Friday, 3 July 2020

+1 maintenance report - 29/06 -> 03/07


This is my first week on +1 maintenance duty, I decided to prioritize Ruby packages (addressed most of the issues) and also worked on some Go packages. Below you can find all the notes I took during the week (I also re-triggered some tests not mentioned here and I might have missed something on the way):

- golang-golang-x-tools: Marked LP #1884314 as Fix Released. The autopkgtest failure was fixed in Debian version 1:0.0~git20200410.79a7a31+ds-2 and it is already in the Groovy release pocket.

- puppetdb: Filed an RM bug against it (LP: #1885568). It was removed from Debian testing because it is stuck in unstable for one year. It never landed in the release pocket, so nothing to worry about reverse dependencies.

- puppet-beaker: Filed LP #1885614 and proposed a patch to Debian and upstream to relax version constraints on ruby-net-s{sh,cp}. Uploaded a new version fixing the issue.

- ruby-guard: ran autopkgtest triggered by ruby-guard/2.16.2-1 and pry/0.13.1-1

- pry: ran autopkgtest triggered by pry/0.13.1-1 and ruby-method-source/1.0.0-2

- ruby-hoe: Reopened the Debian bug #952041 reporting a FTBFS. After some analysis with another DD we decided to skip the failing test. Turned out it is not a bug in the upstream code (it affects downstream only), and the failure does not represent a real bug. Version 3.22.1+dfsg1-2 was uploaded to Debian skipping this test.

- ruby-webmock: Filed LP #1885760 and proposed a patch to fix autopkgtest in Ubuntu, it basically appends to no_proxy env variable because this address is used during the tests.

- devscripts: Filed a component-mismatch bug (LP #1885766) reported by Christian and submitted a patch to fix it, downgrading pristiner-tar (not in main) to Suggests.

- cadvisor: Uploaded debdiff attached on LP #1884663 after mwhudson ack the workaround I proposed. The containerd library package is not working at the moment due to the vendored code. We are working on it.

- ruby-mustermann-grape: ran autopkgtest triggered by ruby-mustermann/1.1.1-1 and ruby-mustermann-grape/1.0.1-1

- rubocop: ran autopkgtest triggered by ruby-parallel/1.19.1-1 and rubocop/0.80.1+dfsg-4

- puppet-beaker: ran autopkgtest triggered by ruby-net-scp/3.0.0-1 and puppet-beaker/4.21.0-2ubuntu1

- ruby-http: A test marked as flaky by upstream was failing in Ubuntu but not in Debian autopkgtest infrastructure. I uploaded version 4.4.1-4 to Debian skipping this test and also reported it to upstream.

- ruby-google-api-client: ran autopkgtest triggered by ruby-googleauth/0.13.0-2, ruby-signet/0.14.0-2 and ruby-google-api-client/0.33.0-2

- ruby-googleauth: ran autopkgtest triggered by ruby-signet/0.14.0-3 and ruby-googleauth/0.13.0-2

- ruby-signet: Uploaded version 0.14.0-3 to Debian skipping the failing test which checks for a $LOAD_PATH change which should not happen in the autopkgtest environment anyway.

- ruby-diaspora-federation: Uploaded version 0.2.6-2 to Debian relaxing the ruby-farday version constraint in gemspec. The package started to FTBFS after new ruby-faraday landed in Debian.

- bundler: ran autopkgtest triggered by ruby-net-http-persistent/3.1.0-2

- ruby-diaspora-federation: ran autopkgtest triggered by ruby-faraday/0.17.3-1

- golang-procfs: Sponsored the upload of version 0.1.3-3 to Debian which fixes the autopkgtest failure in armhf.

- ruby-valid-email: It FTBFS in Groovy because some tests rely on a DNS setup to perform some lookups. I filed a bug upstream to make tests more self contained and uploaded version 0.1.3-3 to Debian skipping those tests for now.

- ruby-arbre: It FTBFS because of some bundler usage in the test helper. I uploaded version 1.2.1-4 to Debian patching this out and closing bug #963379.

- ruby-ahoy-email: ran autopkgtest triggered by rails/2: and ruby-ahoy-email/1.1.0-1

- golang-github-miekg-dns: mwhudson's patch to fix tests in 32 bits platforms was accepted upstream and I uploaded version 1.1.26-2 to Debian applying this patch.

- puppet: Merged version 5.5.19-1 from Debian. It should fix the ruby-puppet-syntax FTBFS. The version 5.5.19-1ubuntu1 was uploaded after sorting out some mistakes made in 5.5.19-1.

- ruby-arbre: ran autopkgtest triggered by rails/2: and ruby-arbre/1.2.1-4

- ruby-hdfeos5: ran autopkgtest triggered by hdf-eos5/5.1.16.dfsg.1-3

- ruby-protocol-http1: Upstream has a couple of tests which checks for memory allocation space and they are failing in 32 bits platforms. I filed a bug upstream and uploaded version 0.13.0-3 to Debian temporarily disabling those tests.

Request for help:
During this week I've faced some autopkgtest failures with ruby-http and ruby-kubeclient in s390x architecture which at a first glance seems an outdated testbed (unsatisfiable dependencies) but after some retries I was not able to fix them, I'd appreciate if someone could help me to find the cause of the failures. FWIW they work fine in the other architectures.

Have a great weekend!