carlos / platform Public

plan: claim Plan 7b rev 3, the run walk across regions (joint round 13 R13-1)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

claude 2 weeks ago c3ea8fbd7d6605944f079ef09421dfd5256da31b parent 0bf2f49

1 file changed, +13 −0

docs/superpowers/plans/2026-10-05-host-lifecycle-claim-7b-genselect.md +13 −0

CI cancelled — superseded by a newer push — run details

diff --git a/docs/superpowers/plans/2026-10-05-host-lifecycle-claim-7b-genselect.md b/docs/superpowers/plans/2026-10-05-host-lifecycle-claim-7b-genselect.md

index c83bc0299..60937523c 100644

--- a/docs/superpowers/plans/2026-10-05-host-lifecycle-claim-7b-genselect.md

+++ b/docs/superpowers/plans/2026-10-05-host-lifecycle-claim-7b-genselect.md

@@ -1329,3 +1329,16 @@ The embedded code was compiled using a Go overlay, with Plan 1's actual `hostcla

Rev 2 was recompiled and the Linux test binary passed **76 table cases**, including **17 prepared-restore cases** and the six direct comparisons with `CalcRestorePlan`. The prepared cases cover absent target markers with covered or trimmed source admissions, ordinary disagreement refusals, empty generations, missing/mismatched pins, marker isolation and same-generation successor advancement. Replacing overlap with exact adjacency made both the model overlap port and the library-comparison overlap case fail, as intended. The unmodified code remains in this plan. The final source also passed package builds for **Darwin/arm64**, **Windows/amd64** and **Linux/arm64**. All three embedded Go blocks were checked byte-for-byte against the formatted source used for those checks.

Paseo could not connect to its daemon during rev 2 verification, so an Opus review was unavailable. The tables above record the source-based self-review; an Opus review remains part of Task 4 before an implementation is landed.

+## Rev 3: joint round 13 (R13-1)

+Round 13 found that Select rejected a valid interrupted clean-return recovery. Claim spec rev 6.22 (§6.2a "Walking the run in one region's bucket", §6.4a operative item 3) now states the walk, and the implementation on `feat/hostclaim-genselect` (d7621a29) follows it:

+- the walk follows each entry's `predecessor` back from `holder.generation` to the marker's generation;

+- a link whose generation keeps its objects in another region is passed over, covered or not;

+- a link in this region must be uncovered; a covered local generation reached as a needed baseline refuses with a distinct error and is never selected;

+- the baseline is the newest generation in the run holding an object in this region, else the marker's generation;

+- the overbroad covered-chain and foreign-baseline refusals are removed; every other refusal stands.

+Table cases named after R13-1: the interrupted clean return (S in IE, T in US, fresh R in IE born covered, no objects: restores S); another interrupted admission after it (still S); a safe covered foreign link (passed over); an unsafe covered local baseline (refuses). All 14 regressions failed before the change and pass after it.