Last week’s update is available at Foundations Team Updates - Thursday 25 August 2022
20.4.5 point release
- Testing RISC-V images
PolarFire Icicle Kit
- Testing images with different versions of HSS firmware
U-Boot
- Fix a bug with EFI system partition detection
GRUB
- Started on rebasing RISC-V related patches for upstreaming
ubuntu-image
- Further review of cloud-init PR
- Merged all unmerged branches together and started building images
- Have been testing on Raspberry Pi images
- This showed some issues that I added to the miscellaneous cleanup PR
- The rest of the issues were able to resolved by changes to the image definition file, as we hoped!
- After resolving these issues I ran through all of the ISO tracker test cases for Raspberry Pi images to confirm they are working correctly
- Created a PR to fix compatibility with snapcraft v7 credentials
- Started discussion about kernel and initrd sourcing for special snowflakes such as the Raspberry Pi
- Began implementing a possible solution that was proposed in the email thread
Miscellaneous
- Helped @waveform with failing dbus autopkgtest
- Sponsored resulting dbus change
- Reviewed many take home tests
- Candidate interviews
- Filled out engineering promotion application
Distro
- MIR: dropped new gsasl->libgssglue dependency (component-mismatch), https://pad.lv/1972866 (cc @jawn-smith)
- systemd-repart SRU, https://pad.lv/1897932
- sponsored sytemd-hwe into Jammy, https://pad.lv/1981599
- systemd Kinetic fixes + symbols versioning/ABI (cc @enr0n)
- review & sponsor systemd Focal SRU
- review & sponsor systemd Jammy SRU
Netplan
- netplan autopkgtest vs OVS/armhf, pending pandoc/armhf build (cc @fnordahl)
- netplan everywhere email
- Netplan everywhere PPA (cc @vorlon)
- netplan status draft, PR#290
- UC20/22 netplan crash bug analysis, https://pad.lv/1959570
- Diataxis emails
Misc
- candidate take-home test & interview
- prepare 360 self-review
- subiquity
- PR: 1386 - complete PR fixing issues related to editing partitions and when to set the wipe value
- PR: 1396 - remove duplicate copy of jsonschema from snap
- PR: 1397 - fix kname lookup that was blocking from returning all OS prober results, as explained in u-d-i issue 1084
- PR: 1399 - quick PR to a feedback item on PR: 1392, so that we can merge these and start a feature freeze branch
- PR: 1402 - some changes for more kind terminology
- PR: 1403 - cleanup on snapcraft file
- Review PR: 1390, 1392, 1394, 1398, 1400
- Start stable branch for work targetting Kinetic
- Focal ISO testing
- casper - successful pmem experiments for chain-booting to a different kernel in the netboot environment
- ubuntu-desktop-installer
- (LP: #1988042) - Suggest client side fix for random startup error
- curtin
- update-notifier - investigate autopkgtest failure of update-motd. recent change may cause update-motd to return non-zero depending on partition config. Open LP: #1987703 and matching MP 428940, merge, upload, migrated.
- oz: syncpackage to merge with Debian
- Candidate packet reviews
- Proposed migration:
- cross-toolchain-base / kmod / procps, retried tests and uploaded initramfs-tools with an update from Debian for depmod from kmod 30
- requested removal of python3-pyfai and python3-freesas binaries on riscv64 (LP: #1979297) to allow migration of pyfai
- Provided additional info for [MIR] lib*-perl for lintian 2.115 (LP: #1980662), libfreezethaw-perl and libhtml-tokeparser-simple-perl now complete, awaiting security review of libwww-mechanize-perl and libmldbm-perl
- Noticed the recent sync of autoconf-archive dropped some patches, pinged the Debian maintainer and got that resolved, sync’d the fixed autoconf-archive
- Still taking inventory of packages with configuration files in /etc
- Tested new Lintian release including a fix for Debian bug #993613, merged and uploaded lintian
systemd
-
ManagedOOMPreference
PR merged upstream: https://github.com/systemd/systemd/pull/23764 - Prepped SRU uploads for Jammy and Focal with @slyon
- Investigated ABI compat way to backport
sd_hwdb_new_from_path
symbol needed for Jammy SRU
- Investigated ABI compat way to backport
- Opened MP for kinetic to fix sd-resolved installation (https://code.launchpad.net/~enr0n/ubuntu/+source/systemd/+git/systemd/+merge/429166)
- Opened request to remove sicherboot from kinetic due to systemd-boot conflict (LP 1988070), and @vorlon did the removal
- Asked reporter for test case on LP 1982218 “wait-online does not correctly identify managed links”
proposed-migration
- siridb-server vs libuv1 (LP 1987558)
- Tracked down the issue to be a NULL dereference, although it was unclear how the new libuv1 revealed this bug. I sent the patch upstream, and the Debian maintainer grabbed my patch before I got a sponsor in Ubuntu. The packaged synced from Debian and unblocked libuv1.
- Fixed a follow on issue in pcre2 that caused siridb-server FTBFS on ppc64el (LP 1987850)
- nodejs vs libuv1 (LP 1987693)
- Pulled upstream patches to apply against nodejs 16.14.2+dfsg1-1ubuntu3, but then nodejs 18.7.0+dfsg-1ubuntu2 in -proposed built successfully and contained the same patches. This has since migrated and unblocked libuv1.
Other
- u-r-u upgrade bug triaging
- Jammy SRU for systemd-hwe (LP 1981599)
- Verified that systemd-hwe is in jammy-proposed
- Merged python-pecan from Debian (LP 1987701)
general
- socket-activated sshd has landed in kinetic release. So far no users complaining about the behavior after upgrade.
- work to support introducing ubuntu-unity as an official flavor
proposed-migration
- managed retries of various autopkgtests failing with the new glibc. Removed llvm-toolchain-12 and revdeps as requested. Patched wcc for new ldd output. Got glibc migrated on Friday by hinting away the remaining failures (pygai, lxc).
- kodi-inputstream-adaptive: new version has incomplete endianness detection, blocked NBS library transitions so fixed and forwarded to Debian
- worked on driving the haskell transition to completion; hindered by very poor ghc performance on the buildds (and only there)
NBS
-
opencv
- removed
pytorch
and reverse-dependencies - removed
python3-ffcv
on riscv64 to letffcv
migrate (newffcv
build-depends onnumba
which isn’t available) -
tomopy
ftbfs on ppc64el because ofskimage
binary skew which is now fixed - uploaded
vitis-ai
to rebuild against currentopencv
, missed before because it’s arm64-only; but it ftbfs on a rebuild, and ftbfs again after patching to ignore possibly-uninitialized variables. This is an Ubuntu-specific package, so bumped the upstream version and made a few changes to get it building again. - removed
os-autoinst
and reverse-dependencies
- removed
-
openexr
-
openexr-viewers
ironically fails to build with newopenexr
. Removed. -
aqsis
ftbfs. Removed. -
openvdb
build failure fixed in Debian unstable. Merged. -
vips
blocked byruby-vips
autopkgtest regression; real regression, have not had time to investigate.
-
-
benchmark
- no-change rebuild of
ros2-performance-test-fixture
, wrong dep only on armhf
- no-change rebuild of
-
nodejs
- no-change upload of the only revdep of
libnode93
,node-shiny-server
.
- no-change upload of the only revdep of
-
rocksdb
- no-change upload of the single remaining revdep,
balboa
- no-change upload of the single remaining revdep,
-
qemu
- only reverse-recommends; removed
-
ldc
- worked through the reverse-dependencies; the D stack is in very poor shape in Debian unstable, had to submit several bug fixes and remove several packages
-
fmtlib
- Removed
opendht
which ftbfs in Debian and Ubuntu - Removed
ensmallen
which was removed from Debian testing and blockingarmadillo
, a dependency ofgnss-sdr
- Tried to reproduce the
gnss-sdr
riscv64 build failure locally and could not, so retried the package build in launchpad and it failed. Reproduced in a riscv64 VM locally but could not make sense of the cmake horribleness in trying to detect lapack; so rather than trying to figure out why it was failing, just overrode cmake to prefer pkg-config for liblapack detection. Then got a new build failure about missing symbols, which maybe was the root cause of the first failure. Still investigating.
- Removed
- Merges!
- dbus in LP: #1959211
- pexpect in LP: #1987420
- u-boot in LP: #1980017
- lvm2 in LP: #1987340
- initramfs-tools in LP: #1988418
- Many thanks to @slyon and @jawn-smith for sponsorship and debugging aid in dbus, and to @bdrung and @vorlon for sponsorship of other merges
- Investigated state of upstream mesa with regard to firefox performance on the Pi (some important improvements but still a few kinks to work out!)
- Verified that LP: #1924251 is indeed fixed on kinetic; still investigating exactly which thing fixed it there (and thus whether a jammy backport is feasible)
- Sponsored Mediatek AIoT 1200 changes to flash-kernel for kinetic (LP: #1986517)
- Point release meeting
- Pi meetings
I am on +1 rotation this week, but still had to do “core” tasks that were time-critical or “crimes of opportunity”.
Misc
-
Investigated lxc FTBFS against glibc 2.36.0. I backported the patches, but we’re still hitting a ppc64el build failure caused by -O3. The root cause of the failure still remains to be found.
-
Packaged, uploaded and verified update 2022c to Kinetic, Jammy, Focal and Bionic (LP: #1986984)
-
Did yet another Nvidia-related install check
Netplan
- Uncovered a bug when parsing routes for the same interface from different sources, worked on reliable reproducer and detailed diagnostic. (LP: #1988444)
- Investigated a
netplan set
reported on IRC by a community member
+1 rotation
-
investigated node-mqtt vs mosquitto, turned out to be a flaky
-
node-fast-safe-stringify/arm64 -> internal 30s timeout on tests that take < 5s on Debian, turn out to take 40+s on Ubuntu. Worked around by explicitly setting a bigger timeout, but would be worth looking into a bit more.
-
presage: ported to C++17 to fix FTBFS
-
lua-cliargs/lua-system/lua-say: follow-up on Bryce Harrington’s work back in July on the Lua ecosystem, those packages just needed to be triggered with each other to migrate in bulk from -proposed.
-
rust-pyo3-macros: retriggered riscv64 build, was waiting for another rust package to build
-
rust-object: investigated s390x autopkgtest regression, turned out not to be a regression and not easily fixable, hinted by tumbleweed
-
busco: needs hinting, see hints-ubuntu MP
- Merged a couple of grub patches into the devel git branch
- Found issues in VPN and NetworkManager OpenVPN support
- Setting up servers and preparing to benchmark stuff (and initial benchmark)