21.05 Retrospective

3

Forgot to kick off the discussion (this will later be used during the meeting, the discourse thread is largely for people to be come aware of the issues and start to formulate their own opinions):

What I liked:

  • RFC80 and RFC85 seemed to have reduced the risk going into the release signifcantly
  • Saw a lot more darwin build fixes for ZHF (thanks @stephank and others)
  • Native M1 Support
  • ZHF on master seemed to have worked really well (way less tedious than 20.09 and prior backporting)
  • Didn’t have to spend any time on plasma. (thanks @ttuegel)

What I feel neutral about:

  • No defined process for onboarding new platforms (e.g. aarch64-darwin) to hydra cc @grahamc @domenkozar
  • No defined support tier support for platforms, which further complicates the above issue. We have language to describe tier support from RFC46, but no current list of what platforms fall under which tier.

What I feel negative about:

  • Losing WorldOfPeace was big blow to the gnome team: @jtojnar had to pick up more slack. We should probably make a community effort to package and stabilize gnome for 21.11 and on.
  • x86_64-darwin hydra jobs were delaying the ability to do staging-next cycles in a timely fashion.
  • Could have had another week “post-branch-off” to further stabilize release notes, hydra builds (especially the stable staging-next), and do some more QA. (e.g. graphical iso was broken)
  • llvmPackages refactor (although well-intentioned) did cause 100+ downstream failures. I would like to add llvm to be added to release critical packages and for the definition of “breaking changes” to include restructuring outputs of the packages as well. Also, any changes made to llvmPackages is a darwin-stdenv-rebuild, which further complicated issues with x86_64-darwin stabilization.
  • I also didn’t make great use of release-engineers. I was also closing on a house, and didn’t have as much free-time so I prioritized reviewing PRs and solitary release work.
25 Likes