Foundations Team Updates - Thursday 08 September 2022

Last week’s update can be found at Foundations Team Updates - Thursday 1st September 2022

1 Like

Microchip polarfire icicle:

  • livecd-rootfs image ok, PR coming
  • flash-kernel fix uploaded

LicheeRV Dock:

  • livecd-rootfs image ok, PR coming
  • flash-kernel new entry uploaded
  • wifi driver dkms ok, but fails in a chroot, ongoing
  • merging u-boot support in u-boot-nezha to avoid duplication

RISC-V Archive rebuild without nocheck:

  • half packages fail for legitimate timeout (riscv builders are very slow), half are ‘true’ failures to fix. Stopped at rustc for now.
1 Like

ubuntu-image

Proposed Migration

  • Working on vorta vs python-secretstorage
  • Have recreated the timeout in a qemu environment
  • It appears to be an issue with gnome-keyring, but attempts to troubleshoot it have not worked so far

Golang

  • Prepared uploads of 1.18.6 and 1.19.1 for Debian
  • Still awaiting sponsorship, will merge to Ubuntu when they are in Debian

Miscellaneous

  • Graded a few take home tests
  • Candidate interviews
1 Like

Short week with US holiday.

systemd

  • Started Focal SRU verification
    • autopkgtest regressions resolved with retries.
    • LP 1959475 “machinectl shell connections immediately terminated”
    • LP 1979951 “127.0.0.1/::1 removed from loopback interface if you configure extra IPs on lo”
  • Started Jammy SRU verification
    • Investigated autopkgtest regressions (LP 1988994). Will need a follow-on upload to skip TEST-58-REPART on ppc64el.
  • Fixed LP 1988772 “kinetic livefs builds are failing on dns errors”
    • Needed to pull some Ubuntu-specific /etc/resolv.conf logic from systemd.postinst, and add it to systemd-resolved.postinst.

proposed-migration

Other

  • Became a member of for ubuntu bug control team
  • u-r-u upgrade bug triaging
1 Like

Short week due to PTO.

subiquity

  • Opened PR 1408 to address some issues reported by mypy
  • Addressed LP: #1894009 (wrong keyboard layout when prompting luks passphrase after an install) through PR 1411
  • Investigating LP: #1986674 (race condition in snaplist causing a jump back to a previous screen) - not yet reproduced
  • Investigating LP: #1988407 (crash when editing LVM partition) - reproduced in dry-run mode.
1 Like
  • Working on rustc 1.61 and cargo 0.62 for the upcoming Firefox release. rustc 1.60 already packaged, cargo 0.62 is (as usual) proving difficult, rustc 1.61 is pending.
  • Tracked down a weird compilation warning on lxc when using gcc-12/ppc64el (FTBFS due to -Werror). Turned out to be the compiler being too smart for itself. See the upstream lxc issue
  • Graded a few take-home tests, and prepared a candidate interview.
1 Like

Light TH1520

  • Mix vendor kernel, U-Boot with Ubuntu image and test board
  • Analyze vendor U-Boot, rebuild, reflash
  • Analyze vendor kernel

Polarfire Board

  • Test with prerelease HSS firmware

Kernel 5.19

  • Bisect configuration to solve
    • LP #1988984 - CONFIG_PWM_SIFIVE=y leads to failure of early console
    • LP #1988990 - Radeon graphics card not usable
1 Like
1 Like
  • subiquity
    • PR: 1404 - fix some partitioning breakage introduced last week. Restore older behavior.
    • PR: 1406 (LP: #1983265) - fix netplan apply warning about gateway directives, and merge to stable branch (PR: 1409)
    • PR: 1407 - merge below curtin fix
    • Review PR: 1359, 1405
  • netinstall - do two protypes of a image choice menu, first with go, then with a /bin/sh script and jq (which is 4 MiB smaller)
  • curtin
    • MP: 429517 - test @jpnurmi’s proposed fix for cryptsetup issues and open MP
    • MP reviews 1, 2
  • proposed-migration - boost vs hilive - long-time buggy test, decided to just retry, which allowed boost migration
  • Candidate packet reviews
  • Specification reviews
1 Like
  • Writing blog post covering planned GPIO changes for 5.19 (and investigating current status)
  • Investigated lack of HDMI output on Kinetic Pi Server images (LP: #1988838)
  • … turned out was not related to another Kinetic display issue, but keeping an eye on that too (LP: #1988859)
  • Sponsored u-boot changes for PolarFire board (LP: #1987505), thanks to @sil2100 for sorting out the new binary there
  • Sponsored flash-kernel changes for LicheeRV and Microchip (LP: #1987735, LP: #1988169)
  • Reviewed flash-kernel merge (LP: #1955353) for @jawn-smith (finally!) with @alexghiti’s input (thank you!)
  • Take home test review
  • Firefox meeting
  • Pi meetings
1 Like
1 Like

Ubuntu QA team’s updates:

1 Like

proposed-migration

  • continued chasing the haskell transition.
    • uploaded haskell-devscripts with a fix that skips the arch-indep package build when building with -B, bypassing a builder timeout problem on armhf.
    • retried pandoc but it got stuck on one of the slow armhf builders and eventually got killed. Restarted and it got a good builder instead.
    • flagged the armhf performance problem to #debian-haskell as ghc being slower on armhf than on riscv64 emulated autobuilders is ridiculous
    • everything finally migrated!
  • removed lomiri-settings-components because it depends on a non-existent package. Recorded in extra-removals and filed a bug in Debian.
  • libigloo: no action from Debian maintainer on bug report about big-endian breakage. Since the package has no revdeps, removed the s390x binary packages to let it migrate.
  • xtpcpp: claimed to be stuck due to no binaries, but that’s because it was superseded by i2masschroq which took over the binary package. Not sure why Debian hasn’t removed this source yet, but removed from release and proposed.
  • tb-goodies: depends on non-existent webext-xnote package. Removed, added to extra-removals.

NBS

  • fixed a bug in ignition-common that caused ignition-launch to FTBFS
  • removed openboard, incompatible with both ffmpeg and poppler and removed from Debian testing
  • ffmpeg
    • took a look at porting ffmpeg2theora, spent way too much time on this before hitting the inner decoding loop that needs porting, then looked around and realized ffmpeg has native support for theora so this is duplicate functionality that should just be dropped as obsolete. Removed from kinetic.
    • telegram-desktop runs the buildd out of memory. Looking around I notice it uses -pipe in its upstream build flags, which increases memory usage. Uploaded with a patch to remove that flag. This helped, but the package still failed to build at the linking stage.

general

  • uploaded @enr0n’s patch for bug #1988772 in systemd-resolved to try to get image builds fixed asap. Daily images now building again!
  • reviewed/sponsored some Farsi fonts
1 Like