Foundations Team Updates - Thursday 31 Aug 2023

Foundations
1

Previous week’s status can be found at Foundations Team Updates - Thursday 24 Aug 2023

2 Likes
2

Silicon vendor collaboration

  • Setup vendor board with Ubuntu, analyze boot chain

RISC-V kernel

  • Test new kernel build for VisionFive 2

Live installer

  • Update package cd-boot-images-riscv64 to current GRUB and u-boot-sifive
  • Add u-boot-sifive to seeds
  • Test rebuilt live-installer

Greenhouse

  • Evaluate written interviews
1 Like
3
  • subiquity
    • PR: #1770 - finish up zsys compat changes / fix CI on mantic
    • PR: #1771 - cmdline arg to disable offering core-boot install options
    • PR: #1772 - merge curtin updates for deb822 + add python3-debian dependency
    • PR: #1773 - zfsutils-linux trigger for install ISOs that don’t have it already installed
    • PR: #1774 - present guided capability flags in a predictable order. The first flag is the generally recommended default now.
    • PR: #1776 - default UEFI for dryrun, to fix the glitches in CI that happen when tests happen to be running on a BIOS system. Specific tests can choose a different bootloader instead, if desired.
    • PR: #1777 - fix autoinstall crash with desktop ISOs
    • PR: #1778 - bump curtin for grub-pc fix
    • lots of canary testing for the above
    • WIP on minimum guided size calculations. We will be making core-boot and classic boot variations of a given install source have the same minimum recommendation.
    • Help investigate packages not being installed from network when they should be
  • curtin
    • MP: #449974 - doc tweak for link to more info about apt and deb822
    • Investigate apparent regression in swap file creation. No bug per se, but curtin’s decision is kind of opaque, so added MP: #450289 to make the calculation clearer.
  • livecd-rootfs
    • LP: #2028862 - start SRU to prevent the same boot blockage that we saw on mantic live-server from happening on jammy. Ideally this would be merged prior to cloud-init 23.3 SRU.
  • livefs-editor
    • PR: #48 - fix resign_pool action, can fail due to gpg glitches around the default key
  • ubuntu-seeds
    • MP: #450177 - fix kernel install problem due to missing package in the pool. Merge remaining differences between ship-live and canary-ship live.
  • ubuntu-desktop-provision
    • learn enough about flutter and related tools to get ubuntu-desktop-installer dryrun working
    • opened PR: #106 as attempted fix about default capability flag choices, but Michael’s PR: #107 was better so I closed 106
  • File LP: #2033308 - conffile conflict on clean install and ubuntu-advange-tools upgrade
  • greenhouse
1 Like
4

dotnet

  • Backlog grooming to triage remaining tasks on the 23.10 roadmap.
  • Attended the weekly .NET partners sync meeting.
  • Attended the Canonical dotnet sync call to report on points discussed in the weekly MS call along with actionables on our side.

distro

  • Investigating nvme-stas autopkgtests on different architectures - reached a block on ppc64el where the QEMU VM won’t boot.
  • Investigating nvme-stas iputils autopkgtest hanging on s390x - bug report.
  • nvme-stas MIR update: nvme-stas and dasbus have security ACK and are ready for promotion.

misc

  • Attended the 360 Knowledge Sharing Session.
  • Started working on my 360 self-evaluation draft.
1 Like
5

Distro

  • vim: patched it to skip a few tests failing in our infrastructure LP#2033072
  • grub2: investigated an issue where grub2 would not generate menu entries if you’re using ZFS LP#2033256
  • network-manager: investigated and fixed some issues causing autopkgtests failures LP#2033391
  • network-manager: investigated an issue leading to veths not being activated automatically (working on a new distro patch) LP#2032824

Netplan

  • Working on netplan diff PR#386
  • Fixed a problem in our CI due to LP#2032824 PR#405
  • Improved support for WPA2 and WPA3 Personal PR#404
  • Fixed an issue in libnetplan where the netdef iterator is returning placeholder netdefs (and it shouldn’t) PR#406
1 Like
6

Ubuntu Packaging Guide

.NET

  • attended weekly .NET Partners Sync
  • attended MIR Meeting

Misc

  • attended 360 Review - Knowledge Sharing Session
  • ubuntu-mir PR #37 got merged
1 Like
7

ubuntu-image

Go

  • Start rebasing FIPS support patches for Go1.21.
  • Ensure packages that are rebuilt with Go1.21 have migrated.
    Found one regression in autopkgtest. Fixed in dh-golang 1.62.
1 Like
8

Distro

1 Like
9

Short week (back from vacations since Tuesday)

subiquity

  • merged recovery key implementation (classic use-case) to main
  • filed a FFe so that recovery key support can be released in the mantic installer: LP: 2033537
  • investigated issue with OEM meta-packages when doing offline installs. Turned out to be a buggy implementation when building our overlay FS. Working on a fix.
  • fixed an issue causing failed installs on certified hardware (wrong parsing of Ubuntu-Oem-Kernel-Flavour attribute)
1 Like
10
  • Merged changes from classic-redesign branch of pi-gadget into classic and proposed the necessary changes to the ubuntu-images repo
  • Confirmed fix for keyboard setting behaviour in cloud-init (LP: #2030788)
  • Still working on GPIO changes to upstream lg-gpio switch
  • Writing up NBD-based net-boot procedure
  • Pi meetings
1 Like
11

Short week (was off until Monday)

Distro

  • Preparations for an openssl SRU (planned: 3 bugfixes and a performance patch series), including more bug work on LP and trying to understand an issue where changelog is missing (been the case for 15 years maybe)
  • Specification work for crypto-configuration

Misc

  • Lots of catching up with messages, e-mails, news, …
1 Like
12

+1 week

  • Follow up on @liushuyu-011’s shift: Included his fix to python-requests-toolbelt in Debian, uploaded 1.0.0-2 to unstable, and synced to Ubuntu.
  • I used visual-excuses to look at clusters. There were no big clusters in universe. So I start looking at the oldest entries and checked their linked update-excuse bugs for updates.
  • bettercap-caplets: It looks like bettercap-ui was rejected from the Debian NEW queue (LP: #2023575)
  • gsl: Retried failing ruby-gsl/2.1.0.3+dfsg1-5build2 on ppc64el. It is still failing and needs to be investigated.
  • Opened python-oslo.log 5.2.0-0ubuntu1 FTBFS
  • dcfldd:
  • simde: The test suite on ppc64el failed. I could reproduce it on a ppc64el host. So I opened bug #2033648 for it. Tested the lastest upstream code which addresses those failures, but introduces two new ones: https://github.com/simd-everywhere/simde/issues/1057
  • spyder: The upstream test case test_get_user_environment_variables failed to get the PATH variable. The behavior could be reproduced in the autopkgtest environment by running env -i bash -l -c env (trimmed down what the upstrem Python code does). The latest upstream code (git master branch) changed to use bash -l inside bash -i which did not work as well. So I reported the failure upstream. (LP: #2033518)
  • dulwich: I guess upgrading breezy from 3.3.2-2ubuntu1 to 3.3.3-2 could solve the test failures (needs testing), but this would need a feature freeze exception.
  • dxvk: Depends on libwine-development-dev which is provided by wine-development. wine-development is not suitatable for Debian testing nor Ubuntu.
  • pycrc: 0.10.0-1 built successfully locally, but failed the same way after cutting the Internet access. Added a patch to use docbook.dtd from docbook5-xml instead of getting it from the Internet. Uploaded 0.10.0-2 to unstable.
  • phylonium: 1.7-1 FTBFS on armhf. The same is true for Debian. Created LP #2033638 and linked the Debian bug.
  • debug-me: 1.20221231-1 FTBFS on ppc64el due to internal ld.gold error. Created LP #2033639 and marked binutils as affected.
  • dipy: 1.7.0-1 fails to build on arm64, ppc64el, and s390x due to a failing test_cross test case. This failure has been reported upstream and test_cross is skipped in 1.7.0-2. So I synced that version.
  • py3dns: testNS and testNSD fails in the autopkgtest environment, but these test succeed in a local schroot autopkgtest. Probably the network setup in the Ubuntu autopkgtest environment let these test fail. Created LP #2033649 for that and marked upstream as affected as well.
  • bladerf: 0.2023.02-1 fails to build from source, because it tries to access the Internet (LP: #2033661). I changed the packaging to use pybuild, uploaded 0.2023.02-1ubuntu1, and forwarded the change to Debian.
1 Like
13

Release

  • Extensive testing of canary image to validate whether the feature was ready to land in the desktop image. Many bugs filed
  • Iterated on livecd-rootfs and live-build for the above
  • Also iterate with Michael on new image build failures on arm due to being unable to unmount /sys. We don’t know what in the build is now mounting a submount of /sys during build, but after a few rounds, a fix is landed that should unmount it.
  • Also extended the autopkgtests of livecd-rootfs to cover the Ubuntu Desktop, which temporarily regressed because of the live-build changes but wasn’t picked up until the regression was in the release pocket. Then struggled to see autopkgtest results because of infrastructure problems (which seem to still be ongoing)

Distro

  • Looked at a misbehaving systemd generator in mantic CPC images

General

  • Two days of patch piloting. queue effectively zeroed!
2 Likes
14
  • Yet another grub ZFS fix
  • Interviewer training
  • Short experiment for using AF_ALG socket for message digests in APT, found bug with kernel not autoloading accelerated sha modules.
  • Patch Pilot
1 Like
15

Been doing Iso testing of canary images, legacy images and standard mantic images.
Couple bugs reported:



Which were discovered whilst investigating a smoke testing failure we’ve had. The investigation has been taken over by @seb128

Discussing an MP to prevent looping in the case of legitimate testbed failures in autopkgtest-cloud with @paride and @bdmurray . Discussion still ongoing for this (MP was created because we had an issue where debvm triggered by mmdebstrap was infinitely looping).

Finished working on an MP for autopkgtest-cloud which reintroduces and updates the unit tests for the autopkgtest-web charm which were written quite a long time ago. Now that the unit tests are working we will look to incorporate them in CI.

Worked to fix the autopkgtest-cloud infra when we had issues with full disks with various juju units. I think this happened yesterday and today.

Updated the charms in staging to be up to date with master and remove all cowboys.

Finally “fixed” a bug in Utah which left desktop VM’s in a “shut off” state on venonat, and not destroying their storage. The fix is cowboy’d on venonat and we’re monitoring it to make sure it actually works. Once we’ve verified it does, I will work on amending the fix to be more elegant and suitable for merging.

1 Like
16

Distro

  • Finished deb822 work in curtin
  • Uploaded systemd to fix LP: #2031416, which unblocked systemd for migration
  • SRU verification for Lunar and Jammy systemd SRUs
  • Prepared/tested patch for LP: #2033325, and will include in the next systemd upload
1 Like
17

openjdk:

openjdk-fips:

  • raised MP implementing argument files and some small fixes in openjdk-fips launcher

aom:

  • working on build-time tests - discussing approach with the maintainer
1 Like
18

Short week, due to a sick day.

Distro

Netplan

Misc

  • pulse planning (roadmap grooming), Debian blue item
  • travel prep, India (visa, documents, expenses, vaccinations, …)
  • 360 review prep
  • Take-home test reviews
  • Interviewer training
  • resilience training
1 Like