Research consolidation: reconcile selected #626 Linux replay evidence

docs/ci-verification.md

@@ -2,6 +2,16 @@
 
 ## Current selected-family replay
 
+The selected #543 + #626 composition uses `.github/workflows/frontier626.yml`.
+Its [completed Linux run](https://github.com/CrocSwap/integer-mult-bounds/actions/runs/38317651752)
+passed on 2026-10-13 at commit `ed49217ad6244f1d3bfe4588f122fa861849739c`.
+The [preserved artifact reconciliation](../research/selected626-linux-reconciliation/README.md)
+checks source equality to the sprint base, all six replay stages and exact
+mathematical assembly agreement, with explicit limits for omitted payloads.
+The older queued integration receipt is retained as historical evidence.
+
+## Historical selected-family replay through #588
+
 `.github/workflows/frontier588.yml` provides one complete cold check of #543
 with #588 on Ubuntu 24.04 and Python 3.12. It installs NumPy, SymPy, SciPy,
 C++17 and Boost, checks the selected record and review evidence, rebuilds both

docs/research/current-status.md

@@ -5,6 +5,12 @@ See the [review](community-k492-review.md), [integration receipt](community-k492
 [selected record](../../certificates/selected-result.json) and [PR ledger](pr-ledger.md).
 Public cold replay: `make frontier626-verify`.
 
+The previously queued #626 Linux run completed successfully on 2026-10-13.
+The [preserved evidence and offline checker](../../research/selected626-linux-reconciliation/README.md)
+bind that run to its tested sources, six replay stages and exact assembly.
+Three generated provenance digests differ from the committed assembly; the
+linked reconciliation records their scope and the unavailable payloads.
+
 The frozen round covers #618–#626. Eight scoped reviews complete; #620 remains
 partial pending portable research replays and validation of its solver/barrier
 claims. Earlier open reviews retain their ledger status. Inherited uniform,

research/selected626-linux-reconciliation/CLAIM.md

@@ -0,0 +1,38 @@
+# Sprint claim
+
+- Track: research consolidation; OS-random assignment seed **95** (`secrets.randbelow(100)`).
+- Exact base: `66849e1475c930fd7732b1a0067ffee7e1e95f60`.
+- Object: the pending/queued Linux status in the selected #626 integration receipt,
+  for run `38317651752` at `ed49217ad6244f1d3bfe4588f122fa861849739c`.
+- Method: inspect that exact completed run and its artifact; check all orchestrator
+  stages, log digests, exact composition and tested-source equality to the base.
+- Deliverable: a preserved, offline-checkable provenance receipt with rejection
+  controls and a narrowly scoped current-status correction. No new kappa claim.
+- Claim time: 2026-10-13 04:08 UTC. Expected finish: 04:30 UTC; final quarter
+  reserved for verification and handoff.
+- Coordination: GitHub's pinned-issue list is empty, Discussions is disabled,
+  and the only open non-PR issue is #650 asking for Discussions. No reservation
+  issue was opened. Open PR titles and #675's body were checked for overlap.
+- Candidates considered: (1) selected source-manifest completeness;
+  (2) #620 partial-review status versus predecessor attribution;
+  (3) queued-versus-completed #626 Linux evidence. Seed `95 % 3 = 2` selects
+  the third candidate after the overlap screen. The first two already have
+  explicit coverage in the selected checker/review; the third has a concrete
+  newly completed result to reconcile.
+
+Prepared by Josh Snider (`jsnider3`) with substantial OpenAI Codex assistance.
+Apache-2.0. This claim is local because no sprint coordination issue exists.
+
+## Result — 2026-10-13 04:21 UTC
+
+Submitted [PR #695](https://github.com/CrocSwap/integer-mult-bounds/pull/695):
+preserved the completed Linux evidence and checked 39 source objects, six replay
+stages, 64 complex receipt substeps and 47 positive assembly margins. All 16
+rejection controls passed, including normal/optimized isolated runs from a
+separate committed-files-only worktree. Selected-record and ledger checks passed.
+The three differing generated provenance digests and unavailable CI payloads
+are documented explicitly; no new kappa or full theorem is claimed.
+
+Final overlap check through #693 found no matching run reconciliation; upstream
+main remained at the recorded base. No pinned sprint issue was available, so
+this is the single local claim update with the result and handoff link.

research/selected626-linux-reconciliation/EXPECTED.json

@@ -0,0 +1,114 @@
+{
+  "evidence_sha256": {
+    "artifact.zip": "1f1ad114e311f75d25fd6500db7ef0b8e15a4814aebc5c1edb01ec9288480355",
+    "artifacts.json": "d70261bfc7af4a9129ed5e2528fb039c72099110604d5ffa1ccb0832f070f379",
+    "jobs.json": "8b18b198ff47192f2e1bc76a79f14915440ae6e56bc5ad9f0833418e949e85c6",
+    "run.json": "6f963a2553e3e429da119f6b84294f90c565f592166f79cf0a63f3aa5659f626",
+    "workflow.log.gz": "0002ce6ad55392dcd4c4ac3f96635affff4169dfa0e5367a2d43c350fb8bb2c9"
+  },
+  "workflow_steps": [
+    "Set up job",
+    "Run actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1",
+    "Run actions/setup-python@5fda3b95a4ea91299a34e894583c3862153e4b97",
+    "Install verification dependencies",
+    "Check selected record and review evidence",
+    "Cold finite replay and exact composition binding",
+    "Require committed files unchanged",
+    "Run actions/upload-artifact@v4",
+    "Post Run actions/setup-python@5fda3b95a4ea91299a34e894583c3862153e4b97",
+    "Post Run actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1",
+    "Complete job"
+  ],
+  "complex_steps": [
+    "actual-recompression-and-rebound-controls",
+    "admission-controls",
+    "bind-providers",
+    "ck-all-columns-native-frames",
+    "ck-all-phase-edges",
+    "ck-both-reflected-ledgers",
+    "ck-independent-local-controls",
+    "ck-normalized-wrappers",
+    "ck-step0-port_stage-construction",
+    "ck-step0-port_stage-native-audit",
+    "ck-step1-port_stage-construction",
+    "ck-step1-port_stage-native-audit",
+    "ck-step2-port_stage-construction",
+    "ck-step2-port_stage-native-audit",
+    "ck-step3-port_stage-construction",
+    "ck-step3-port_stage-native-audit",
+    "ck-step4-port_stage-construction",
+    "ck-step4-port_stage-native-audit",
+    "ck-step5-frame_regroup-construction",
+    "ck-step6-hyperplane_end-construction",
+    "ck-step6-hyperplane_end-controls",
+    "ck-step7-cold_end-construction",
+    "ck-step7-cold_end-controls",
+    "ck_504-step0-greedy_kernels-construction",
+    "ck_504-step0-greedy_kernels-omissions",
+    "ck_504-step1-frame_coalesce-construction",
+    "ck_504-step2-gate_move_coalesce-construction",
+    "conservative-and-native-exact-assembly",
+    "emit-controller",
+    "fresh-invoice",
+    "frozen558-full-replay",
+    "independent-controller-and-controls",
+    "jp8-all-columns-native-frames",
+    "jp8-all-phase-edges",
+    "jp8-base-k_triangle-construction",
+    "jp8-both-reflected-ledgers",
+    "jp8-independent-local-controls",
+    "jp8-normalized-wrappers",
+    "jp8-step0-port_stage-construction",
+    "jp8-step0-port_stage-native-audit",
+    "jp8-step1-port_stage-construction",
+    "jp8-step1-port_stage-native-audit",
+    "jp8-step2-port_stage-construction",
+    "jp8-step2-port_stage-native-audit",
+    "jp8-step3-port_stage-construction",
+    "jp8-step3-port_stage-native-audit",
+    "jp8-step4-frame_regroup-construction",
+    "jp8-step5-hyperplane_end-construction",
+    "jp8-step5-hyperplane_end-controls",
+    "jp8-step6-cold_end-construction",
+    "jp8-step6-cold_end-controls",
+    "jp8_504-step0-cold_kernel-construction",
+    "jp8_504-step0-cold_kernel-omissions",
+    "jp8_504-step1-multicut-construction",
+    "jp8_504-step1-multicut-omissions",
+    "jp8_504-step2-frozen558_scalar_audits-rank5",
+    "jp8_504-step2-frozen558_scalar_audits-static44",
+    "jp8_504-step3-joint_prefix-all-kernel-omissions",
+    "jp8_504-step3-joint_prefix-construction",
+    "jp8_504-step3-joint_prefix-independent-joint-prefix-and-coverage",
+    "jp8_504-step3-joint_prefix-rank5-omissions",
+    "k-triangle-control-empty-schedule",
+    "k-triangle-control-pr531-construct504",
+    "phase-timing-portability-controls"
+  ],
+  "assembly_digest_differences": [
+    {
+      "path": [
+        "template_sha256"
+      ],
+      "linux": "8aac9a65fe3b15889a0cdbc125bd80ab0e141578efccd25a7d338fd5ee3aae6e",
+      "committed": "22b41c45fdac1dcc7495311fe701f4697219654843558785ed72fae91df531b0"
+    },
+    {
+      "path": [
+        "input_sha256",
+        "owner_audit"
+      ],
+      "linux": "1de45f6b911f8ef6f9c134a1eaf1e1f03feac808e048f896c1099b121c061ce9",
+      "committed": "ec7bc4157e7ab1707bddf99994149c932b51365de7e742508fd182051ca57fc1"
+    },
+    {
+      "path": [
+        "input_sha256",
+        "fresh_invoice"
+      ],
+      "linux": "2219b0c60373d93f256ff255ce6de2af1605b4c61c18f45cb189f788c5bb5d96",
+      "committed": "367a73f744ea9b7dd6cca5fa3fda47175f560aa3d37e708c9fa1294dbd4b56eb"
+    }
+  ],
+  "workflow_log_sha256": "fbdc929e9bbf4ba9afb050b4c74f51322499dd388be3c861ff51a5754b39ebba"
+}

research/selected626-linux-reconciliation/README.md

@@ -0,0 +1,126 @@
+# Selected #626 Linux evidence reconciliation
+
+The published #626 integration receipt records run **38317651752** as queued.
+That exact run completed successfully on **2026-10-13 at 02:10:06 UTC**.
+This package preserves its artifact and checks the evidence linking that
+execution to the selected #543 + #626 composition. Status: **research
+consolidation / provenance verification; no new kappa**.
+
+The conditional result remains exactly
+`51110230636731/31250000000000000 = 0.001635527380375392`.
+The earlier queued receipt remains an unchanged historical snapshot.
+
+## Reproduce
+
+From a clone containing the two commits below, using Python 3.11+ and Git:
+
+```sh
+python3 -I -B research/selected626-linux-reconciliation/verify.py
+```
+
+Expected:
+
+```text
+PASS_PR626_LINUX_RECONCILIATION 6 stages, 64 substeps, 47 strict margins, 16 rejection controls
+```
+
+No network, third-party Python package, compilation, or supplier rerun is needed.
+`--out /tmp/pr626-linux-report.json` writes the deterministic report.
+The same command also works with `-O`; rejection checks use explicit exceptions.
+A shallow clone must fetch the pinned commits first. This is an evidence check,
+not a fresh execution of the supplier pipeline; that remains `make frontier626-verify`.
+
+## Exact evidence and source scope
+
+| Object | Pin |
+|---|---|
+| Sprint base | `66849e1475c930fd7732b1a0067ffee7e1e95f60` |
+| Tested CI commit | `ed49217ad6244f1d3bfe4588f122fa861849739c` |
+| Run, attempt | [38317651752](https://github.com/CrocSwap/integer-mult-bounds/actions/runs/38317651752), attempt 1 |
+| Job | [115038013572](https://github.com/CrocSwap/integer-mult-bounds/actions/runs/38317651752/job/115038013572) |
+| Artifact | `11784169747`, `frontier626-linux-python312` |
+| Original ZIP SHA-256 | `1f1ad114e311f75d25fd6500db7ef0b8e15a4814aebc5c1edb01ec9288480355` |
+| Selected #626 source | `660de893d27bb7abd1d5b01b1640207528a5c851` |
+| Retained #543 source | `2065fe4f72a914d907f1046603f5a36f17018535` |
+
+`evidence/` retains the original ZIP, run metadata (selected API fields), full
+job/artifact metadata, and losslessly compressed `gh run view --log` output.
+`EXPECTED.json` also pins the decompressed log bytes. It pins those evidence
+bytes, all workflow step names, all 64 complex receipt substeps, and the three
+assembly digest differences. The ZIP hash matches GitHub's artifact metadata;
+its bytes are an immutable downloaded input, not a regenerated container.
+
+`SOURCE-PINS.json` records Git object IDs. The checker independently queries
+both commits and requires 39 identical objects for every selected package root,
+all proof supplements, the complete `scripts/` tree, the workflow, Makefile,
+selected record, source manifest, and selected assembly. It derives this coverage
+from the selected record rather than accepting an arbitrary shortened list.
+It also binds the complex receipt's package/spec and six dependency-manifest
+SHA-256 values to the base's bytes. The intervening publication/ledger edits
+are outside this source-equivalence claim.
+
+The archived run reports Ubuntu 24.04.5, image `20261004.327.1`, Python 3.12.15,
+NumPy 2.5.4, SymPy 1.14.0, SciPy 1.18.1 and mpmath 1.3.0. The workflow installs
+dependencies without version pins; these are observed versions from the log,
+not a claim that a future installation will reproduce that environment.
+
+## What agrees, and what remains unavailable
+
+All six orchestrator stages have exit code zero and retained logs matching
+their SHA-256 values: bit543, declarations, complex626, independent-owner,
+helper-interface, and bindings. The complex receipt lists 64 successful
+substeps; its result explicitly reports no baseline or local-output reuse.
+The composition links to the exact archived complex receipt and assembly.
+
+Every assembly JSON field agrees with the committed certificate except these
+three explicitly recorded paths:
+
+- `template_sha256`
+- `input_sha256.owner_audit`
+- `input_sha256.fresh_invoice`
+
+In particular, the paid profiles, moment intervals, source-engine hashes,
+supplier endpoints, bootstrap data, controls, and all 47 strict outer margins
+agree exactly. The checker additionally evaluates all 47 rational margins as
+positive. The two source-engine/margin mutation controls update receipt links
+consistently, so rejection tests mathematical-field comparison rather than
+merely noticing a stale file hash.
+
+The workflow did **not** upload the generated template, owner audit, invoice,
+all 64 substep logs, or separate `OWNER-CHECK.json`/`INTERFACE.json`. Their
+execution is supported by the retained stage logs and receipts, but this package
+cannot independently compare those omitted payloads or explain the three byte
+digest differences. It does not assert that those differences are only gzip
+metadata. The source already includes recompression controls; their recorded
+success is evidence of execution, not a replacement for the missing payloads.
+
+Negative controls reject wrong heads/attempts, queued runs, skipped workflow
+steps, detached artifacts, missing/failed replay stages, changed logs, broken
+receipt links, wrong kappa, changed positive margins, altered engine hashes,
+unrecorded provenance differences, and changed tested source objects.
+
+This is an offline consistency check of preserved GitHub evidence, not a signed
+attestation or an independent mathematical reimplementation. It confirms one
+completed run at a pinned source state. It does not establish other platforms,
+newer PRs, optimality, or a full multiplication theorem. The selected review's
+all-size compiler, common-label/mutable-basis, pooled hosting, residual,
+precision/tape/recovery and analytic assumptions remain.
+
+## Attribution and sprint record
+
+Douglas Colkitt supplied the selected review, replay orchestration and CI.
+Chafik Boukhalfa supplied #626's owner repair; Jeremy Kothe supplied #622/#623's
+construction; eumemic supplied #543's bit supplier. GamingPuzzled's #620 lead
+and the preceding supplier, recurrence and verification contributors retain
+the attribution in the selected review and package notices. PR #675's
+independent owner-boundary audit is distinct from this CI provenance check.
+
+Consolidation by Josh Snider (`jsnider3`) with substantial OpenAI Codex assistance.
+Apache-2.0; this is not independent human peer review. [CLAIM.md](CLAIM.md)
+records OS-random seed 95, candidates, scope and coordination checks.
+
+Local validation: normal and optimized isolated Python runs produced identical
+reports; all 16 rejection controls passed. `make selected-record-check`,
+`python3 -B scripts/verify_pr_reconciliation.py`, and `git diff --check` passed.
+The selected suppliers were not rebuilt locally because this change reconciles
+the already completed, preserved CI execution and changes no construction.

research/selected626-linux-reconciliation/RESULT.json

@@ -0,0 +1,54 @@
+{
+  "status": "PASS_PR626_LINUX_RECONCILIATION",
+  "run": 38317651752,
+  "ci_head": "ed49217ad6244f1d3bfe4588f122fa861849739c",
+  "base": "66849e1475c930fd7732b1a0067ffee7e1e95f60",
+  "kappa": "51110230636731/31250000000000000",
+  "replay_stages": 6,
+  "complex_substeps": 64,
+  "strict_assembly_margins": 47,
+  "assembly_digest_differences": [
+    {
+      "path": [
+        "template_sha256"
+      ],
+      "linux": "8aac9a65fe3b15889a0cdbc125bd80ab0e141578efccd25a7d338fd5ee3aae6e",
+      "committed": "22b41c45fdac1dcc7495311fe701f4697219654843558785ed72fae91df531b0"
+    },
+    {
+      "path": [
+        "input_sha256",
+        "owner_audit"
+      ],
+      "linux": "1de45f6b911f8ef6f9c134a1eaf1e1f03feac808e048f896c1099b121c061ce9",
+      "committed": "ec7bc4157e7ab1707bddf99994149c932b51365de7e742508fd182051ca57fc1"
+    },
+    {
+      "path": [
+        "input_sha256",
+        "fresh_invoice"
+      ],
+      "linux": "2219b0c60373d93f256ff255ce6de2af1605b4c61c18f45cb189f788c5bb5d96",
+      "committed": "367a73f744ea9b7dd6cca5fa3fda47175f560aa3d37e708c9fa1294dbd4b56eb"
+    }
+  ],
+  "source_objects_matched": 39,
+  "negative_controls": [
+    "wrong_ci_head",
+    "wrong_attempt",
+    "queued_run",
+    "skipped_workflow_step",
+    "artifact_other_run",
+    "artifact_bad_digest",
+    "missing_helper_log",
+    "corrupt_owner_log",
+    "missing_helper_stage",
+    "failed_bit_stage",
+    "wrong_kappa",
+    "broken_receipt_link",
+    "changed_positive_margin",
+    "changed_source_engine_digest",
+    "changed_recorded_digest_difference",
+    "changed_tested_source"
+  ]
+}