branding-openSUSE wallpaper-branding-SLE

## Maintenance team (Marina or Marcus, Maurizio (m4u))

- Marina: SLE-PackageHub overlaps
(https://github.com/openSUSE/openSUSE-release-process/issu...) work
in progress following the initial list of overlapping packages
(https://progress.opensuse.org/issues/106610#note-3)
- Marcus: 15.4 is set up channel wise, updates will go in testing
occasionaly, if you need more ping me, no openqa setup yet
- Maurizio: issues affecting PackageHub are now more visibile to us as
they are spotted by our tools, one in particular that needs attention
is  python3-avahi cannot be installed due to missing python3
dependencies (bsc#1196282). This is a problem that affects SLE Desktop
users because we don't have the required dependencies in the Desktop
modules either.
lkocman: to talk to Packaging Team (Matej)

Marcus: 15.3 in QA is stuck because of libreoffice udpate.  Container
image is red for quite a long time.
I had to update openstack-helper package with help from Dirk, this
duplication is not easily handled with PackageHub, rpmlint removes rpms
during the build which I did not expect. That behavior was introduced
for being able to build either the python2 or python3 flavors from
single source while removing the flavors provided in SLE (sbahling). It
only affects python builds.

Maling list archive seems to be not working reliably. There is already
a ticket openned, team is on it.

## Adrian - OBS 
Not avialable

Rudi is fighting with host kernel issues in IBS. There are quite some
issues with the SP4 kernel.
x86_64 - seems like the issue not caused by SP4/numa. 

Discussion about the Leap build pace.  
Seems like switching the rebuild strategy would only result into more
issues.
Adrian: could be caused by numa.  Let's focus on this once we'll have
stable workers again.