carlos / platform Public

Clone
git clone https://amadan.net/carlos/platform

Plain git — no account needed to clone.

Download

Download this file

Astra: relocation × main merge check (round 6)

Round 5 approved relocation at f4395941. Before it could merge, main moved two merges: d6122533 moves-phase1-5f and 01edf901 feat/claim-admitted-provenance. They are merged into region-ha-1b-next-build at the commit under review (merge 87146236).

Conflicts and how they were resolved (see the merge commit message):

  • hostfiles/publisher.go: both sides kept. Main's new OpenAt now applies manifestWithinCap, as Open does.
  • hostfiles/prepared.go: Prepared.Stage and StageSkipped now hold withinTreeCap.
  • Census: relocation's Publisher.Commit digest was kept, and main's prepared.go rows were added.
  • Copy: relocation's row is renumbered MV-N450 → MV-N500, because moves reserved MV-N450 to MV-N499.

Judge only what the merge changes for relocation. Round 4 found that main's Staged launch field was dropped by relocation's own builder. Look for the same class of problem. Did main add a field, a guard, a phase, a journal key or a new contract that relocation's code (under internal/host/relocation*, internal/hostrelocate, internal/hostops completion and hostclaim relocating phases) must now carry, check or honour, and doesn't? Check in particular:

  • claimfiles.go, claimoriginproof.go, movesfiles.go, Manifest Origins/SetOrigins, OpenAt and NewFromCopy origins. Do relocation's files publication and restore need the origin set?
  • Changes to claimmoveparks, claimproduction, claimlaunch_production, claimstop_run, and to genselect (generation max).
  • Claim-admitted provenance: do relocation's admissions need to record the provenance main now requires?
  • Whether the conflict resolutions are correct.

Verdict: may relocation merge dormant at this commit? Give findings P1/P2/P3 with file:line. Report only new items.