Welcome to Week 17 of the DevOps Log.
This week I continued to reduce spurious failures in GHC CI.
As it turned out, there were legitimate failures this week due to CI’s incompatibility with MacOS Monterey. @angerman is tackling this problem in #22286. There may also be issues with GHC itself on Monterey (#22295), but let’s remove the confounding factors before we jump to conclusions.
In other work, I marked one GHC test as fragile and successfully split some fragile tests out of the head.hackage pipeline. Naturally, I committed the sin of making an unrelated “upgrade” while changing the head.hackage pipeline, which meant I got to fix broken CI a day later.
- We want to begin notarizing GHC packages for MacOS for the 9.6 release. I will be working with @angerman on that one as well.
- Some infrastructure for our GitLab instance need to be migrated to new servers before November 30. I am waiting on having the new machine available, and then I’ll start working on that.
- If I find myself blocked on those two tasks, I will continue looking at reported spurious failures and attempting to mitigate them. But the priority is 1 and 2, since they have hard deadlines.
Starting next week, I am shifting the timing of these updates to Wednesday. There are two reasons. The first is that the GHC team already has a check-in on Wednesday, so it will reduce my overhead to review my notes once a week rather than twice. The second is that Haskell Weekly often kindly links to these updates of mine, but since it’s released on Thursday, my updates are either a week old or one day too late, depending on how you look at it.