Incremental Update — Buildfarm Preparation (May 16)
The thread advanced on the operational/infrastructure side, not on patch content.
Key Development: Buildfarm Client Updated
Andrew Dunstan asked whether the decision to bump to v10 was final so he could push a pre-prepared buildfarm client change. Tom Lane confirmed consensus ("I think we've agreed to do this for v20") while noting he hasn't reviewed the patch itself. Nathan +1'd.
Andrew then pushed the buildfarm client change:
- Commit: 4c3ed0c37bbcaf427e61384bb351364ecae9b4d9
- Immediate effect: Andrew's buildfarm animal
crakewill stop running cross-version upgrade tests against pre-v10 servers on HEAD immediately. - Broader rollout: A new buildfarm client release will happen when the v20 branch is cut, affecting all animals.
This is significant because it removes the last practical blocker Andrew had flagged earlier — buildfarm animals wasting cycles on cross-version tests against ancient servers that nobody meaningfully validates. The infrastructure is now aligned with the policy decision, ahead of the actual code commit.
No Patch Progress
No new patch versions were posted. The pg_restore archive-format sub-discussion has not advanced. The patches themselves (psql, pg_dump, pg_upgrade) remain in the state Nathan last posted (corrected v11 code retention, cleared multixact concerns).