Foundations Team Updates - Thursday 25 August 2022

Project Discussion Foundations
1

Last week’s update is available at Foundations Team Updates - Thursday 18 August 2022

2 Likes
2

proposed-migration

  • lua-cliargs: not failing to build in Debian. Tracked this down to lua-system not being rebuilt for lua 5.4. Did a no-change rebuild, but turns out lua-system was dep-wait in -proposed, on a package that ftbfs because lua-system isn’t built for lua 5.4. Uploaded lua-mediator, lua-say, and lua-system with build-time tests skipped to unblock this
    lot, then reverted those changes.
  • tpot: sergiodj identified an upstream fix for this during his last +1 maintenance shift but did not upload. Uploaded now.
  • probabel: Debian dropped all archs except amd64. Removed the binaries from kinetic-proposed to let this migrate.
  • libmojo-ioloop-readwriteprocess-perl: armhf-only autopkgtest regression, has been retried a few times by several people. Failure is not reproduced in Debian, so I checked whether I could reproduce it myself in an armhf container in canonistack. And this failed, because there were tests of containers, and my lxd environment was probably not set up correctly for nested containers. Although this was not the test that failed in autopkgtest, the fact this passes in Debian suggests to me this is a containerization issue (since armhf autopkgtests run in containers but no other archs do) and I added a badtest hint to let this migrate.
  • golang-github-vishvananda-netlink: test data is little-endian so fails on big-endian archs (s390x). This is a bad test and Debian has let it into testing as-is, so added a badtest hint.
  • ruby-jwt: LP #1981456 documents that the package fails to build with openssl 3.0. That’s not going to change without a sourceful fix, so there’s really no reason to leave the broken package in -proposed. Removed it, pending a Debian fix. Would warrant removal from the release pocket, except there are non-trivial reverse-dependencies.
  • salmon: another instance of a ftbfs with current glibc due to embedded catch2. Cherry-pick the new version from upstream git. The package still fails to build with a test failure, but uploaded anyway since this gets us closer to building.
  • ubuntu-kylin-wizard: was removed from the release pocket recently but I missed removing it from -proposed. Corrected now.
  • geneweb: FTBFS in Debian and Ubuntu, removed from Debian testing, removed now from kinetic.
  • python-djvulibre: bad test on big-endian archs, the code itself is fine. Debian has let this into testing, so add a badtest hint to let this into kinetic.
  • python-django-celery-results: tests pass in Debian but fail in Ubuntu. Retried with --all-proposed, still failed.
  • node-puppeteer: uninstallable because chromium packaging differs between Debian and Ubuntu. Updated the package to depend on chromium-browser only.
  • magit-forge-el: FTBFS because it assumes it can query source package names with apt-cache showsrc in the build environment. Fixed to use dpkg -l instead and forwarded to Debian.
  • lomiri-url-dispatcher: build-depends on lomiri-app-launch, which was removed because it needs porting to current mir to be included in Ubuntu. Removed, and added to the sync-blacklist extra-removals.
  • libigloo: ftbfs on s390x, looks like code has endianness issues. Bug reported to Debian and update-excuse bug opened for tracking.
  • adwaita-qt: needed symbols file adjustments for ppc64el due to use of-O3. Uploaded, forwarded to Debian.
  • xtpcpp: xtpcpp-tandemwrapper binary package dropped, remove so this can migrate.
  • tcpreplay: Debian dropped the armhf binary, so followed their lead.
  • sklearn-pandas: one-line change to fix a test failure caused by a scipy namespace change. Uploaded and forwarded to the Debian bug.
  • rumur: FTBFS in Debian as well, and in the release pocket. No reverse-dependencies, so removed, leaving to the Debian maintainer to fix.
  • python-duniterpy: blocked because silkaj is incompatible with the new version. silkaj has been removed from Debian testing to unblock python-duniterpy, so removed from kinetic as well.
  • primus: stuck because it depends on bumblebee which is in multiverse. In Debian, bumblebee source is in universe and bumblebee-nvidia binary package is in contrib. Move these packages around in Ubuntu to match.
  • ocrad: autopkgtest regression, reported in Debian; removed package from
    -proposed.
  • node-wikibase-cli: ftbfs because testsuite requires Internet access. Patch to ignore testsuite failures and forward to Debian.
  • golang-github-sylabs-sif: autopkgtest regression in Debian and Ubuntu. Filed Debian bug and removed from -proposed.
  • golang-github-containers-storage: dep-wait on new golang-github-containerd-stargz-snapshotter, which in turn ftbfs because of a bug in our containerd package (referencing other go modules which don’t exist). Pinged @kanashiro about this, as our containerd source is packaged differently than Debian’s.
  • got a request for getting openexr built on i386; so added necessary bootstrap entries for openexr, cargo, and nvidia-graphics-drivers-tesla.
  • dtfabric: determined that the s390x failure is due to introduction of a new test that checks endianness handling, so this is not a regression in the package, just better tests. Marked badtest to let this migrate.
  • dropbear: evil autopkgtest that declares “breaks-testbed”, but just plain fails when trying to create ~/.ssh because the directory already exists. Uploaded fix to use mkdir -p; forwarded to Debian.
  • azure-devops-cli-extension: build-depends on azure-cli which was removed per Microsoft. Removed and added to sync blacklist.
  • node-configurable-http-proxy: ftbfs because missing build deps, fixed in Debian but not autosynced because of an Ubuntu delta. Merge the Debian version, it still fails. Not pursuing further at this time.
  • dislocker: patch the upstream build system to not install to the wrong target when the build environment is usrmerge’d. Forwarded to Debian.
  • flask-limiter: autopkgtest failure in Debian and Ubuntu blocking the package from being releasable. Removed from kinetic-proposed, leaving to Debian maintainer to fix.
  • greenbone-security-assistant: tries to download source from the network at build-time. Filed bug in Debian and removed.
  • nim: build-depends on libssl1.1. Removed.
  • python-aiortc: confirmed that this is an OOM kill of the autopkgtest on ppc64el. Added to big_packages and reran - passed.
  • heudiconv: appears to also be an OOM error, added to big_packages - passed.
  • autofdo: was an Ubuntu-specific build failure, now FTBFS in Debian also. No revdeps, so removed.
  • rust-uucore: depends on obsolete versions of rust-nix and rust-time; removed.
  • rust-fasteval: FTBFS in Debian and Ubuntu. Removed.
  • rust-average: depends on obsolete version of `rust-serde-big-array’; removed.
  • indi-pentax: Debian dropped arm64 as a supported arch, remove the binary.
  • qtmir: build-depends on lomiri-url-dispatcher which we just removed; remove this also and add to extra-removals.
  • python3.9: autosynced package that we don’t want. Removed.
  • kdevelop-python: hard-coded dep on python3.9. Removed.
  • shaarli: jbicha had reuploaded the php build-deps; let these through NEW to get the stack built again, but still FTBFS, so removed.
  • ploop: fix build failure caused by -Werror=format-truncation being overly aggressive.
  • lepton-eda: Debian dropped s390x; dropped as well for us.
  • jruby-openssl: depends on jruby which was removed in jammy. Remove.
  • removed various rust bindings that depend on non-existent packages; documented
    these in extra-removals in case they ever show up.
  • haskell-*: uploaded haskell-devscripts, then rebuilt the world.

NBS

  • fired off no-change rebuilds against libfoonathan-memory0.7.2, libfst22, libmeshoptimizer2d, libticalcs2-13, libpbbam2.1.0, libwebsockets17, libfmt9.
  • removed krosspython, only has reverse-recommends.
  • retried freecad autopkgtests with --all-proposed to see if they pass this way since they pass in Debian. Still failed.

SRU

SRU rotation on Friday, made more progress on the jammy backlog.

General

Uploaded openssh to enable systemd socket-based activation by default and started drafting documentation: SSHd now uses socket-based activation (Ubuntu 22.10 and later)

2 Likes
3
  • Worked on inventory of packages with configuration files in /etc
  • Proposed migration:
    • Scheduled all autopkgtests blocking python3-defaults and python3-stdlib-extensions with both packages as triggers
    • Hinted (and) packages blocked on a zfs-linux/s390x regression with the kernel from -proposed
    • Uploaded skimage fixing the FTBFS on ppc64el and s390x, allowing migration
  • 20.04.5 point release preparations:
  • Reviewed (LP: #1986664) hart-software-services for @xypron, accepted by @sil2100 (with minor change)
  • Sponsored mokutil sync for @xypron
  • Merge Party:
  • Sync’d/merged the following TIL universe packages: fpc, lazarus, scipy, box2d, dask, librandomx, openmsx
  • Sync’d pocl switching from LLVM 12 to LLVM 13, and requested removal of riscv64 binaries (LP: #1987621), which should unblock removal of llvm-toolchain-12
  • Travel booking
  • Candidate interview
2 Likes
4

Distro

Netplan

Misc

  • candidate take-home test
2 Likes
5

ubuntu-image

Go 1.19

  • Uploaded new version of go-gir-generator to fix FTBFS
  • Got golang-github-lucas-clemente-quic-go to migrate

Merge party

  • Merged busybox
  • Merged golang-1.18
  • Filed removal request for golang-1.17
1 Like
6

Short week due to PTO.

systemd

proposed-migration

  • siridb-server vs libuv1 (LP 1987558)
    • Unclear how libuv1 causes this, but I tracked the hang down to a NULL dereference in siridb-server and proposed a fix upstream: https://github.com/SiriDB/siridb-server/pull/182
    • The version of pcre2 in kinetic is newer than what siridb-server was last built with, and causes FTBFS on ppc64el. I believe I have a patch for pcre2 that will fix this.
  • nodejs vs libuv1
    • This issue has been discussed at length upstream https://github.com/nodejs/node/pull/42340. There is a patch for one of the tests, but the root cause of the second failure is still being discussed upstream.

Other

1 Like
7

Organisational

  • short week, had a 5 day weekend, only returned Tuesday :smiley:
  • booked trains
  • written interviews

APT

  • APT 1.6.17 in bionic unapproved, fixes regression in proposed 1.6.{15,16}

grub

  • Merged @mfo’s patch to support multiple initrd on xen
  • Merging @xypron’s typo fixes

Other Distro Stuff

  • Reviewed @schopin’s dh-cargo perl code for vendoring
  • Provided some assistance to @bdmurray for reproducing a release upgrader issue

QA

  • I quickly restarted cloud-worker-maintenance on lxd worker on Wednesday as it was stuck and left the QA team some guidance as to what should be improved there (timeouts when running lxc should be added)
1 Like
8
  • Became core-dev!
  • Clicked tests!
  • Specification work
  • subiquity
    • PR: 1386 - update existing PR for manual partitioning with ubuntu-desktop-installer for reviewer feedback and more test cases
    • collaborate with @chad.smith on PR: 1359 - something in the way cloud-init does schema validation breaks later users of jsonschema.
    • Review PR: 1387, 1388, 1391, 1392, 1393, 1394, 1395
  • curtin
    • MP: 428806 - nicer terminology where it doesn’t require API changes
  • casper
    • learned how to reconstruct a casper initrd given a valid one as a start point, testing kexec chainbooting as a strategy for network install improvements - kexec is just fine but needs more initrd magic.
  • ubuntu-cdimage:
    • MP: 428648 Fix simplestreams hash key collision when there were two types of ISO files, as in the case of netinst
1 Like
9
  • Rust: I rewrote my private dh-cargo code for vendored source tracking (the overall approach was broken if the crate had dev-dependencies), and uploaded it, with rustc and cargo waiting in the wings for uploads adding the XS-Vendored-Sources-Rust field to their source control file.
  • glibc:
    • looked into the llvm-toolchain-12 failure, ended up investigating removing the package altogether (cc @ginggs)
    • looking into backport lxc patches to fix a FTBFS against 2.36
  • Merged openssl 3.0.5-2 from Debian
  • Merged libvpx
  • Synced a bunch of universe packages where I was TIL for the OpenSSL 3.0 transition which has since reached Debian
  • Reviewed and sponsored s390-tools and p11-kit uploads for @Frank
  • Reviewed some apport patches
  • Travel planning
1 Like