Skip to content

[Codex] Index pinned library branches without double-counting shared inputs #201

Open

Description

@guluchen

Engineering task under the active CSS/qLDPC coverage goal. GH200source-dimension decision remains pending; no new source-dependent scientific batch will advance here.

Build a read-only, deterministic cross-branch inventory from current main and all38library work-branch heads, pinned to full commits. Group repeated stems/file identities rather than summing per-branch manifests. Preserve every differing manifest record and source claim, isolate quarantined files from accepted flat manifests, flag same-(n,k) review groups without declaring physical equivalence, and link retained server queues without treating historical/optional plans as new required execution. Use Git object metadata and manifests only; no solver or broad matrix computation.

Outputs: exact refs, reproducible inventory script, JSON and concise Markdown index, fixture tests for missing files/unregistered quarantine/conflicting versions/repeated branches/job unknown limits. Counts describe stored inputs/versions, never independent families, accepted scientific proofs or globally open distances. No matrix/manifest/solver/build/default/tier/QDistSAT/main integration, no external communication or server allocation. Read existing evidence; do not overwrite or synthesize raw bound fields. Branch codex/library-coverage-index from current main6083e662. Unit-only validation and ordinary CI; no scientific Actions runs. Broader goal remains active.

Activity

  1. guluchen commented on Oct 29, 2026

    @guluchen
    OwnerAuthor

    Inventory checkpoint pushed: data6aa4c771461a03984ee6e6349fd37b820857cb1a, source/workflow38954311a4d830648c54ff0d439e4800d071c1b0, tree3c4f59a7c8a422c62e8bb79513553d988f7f7bd8. Local19tests and full39ref byte-identical snapshot replay PASS. 1068manifest memberships group to192stored stems/192four-file versions;198distinct raw metadata records preserved,876repeated memberships. Main unchanged51tiered+24untiered,5GH200quarantine candidates separate; no physical-inequivalence/family/proof/adoption count inferred. Same(n,k)review groups include main tiers.

    33canonical queue versions linked with source/runner identities;17other historical job-file versions retained without interpreting them. Unknown limits remain null, no workload total or execution assumptions. No matrix/manifest/solver/build changes, new scientific inputs or server execution. GH200decision still pending and that batch does not advance. Linux contract run40954290453 and ordinaryCI40954290324 started; contract tests cover19fixtures+real pinned-main metadata, not full39branch Git replay (that was local). Goal remains ACTIVE.

  2. guluchen commented on Oct 29, 2026

    @guluchen
    OwnerAuthor

    Terminal engineering evidence retained inbbb684ead66fc64ac1ed137780ca3581945b68f5. Linux contract40954290453/job123730222023 and ordinaryCI40954290324/job123730220643 both SUCCESS. Full Linux contract log/SHA256receipt committed;19unit contracts cover fixtures+real main metadata. Full39ref snapshot byte replay was local. 1068memberships ->192stored stems with198raw metadata variants and876repeated memberships;5quarantined GH200candidates excluded. Main remains51tiered+24untiered;192is not a family/equivalence/proof/adoption count. No matrix/manifest/solver/build/main/QDistSAT/source-hold change, no server execution. Source-dependent GH200 remains stopped pending PI; active full coverage goal not complete.

  3. guluchen commented on Oct 29, 2026

    @guluchen
    OwnerAuthor

    Final pushed packagebbb684ea ordinaryCI40954546269/job123731110129 terminal SUCCESS. Worktree clean. Metadata-only final evidence commit did not rerun the inventory-contract workflow; no new scientific workload. Engineering task complete, full coverage goal ACTIVE and source-dependent GH200PI decision still pending.

  4. guluchen commented on Oct 30, 2026

    @guluchen
    OwnerAuthor

    PI decisions recorded from the current task on 2026-10-30 (first six consolidated options: A):

    Main and all scientific raw evidence remain unchanged. Routine index refresh/deduplication/documentation can proceed independently; the older 39-ref inventory is a historical snapshot, not a current count or a list of all unfinished jobs.

  5. guluchen commented on Oct 30, 2026

    @guluchen
    OwnerAuthor

    All fifteen PR-triggered checks on the five authorized library PRs completed SUCCESS. These are ordinary CI, existing bounded HGP smoke, and the short QDistSAT cross-repo benchmark; no new family-specific scientific workflow or server continuation was launched.

    PR Check Run
    #210 CI 41045225717
    #210 HGP library smoke 41045225832
    #210 QDistSAT cross-repo benchmark 41045225640
    #211 CI 41045233068
    #211 HGP library smoke 41045232809
    #211 QDistSAT cross-repo benchmark 41045233016
    #212 CI 41045241503
    #212 HGP library smoke 41045241455
    #212 QDistSAT cross-repo benchmark 41045241428
    #213 CI 41045249948
    #213 HGP library smoke 41045249976
    #213 QDistSAT cross-repo benchmark 41045250003
    #214 CI 41045257402
    #214 HGP library smoke 41045257487
    #214 QDistSAT cross-repo benchmark 41045257375

    No merge, default/stem/tier change or source-hold release. Main remains 6083e66. Engine decision item seven remains pending the PI's response after explanation.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions