Repository navigation
Audit the hosting boundary of PR1817’s retimed rank-one word - #1833
Conversation
|
Batch review at Merged to Reviewed scope: Exact original-rank retiming preserves all per-role occurrence traces and complete phase boundaries. Independent144-call geometry checks69120 reservations and exported charts/windows; all144 program hashes require new admission bindings. Does not realize rank16 batching or replay native/compiler obligations. The pinned scoped command passed; the batch ledger contains its command, input hashes, environment and log. This incorporates the research/tooling contribution and does not change the selected κ or discharge the inherited multiplication-theorem assumptions. Batch review, attribution and evidence. Thank you for the contribution. |
2 participants
Change and mathematical scope
This independently checks one explicit gap in PR1817: its exact within-phase retiming preserves the selected coarse whole-call hosting projection, under the retained completed-primitive workspace contract. Both 24,767-event orientations preserve every shared-role occurrence order and all ten phase cuts. Independent reconstruction verifies all 144 calls, 69,120 helper reservations, 288 strict busy windows, 10,002 orthogonal 73-axis charts and 7,526 idle suffixes.
All original rank-one children remain. The literal local/global frame profiles are unchanged (global rank mass 1,041,408; deficit 5,120). No rank-16 replacement,5% construction-gate success, supplier admission or new kappa is claimed. All144 frozen call bindings name the old program hashes: they cannot silently admit the retimed streams. New native/phase, precision/tape and controller/invoice evidence remains necessary, through fresh replay or explicitly justified source-bound reuse.
The written proof uses shared-role dependency order and constant injective whole-invocation owners. Its physical application assumes complete primitives restore shared implementation workspace and do not depend on event position. The checker explicitly validates private namespace disjointness; this does not prove the native scratch contract itself.
Task and overlap
Base current main: 0b1cccf. Revision 3 card complex-rank-shape-decoder; OS nonce 48d7645b26b9f89e9337e038a86403f8; scope 02b31ebd160542c86299/PR1817/whole-call-owner-window-invariance. Scope is changed from physical block-decoder synthesis to this missing boundary audit; the construction acceptance gate is not claimed satisfied.
PR1817 head ac3a759 supplies the unmodified retiming generator and the count of 134 groups. PR1779 supplies the diagonal premise, and PR1605 supplies price budgets. Exact heads and source/blob/SHA256 pins are recorded in TASK/SOURCES. Only research/diagonal-retiming-host-boundary/ is added; the vendored generator retains its original bytes and notice.
Claim/handoff. Final overlap covered 489 open PR bodies including drafts and 873 local notes. Relevant live Discussion threads were read directly; there is no complete Discussion census, and numbered PR pages are non-atomic. Main and predecessor heads were rechecked.
During the session Ross Richardson completed a separate native-contract audit, applying Greer Guthrie’s PR679 support/volume obstruction to the monomial tensor-child proposal. It is read and credited, not replayed here. Our original-rank boundary projection remains distinct from that obstruction and from an exact phase-B batch optimum.
Validation
Python 3.12.14, macOS arm64: registered replay passes in 7.59 s. Nine semantic/geometric mutations reject. Three oracle tests classify all 720 permutations of six mixed scalar operations and check every accepted ordering on all 625 dirty F5 inputs; occurrence duplication and unknown-event controls reject. These finite tests corroborate the written implication, not the entire native compiler.
No-Git exports pass normally and under -O. Actual input, predecessor-code and receipt corruption reject; an occupied output is preserved. Metadata/source pins and staged whitespace checks pass. VALIDATION.json records the initially rejected numeric-key serialization comparison and its repair; no mathematical claim was changed.
Not run: full frontier1094 cold replay, new native/phase/precision/controller invoice regeneration, actual diagonal-child construction, Ross/PR679 replay, Lean or independent human review. Selected record and frozen inputs are unchanged.
Credit
Credit Josie Elliston and her disclosed Claude assistance (PR1817); Jose Manuel/jmbarrancoml and publishing account Cinna Davis/1337isnot1337 (PR1779 source attribution); Arham Tahir (PR1605); Douglas Colkitt/PR1240; Sagar Thakrar/PR1012; the PR442/645/858 hosting/controller lineage; Jeremy Kothe, Hao Yan, Rohan Arun/PR1094, icekylinx and retained contributors. Credit Ross Richardson and Greer Guthrie/PR679 for the distinct native-contract finding. Prepared by @ToikY with substantial OpenAI Codex assistance, including bounded geometry implementation, overlap review and separate static review. Apache-2.0; original notice retained.
Open next questions
See the exact-head README. Before a diagonal implementation, name the PR679 hypothesis that changes and give the complete dirty-input contract and cost. For an original-rank retimed supplier, supply the strict new-hash native/phase and precision/controller evidence. No merge or frontier update is requested.