Last week’s update is available here: Foundations Team Updates - Thursday 28 April 2022
- Uploaded APT 1.6.15 to bionic, marking a complete set of updates for the backport of the current kernel autoremoval code
- Worked with @gjolly on nullboot to add support for writing JSON files (PR#17, PR#18)
- Notified coveralls.io support that the pages for nullboot are not visible; it got fixed
- Analysed, proposed fixes for regression from recent networkd-dispatcher CVE: https://bugs.launchpad.net/ubuntu/+source/networkd-dispatcher/+bug/1971550; also forwarded bug to systemd about incomplete networkctl manual page.
- Initial work on improvement to apt phased update support
- Testing of curtin changes - have now seen Subiquity (when driven by the API, not in the Server Terminal UI yet) be able to resize and ext4 or ntfs partition and install into the remaining space while preserving the boot of the other OS
- Chat with @chad.smith about better interactions between subiquity and cloud-init.
- Review: PR: 1286
- MP: 420764 - Support setting arbitrary partition type values. Required in order to preserve the type values on existing partitions when they do not map to an existing curtin flag. Also preserve the disk label id, an absolute requirement for some dual boot cases.
- cyrus-imapd (LP: #1971469) - Fix unit test failure related to OpenSSL 3 and EVP_PKEY_base_id
- osmo-mgw (LP: #1971620) - adjust debian/rules to not install the upstream systemd unit file, as the Debian packaging already provides one.
- openmpi - investigating. nwchem might be simple retest, gyoto has a dependency that is LTO grumpy, armci-mpi new autopkgtests are a bit broken
Awesome job everyone is accomplishing, Foundations is the
beginning of every new operating system. This is where everything
starts and eventually through hard work and integrity with hard
work becomes an O.S. Named Ubuntu…!
Thank You for all your hard work and dedication for Ubuntu !
Awesome Job… !
- submitted a firewall change merge proposal which allows the staging Error Tracker access to radosgw in PS5
- submitted another firewall change adding the failed-queue retracer app IPs to the retracer group
- updated the errors code switching from jammy to 22.04 and adding support for kinetic
- multiple daisy code updates
- use swift version 3.0 for core file storage
- added in retracer config for kinetic
- built apport-test-crashes for kinetic
- tested them in the staging Error Tracker
- set it up for Kinetic Kudu
- reviewed, merged and sponsored @enr0n’s Kinetic u-r-u changes
- tested a dist-upgrade bug from 20.04 to 22.04 and recorded my test using script
- worked on the archive opening for Kinetic Kudu
- updated archive opening checklist template with more detailed information
- uploaded ubiquity for kinetic
- SRU reviews and releases
- attended Ubuntu Developer Membership Board meeting
- point proposed-migration reports at kinetic
- unblock proposed-migration
- hand over jammy-proposed from release team to SRU team
- turn on autosyncing and flood the autopkgtest queues
- update llvm-toolchain-14, because llvm has a hard-coded list of Ubuntu release names that is used in order to set default compiler options hnngh
- IS asked if we could reduce swift usage by autopkgtests so I started doing the removal of hirsute test results for the (yet to be completed) hirsute EOL process. As in past cycles, this has made the swift cluster very angry due to the large number of deletion operations coming in all at once, so IS also started getting alerts and wondering what was going on so I pointed IS at IS and gave the deletion a few coffee breaks. The deletion is still ongoing (amd64 results deleted, now going through arm64).
- also found that there are other buckets still around (ppas, etc) for other EOLed releases. Will work on clearing those out once autopkgtest-hirsute is done.
- Established that the exim4 autopkgtest regression on ppc64el showing up against pam was not caused by pam but was random exim4 breakage; hinted exim4 away and pam has migrated, letting us drop krb5 out of the essential package set! now have my eye on getting nis dropped from libpython3.10-stdlib as well, since the nis python module is already deprecated for >python3.10 and this would let us drop it out of
Priority: importantand the default install.
- first pass at a patch to make openssh use systemd socket activation, reducing the baseline memory consumption of idle systems.
- Raised an MP against launchpad-buildd to allow cherry-picking packages from -proposed during an image build, analogous to autopkgtest’s interface (-proposed test images!)
- Worked on configuring next-generation UEFI signing keys for SecureBoot.
- In Copenhagen this week for the 22.10 product sprint.
- Rebased both VisionFive/Nezha kernels on top of latest Ubuntu unstable branch and handed that to the kernel team
- Start working on Wifi driver for the Nezha board (which only works on 5.4 kernel, which is the kernel used by Fedora and Debian)
RISC-V large initrd size investigation:
- actually ‘normal’ since cryptsetup-initramfs and linux-modules-extra are installed by default on preinstalled images
- PR to initramfs-tools to warn the user if the initrd is larger than X% of ram size, because it silently fails to boot otherwise: https://salsa.debian.org/kernel-team/initramfs-tools/-/merge_requests/62
- Will try to remove those packages from RISC-V installations (thanks to @jawn-smith for the pointers)
u-boot-nezha: Needs sponsor to upload latest upstream version which contains:
- automatic patching of kernel dtb with memory nodes from nezha-boot0
- add opensbi version in deb package
- fix expected model name
u-boot-licheerv: Needs sponsor but is low priority, was a nice and simple consequence of the new u-boot-nezha
Sorted and updated list of rebuild failures for RISC-V, started to fix packages:
- aseba FTBFS needs sponsorship
- zzuf FTBFS needs sponsorship
Prepare RISC-V rebuild without nocheck:
- ‘Benchmarked’ qemu 7.0.0 which supports up to 32 cores compared to qemu 4.2.0 with only 8 cores: not very conclusive, requires more investigation as to why it does not perform better with more parallelism.
- Added visual indication for starred snaps publishers PR 1292. Noticed and got rid of unexpected green font color in basic mode (e.g., over a serial console) where we only use monochrome.
- Fixed crash when going back from FS screen. 22.04.2 was not affected by the regression so no need to roll out a fix yet. PR 1291
- HTTP proxy settings being discarded: merged curtin workaround MP 420651 (i.e., disabled code that removes HTTP proxy settings). Investigated root cause and traced it down to the way we handle debconf-selections. Enumerated different options to fix the issue consistently.
- Made recent changes in /source API backward compatible with flutter desktop installer PR 1287. New expected arguments were added to /source as part of the redesign for third-party drivers
- Rediscussed & redesigned copy-API requirements for flutter desktop installer with @jpnurmi. This will most likely be replaced by a new network endpoint having the ability to copy NetworkManager configuration files to the target.
- small fixes around the ABI checker
- some bug triage
- Drafted PR#274 (ip6-gre[tap] tunnels), https://pad.lv/1966476
- Working on NM unmanaged devices PR#276, https://pad.lv/1951653
- Working on specification for VRF / VXLAN yaml schema
- Initial review for PR#272 (VRF & VXLAN support)
- aspectc++ LP: 1970947
Part of the build issues can be fixed by suppressing a compiler warning. But our current libc6-dev and llvm don’t fit together. The problem is described in https://bugs.llvm.org/show_bug.cgi?id=52265
- Looking into SPDK and DPDK
RISC-V support is provided by in https://github.com/ventanamicro but the complete change history of the the Ventana changes is missing.
Product sprint in Copenhagen
- start implementing classic image redesign
- Introduce yaml parsing and schema validation
- Still bikeshedding certain key names for the yaml definition
- Investigating occasional test failures in TestGadgetEdgeCases
- Current theory is that there’s a map somewhere within snapd that isn’t having the order preserved, leading to occasional failures
- start implementing classic image redesign
- Raspberry Pi
- Fixing LP: #1970761
- Have local changes, but testing has been slow.
- Working on adding unit test cases
- Investigated totem failure on Pi desktop (LP: #1969512)
- Investigated reports of excess initramfs size (from IRC); flash-kernel probably needs to do some more useful/sensible with backups
- Auto-connection of
pi-bluetoothon uc22 was resolved with some store updates
- Figured out the CM4 USB issue (flashing cable attached while USB-keyboard plugged in … doh!) (LP: #1969689)
- … amended Jammy release notes accordingly
- Working on picotool packaging
- Pi meetings
- Finalized PR #254
- Had fun with C string manipulation in PR #275
- Rebased the final YAML consolidation PR #255
- Prepared a SRU for the TR locale issue
- Uploaded latest release 3.0.3 to kinetic
- Prepared the SRU for the ruby-net-ssh issue
- Working on safely reverting the grub-customizer mess
- remote product sprint participation
- synced util-linux, realized I shouldn’t have done that and merged it properly
- subiquity and curtin reviews
- glibc 20.04 SRU verification
- uploaded glibc for bionic
- Started working on LP 1966800 (“systemd locks up due to incorrect handling of time zone changes”) and LP 1959475 (“machinectl shell connections immediately terminated”) for eventual focal SRU
- Continuing to investigate upgrade bug reports
- Updates for Kinetic
- Copenhagen induction sprint