Last week’s update is available at https://discourse.ubuntu.com/t/foundation-team-updates-thursday-23-june-2022/
Package management:
- Implemented simple workaround for phasing issue, noticed test suite failing as phasing was only enabled for *upgrade commands, not policy that was tested - weird behavior
- Implemented different approach to phasing updates (only) in https://salsa.debian.org/apt-team/apt/-/merge_requests/245, seeking feedback
- Added more tests. Moar tests. Testing good.
- Uploaded those fixes for kinetic (sync from Debian soon), jammy, impish (all affected versions)
- Assisted @ogayot with the unattended-upgrades regression in focal. It seems there are a couple of bugs in handling python-apt C->Python->C->Python type call chains that probably cause some issues here
Secure Boot:
- Found out dkms modules were no longer signed, filed LP: #1980057 for now with analysis
- subiquity
- PR: 1335 - Some of the unit tests using asyncio were fragile, in the sense that the addition of other unit tests could break them. Switch over to using IsolatedAsyncioTestCase instead of using
asyncio.get_event_loop().run_until_complete()
tricks. - PR: 1337 - First PR extending the work for use_gap into being able to use partition table editing for guided install. This PR calculates some sizes, minimum recommended size for the install and suggested values for resize.
- PR: 1338 - Merge a cloud-init fix to the stable branch.
- PR: 1339 - bundle ntfs-3g in snap for ntfsresize
- (no PR yet) - work for Guided Install to work by resizing then performing guided in the space remaining. Solving issues with boot and adding the API endpoints.
- Review PR: 1331, 1334, 1336
- Spec reviews, including libnetplan
- Investigate and file LP: #1980008 - we need some sort of fix for Canary isos to make LVM + LUKS work
- release 22.06.1, with fixes for dasd/bcache, and a cloud-init output format fix
- PR: 1335 - Some of the unit tests using asyncio were fragile, in the sense that the addition of other unit tests could break them. Switch over to using IsolatedAsyncioTestCase instead of using
- curtin
- release 22.1, with primary changes being partition editing capabililties and RISC-V support
- probert
- PR: 120 - handle filesystem resize tools not being present
- ubuntu-desktop-installer
- PR: 956 - bundle ntfs-3g in snap for ntfsresize
- misc
- glibc
- discussed some topics with @schopin
- wrote a tool to help manage backport of upstream patches
- build a 2.36 snapshot
- various spec and code reviews, including first version of deb822 sources by default
Excellent name for a merge:
git remote add ubuntu-installer git+ssh://bernard010@git.launchpad.net/ubiquity
I am bermard010 why are you ssh-ing with bernard010 at git.launchpad.net ?
I would like to thank the entire team for all of your hard
work and dedication for Ubuntu !
- FTBFS reports: kinetic baseline and kinetic GCC12
(finished on amd64, i386 and ppc64el, still running on arm64, armhf, riscv64 and s390x) - Finished my +1 shift last week and sent my +1 Maintenance Report
- Started investigating which packages keep configuration files in
/etc/
- Reviewed a MR adding support for LXD VMs to
autopkgtest
- While looking at failed build logs, I noticed python-suntime failed to upload due to a missing ‘Changed-By’ filed in the changes file, and fixed it with a no-change rebuild
- Sync’d m2crypto (TIL universe package)
- Uploaded dotnet6 fixing some dependency issues
@bernard010 are you referring to the merge guidelines?
You will only see them if you are logged in to Launchpad, and they will be customized for you.
subiquity
- Reworked implementation of Ubuntu Pro (with manual token input) to match new mockups. PR 1331. Added better support for answers-based automation.
- Added new information to
/ubuntu_pro/check_token
API in PR 1340. Feedback from the desktop team would be appreciated since it is a backward incompatible change. - Refactored the code handling validation fields in forms. PR 1336
Experimenting with ways to define relationships between fields in a declarative way. - Attended different meetings for Ubuntu Pro with magic-attach.
- Came up with initial implementation of magic-attach in Subiquity.
- Brainstormed about ways to run automated testing for magic attach without USSO accounts.
distro
- Submitted workaround for unattended-upgrade autopkgtest regression on focal (thanks @sil2100 for sponsoring!). Root cause is yet to be identified. LP: #1974223. Thanks @juliank for helping in the investigation.
- SRU validation for MOTD asking for reboot when Livepatch is enabled LP: #1747499
Distro
- reviewed & staged systemd SRUs (EFI pstore), https://pad.lv/1978079
- usb-creator MIR sponsorings not needed (added comments, cc @alexghiti), https://pad.lv/1977959
- sponsored systemd-oomd SRU, https://pad.lv/1972159
- reviewed & sponsored the systemd v251 merge for @enr0n
- good job on keeping the June 30 schedule!
- https://code.launchpad.net/~enr0n/ubuntu/+source/systemd/+git/systemd/+merge/425171
Netplan
- Merged v0.104 into Debian unstable, migrated to testing
-
netplan status
spec comments - finalized & merged InfiniBand (IPoIB) PR#283 (cc @vorlon)
- discussions around regulatory-domain PR#281
- API & bindings spec publication & final review of PR#255
- netplan website discussions + small example fix
- investigated UC22/network-manager snap issue, https://pad.lv/1979277
- investigated netplan vs sd-networkd loopback IP assignment regression, https://pad.lv/1979951
- YAML consolidation preview deb for @mvo to run some tests on, PR#255
- running some tests, adding more test cases (after failure)
- testing YAML parser vs snapd on UC22
- Focal & Jammy SRUs, https://pad.lv/1956264 (and some UC20 related fixes)
- started working on VRF/VXLAN PR#272
Misc
Test environment for network applications (DPDK, SPDK, OpenVSwitch, rdma-core)
- package QEMU 7.0 in a ppa - RISC-V kvm requires QEMU 7.0
- create first Python classes for
- cloud-init data generation
- running QEMU with detection of cloud-init finalization
- SSH login
VisionFive kernel
- test kernel provided by @esmil
ubuntu-image
- more discussion on the germinate PR. It is almost ready to merge!
- Released the 2.x snap to stable \o/
- Two bugs were reported and fixed:
- manifest file formatting didn’t match 1.x
- Empty system-seed/system-data partitions were being treated as nil
- Continued work on pre-seeding classic images. PR to come
Miscellaneous
- Proposed migration: resolved vim FTBFS
- Created unicorn hat blog post
+1 Maintenance
- libdigidoc - fixed opensslv3 related regression
- rinutils - filed launchpad bug to explain why it can’t build unless we package some other dependencies
- golang-github-masterminds-sprig - fixed regression that should unblock the following other universe packages:
- golang-step-crypto
- golang-step-cli-utils
- golang-github-smallstep-certificates
- golang-oras-oras-go: Fixed regression caused by missing dependencies
General
- SRU reviews and releases
- Management trainings, courses and inductions
- 1:1 meetings with people from my squad
- Scheduled a meeting with Daniele about netplan documentation improvement progress and plans
- HWE meeting
- Reviewed and sponsored an SRU for Olivier
- Various FDE-related discussions and planning
- Some Jira project cleanup work
- DMB meeting
- Fixed some of the application pages to include information about our CoC signage requirements
- Started work on some new specs regarding image build tooling
Release
- 22.04.1 prep tasks:
- Created tracking discourse page
- Reached out to the kernel team for their plans and schedule for the upcoming point-releases
- Investigated issues with
rebuild-requests
not quite working from the crontab. It seems to put the image builds into ‘queued’ state but possibly fails running the build commands. Requested a bug filled and will re-visit when having more free cycles - After some final tweaks to debian-cd, built focal ubuntu-server test images for intel-iot with the new kernels
- Submitted the recent alpha image intel-iotg livecd-rootfs changes for review
- Prepared SRU bugs
- Now reviewed, started cutting a release for focal and jammy
- Filled in an RT for fixing cd-build-log mirroring from cdimage
Ubuntu Image
- Reviewed the germinate-state PR from William
- We discussed some of my concerns on the fortnightly meeting and I feel much better about the vcs field
- While testing, came up with a new work item to think about - progress reporting, which might need some design
- Reviewed the --thru/–until PR from William
Short week, was off Friday
systemd
- @slyon sponsored systemd merge (251.2-2ubuntu1) to kinetic
- SRU for LP 1972159 “systemd-oomd frequently kills firefox and visual studio code” (@slyon sponsored)
- Opened PR to fix a TPM2 test on Debian-like systems: https://github.com/systemd/systemd/pull/23858
- Sent an update to ubuntu-devel list regarding systemd-oomd plans
- Began working on src:systemd-hwe package
OpenSSL
- Uploaded 3.0.4-1ubuntu1, with a hotfix from upstream for the AVX-512 buffer overflow
- Uploaded a fix for stunnel4 autopgktests having a strict test on libssl version (previously forwarded to Debian but was never uploaded there)
Netplan
- Fixed some in the YAML consolidation PR related to field dirtiness tracking and configuration revertability
Misc
- Looked into glibc maintenance to help @mwhudson for 2.36 and the next Jammy SRU
- Fixed sbsigntool which I mistakenly synced in kinetic.
Rust
TL;DR: finally sorted out the cargo libgit2 bundling and autopkgtests situation, now having issues with libssh2 and libgit2.
As some might be aware, my goal is to have src:cargo in main this cycle. This package in turn depends on libgit2 and libssh2.
I uploaded a new version of cargo to kinetic that de-bundled libgit2 (which we kept vendored so far for ease of backporting) and fixed the autopkgtests. Before re-opening the MIR, I then looked into libssh2 and libgit2 to make sure there weren’t any surprises there. As it turns out, there were some.
libgit2
actually builds against mbedTLS for its HTTPS use, which is not in main. I’d like to keep that one out in universe as we already have two TLS libs in main, GnuTLS and OpenSSL. Luckily, upstream also has an OpenSSL backend, and my tests indicate that it works just fine. However, I’m still looking for feedback before implementing this, as I’m not sure whether it constitutes a breaking change or not.
libgit2
also depends on libhttp-parser, which comes out of the NodeJS project. Sadly, the upstream project explicitly states that it is “not actively maintained”. I looked into porting libgit2 to the recommended alternative, but it turns out that there isn’t feature parity.
libssh2
simply refuses to build in my sbuild environments. A test fail with what seems to be a crypto failure. Since the test didn’t fail when building in a VM or on Launchpad, I thought at first it could be an optimized assembly routine in libssl that’s misbehaving, but it turns out that builds in a LXC container are succeeding! I’m still trying to find the root cause of this issue, which I could also reproduce on some other hardware (different CPU vendor). Now trying to understand the difference between LXC and schroot environments.
- Reviewed and tweaked last bits on the Unicorn HAT blog post
- Built a tentative merge of u-boot 2022.07-rc4 (LP: #1980017) from Debian, primarily for RISC-V testing (including some minor bits upstreamed)
- raspi-config:
- integration work (LP: #1972982), including some potential upstreaming
- more or less re-writing all the overlay stuff in there to use overlayroot on the server images
- investigated ondemand cpufreq settings (it appears RaspiOS runs ondemand quite a bit more aggressively than Ubuntu; looking into the impacts of this)
- Pico-W developer experience work
- Rebased my long-standing tidy-up branch for livecd-rootfs
- Firefox performance measurement and work
- Pi meetings
apport
- Fixed bugs:
- Replace deprecated ‘imp’ module (LP: #1947425)
- whoopsie-upload-all: Fix ‘EOFError’ object has no attribute ‘errno’ (LP: #1979681)
- Fix AttributeError: module ‘importlib’ has no attribute ‘util’ (Follow-up for LP: #1947425)
- Fix calling in_session_of_problem with a dict (Follow-up for LP: #1978487)
- tests: Fix race with progress window in GTK UI tests
- Released Apport 2.22.0 and uploaded 2.22.0-0ubuntu1 to kinetic
- Cleaning up list of upstream bugs
- Uploaded apport 2.20.11-0ubuntu82.2 SRU for jammy
- Fixed apport-gtk: Fix importing the wrong Gdk version and uploaded 2.22.0-0ubuntu2 to kinetic
- Reported apport-gtk crashes if killing gnome-shell (LP: #1980197) which was already reported as LP: #1842439
- Prepared jammy SRU and uploaded apport 2.20.11-0ubuntu27.25
Distro
- Sponsored upload of sbuild 0.83.1ubuntu1 for @waveform (LP: #1974177)
The Ubuntu QA’s team update can be found here:
Thanks, that is very much appreciated.