Skip to content

CF-06: qualify upstream comparator source after HL7 fix #15

Open

Description

@TheHalfMoon

Decision boundary

This issue records the next CF-06 qualification gate only. It does not authorize a production oracle pin change, a semantic reinterpretation of oracle failures, a commandF PR #11 merge, or CF-12.

Current production contract

CF-06 remains pinned to:

project: hapifhir/org.hl7.fhir.core
release: 6.10.2
source: d06577dbc5c62c74a2a8823fbc4830a3024d5b0b
validator_cli.jar sha256: a3addadfa18dfa23146a0a243b6ede68eaad92157a5407738c468bb3d7e4ccd6

GitHub releases/latest still reports 6.10.2 as the latest official release as of 2026-08-16.

Upstream defect/fix evidence

Upstream issue/PR:

hapifhir/org.hl7.fhir.core#2553
hapifhir/org.hl7.fhir.core#2554

PR #2554 fixes two independent DefinitionNavigator defects:

  1. inherited differential slices whose slicing master exists in the snapshot but is omitted locally from the differential;
  2. contentReference traversal incorrectly treating same-path named slices of the referenced master as children.

The commandF qualification branch has already proven:

C001 isolated real probes: 6/6 PASS
C002 isolated real probes: 6/6 PASS
combined real probes:      12/12 PASS

The full frozen CF-10 A/B patched-corpus proof is run 31918255417 and remains the required evidence gate before any later source qualification decision.

Required qualification sequence

If and only if the patched full-corpus proof completes cleanly:

  1. retain the current 6.10.2 production pin unchanged;
  2. wait for upstream #2554 to be reviewed/merged or otherwise obtain an upstream-authoritative source identity;
  3. record the exact upstream merge/source SHA;
  4. qualify that exact upstream source against the full CF-06 regression surface;
  5. rerun frozen CF-10 A/B with byte-identical deterministic summaries and all three cases operationally complete;
  6. verify CF-01..CF-09 and security/process regressions remain green;
  7. review the resulting evidence independently;
  8. require a separate founder decision before changing the production CF-06 release/source identity.

Fail-closed rules

  • CF-03 remains commandF semantic authority.
  • HL7 remains an independent advisory oracle.
  • comparator exceptions remain operational failures under the current production contract.
  • no exception may be silently converted to agreement or uncomparable.
  • no frozen CF-10 case may be removed or altered to improve results.
  • an unmerged fork commit must not become the production oracle pin solely because it passes commandF qualification.

Exit condition

This issue is ready for a production-pin decision only when an upstream-authoritative source identity has passed the complete qualification sequence above with retained evidence.

Activity

  1. TheHalfMoon commented on Aug 16, 2026

    @TheHalfMoon
    OwnerAuthor

    Patched full frozen-corpus qualification is now complete and green.

    Evidence:

    workflow: hl7-upstream-fix-qualification
    run:      31918255417
    job:      95093683595
    result:   SUCCESS
    branch:   exp/hl7-comparator-fix-qualification
    head:     c94748b73002d9f3cbec6f26f3bcd16ba3b1dfd9
    

    Artifact:

    name:   hl7-upstream-fix-qualification-evidence
    id:     9256943692
    sha256: e405990cff49afa40ecebae314241302a8bb7024d92abb1355a40aea06e0f1fc
    

    Deterministic full-corpus A/B result:

    status_a:   0
    status_b:   0
    comparison: 0
    A/B summary bytes: IDENTICAL
    summary sha256:
    074594e630c02f10c4bec9bcb903e153edaeb88a04c65636ff11c11301547409
    

    All unchanged frozen cases completed:

    C001 complete — 67 oracle comparisons
    C002 complete — 27 oracle comparisons
    C003 complete — 43 oracle comparisons
    

    Narrow classification:

    CF10_PATCHED_HL7_ORACLE_FULL_FROZEN_CORPUS_PROVEN
    

    This proves the two submitted upstream DefinitionNavigator fixes are sufficient for the unchanged CF-10 corpus when applied to the exact 6.10.2 source. It does not authorize using the unmerged fork/source as the production CF-06 oracle.

    Issue #15 therefore advances to its next external gate: obtain an upstream-authoritative merged/source identity for hapifhir/org.hl7.fhir.core#2554, then qualify that exact source through the full CF-06/CF-10 sequence already recorded here. The production 6.10.2 pin remains unchanged.

  2. TheHalfMoon commented on Aug 16, 2026

    @TheHalfMoon
    OwnerAuthor

    Current upstream gate rechecked after the patched full-corpus proof.

    Live upstream state:

    hapifhir/org.hl7.fhir.core PR #2554
    state: OPEN / NON-DRAFT / UNMERGED / MERGEABLE
    head:  e67d1c3f1d984fde11b2dfeb522bf9cba8c8d066
    base:  master
    master SHA: b06c7ee7c57367305ff72b6bb09ce779b5dbd6ce
    reviews: none
    comments: none
    

    Upstream master protection requires Pull Request Pipeline. For the fork PR head, the visible GitHub workflows (Trivy Security Scans, BIDI CHECK, License Check) are all action_required, so this is treated as maintainer-approval/routing state, not as a test failure. No required Pull Request Pipeline success is currently visible.

    The upstream README identifies grahamegrieve, jamesagnew, dotasek, and markiantorno as project maintainers. A review request to grahamegrieve was attempted through the current GitHub integration and was rejected with 403 Resource not accessible by integration; no upstream mutation occurred.

    No CF-06 production pin/source change is authorized or made. PR #11 remains Draft/unmerged. CF-12 remains not started. The next executable CF-06 qualification step still requires an upstream-authoritative merged/source identity.

  3. TheHalfMoon commented on Aug 16, 2026

    @TheHalfMoon
    OwnerAuthor

    Upstream follow-up package prepared; no upstream write was possible from the current GitHub integration.

    Current upstream state checked 2026-08-16 (Asia/Riyadh):

    If there is still no maintainer interaction by 2026-08-24, the appropriate next action is one concise follow-up on the upstream PR. Suggested text:

    Friendly follow-up: this patch fixes two DefinitionNavigator failures reproduced on published US Core and IPS profiles, including self-comparison failures. The two focused regression tests pass with the fix and fail with the original exceptions without it. In addition to the 12/12 targeted real-package probes, we completed two independent full runs of an unchanged three-IG corpus: 67 US Core + 27 IPS + 43 mCODE StructureDefinition comparisons per run, all operationally complete, with byte-identical summaries (SHA-256 074594e630c02f10c4bec9bcb903e153edaeb88a04c65636ff11c11301547409). Happy to adjust the patch to project conventions if needed. Could a maintainer please approve the PR workflows / take a look when convenient?

    This note does not authorize a commandF production CF-06 pin/source change, PR #11 merge, semantic reinterpretation, or CF-12.

  4. TheHalfMoon commented on Aug 18, 2026

    @TheHalfMoon
    OwnerAuthor

    Live upstream gate check — 2026-08-18

    hapifhir/org.hl7.fhir.core#2554 remains OPEN / non-draft / unmerged.

    Exact upstream candidate identity remains:

    head: e67d1c3f1d984fde11b2dfeb522bf9cba8c8d066
    base-at-submission: b06c7ee7c57367305ff72b6bb09ce779b5dbd6ce
    

    No maintainer review comments or review submissions are currently visible. The upstream GitHub Actions entries visible for this fork PR (Trivy Security Scans, BIDI CHECK, License Check) are action_required, so they are not treated as test failures or as green qualification evidence.

    Therefore the issue #15 exit condition is NOT met: there is still no upstream-authoritative merged/source identity to qualify. Production CF-06 remains pinned to HL7 6.10.2 / d06577dbc5c62c74a2a8823fbc4830a3024d5b0b.

    No production pin change, CF-10 merge, semantic reinterpretation, or CF-12 work is authorized by this status check.

  5. TheHalfMoon commented on Aug 18, 2026

    @TheHalfMoon
    OwnerAuthor

    Upstream-gate status update (read-only verification, 2026-08-19 local founder time):

    • hapifhir/org.hl7.fhir.core#2554 remains OPEN / non-draft / unmerged on exact head e67d1c3f1d984fde11b2dfeb522bf9cba8c8d066.
    • Upstream master remains b06c7ee7c57367305ff72b6bb09ce779b5dbd6ce; no rebase/update is currently required.
    • The PR is technically mergeable but mergeable_state=blocked.
    • Protected master requires the status context Pull Request Pipeline; that required status is not present yet on the PR head.
    • The visible GitHub workflows (Trivy Security Scans, BIDI CHECK, License Check) remain action_required, consistent with an external-contributor maintainer approval gate rather than a test failure.
    • No upstream review submissions or PR comments are present yet.

    Reviewer targeting evidence: DefinitionNavigator.java has a prior direct slicing fix authored by Grahame Grieve (d3517bb539d8671352f331e4b8f8ba3432479b03), and in the closely related upstream PR #2002 he requested a regression test before acceptance and later approved the fix once the failing-master/passing-branch evidence was supplied. Therefore @grahamegrieve is the best-supported targeted maintainer ping for #2554; @dotasek also participated substantively in that prior review, but no broad/random reviewer request is justified.

    No commandF product change, production HL7 pin change, CF-10 merge, or CF-12 work is authorized by this status update.

  6. TheHalfMoon commented on Sep 5, 2026

    @TheHalfMoon
    OwnerAuthor

    Upstream review update — 2026-09-05

    Graham Grieve reviewed hapifhir/org.hl7.fhir.core#2554 today and provided substantive revision requests. The upstream PR remains open and unmerged.

    Confirmed review direction:

    1. Rebase the contribution onto current upstream master; the branch still carries an annotation that upstream has since removed.
    2. Replace the current contentReference workaround with a fix that prevents duplicate child expansion, and add a realistic recursive/sliced snapshot regression test asserting exactly one set of children.
    3. Preserve the existing extension / modifierExtension special-case ahead of the inherited-slicing fallback, and add a regression test proving that precedence.
    4. Strengthen the inherited-slicing regression test by asserting the discriminator; a genuine profile-backed fixture is preferred.

    Graham explicitly confirmed that the first reported issue is a real bug, but the submitted implementation is not yet acceptable as-is because it can change the failure mode into duplicate child expansion.

    Current disposition:

    UPSTREAM_REVIEW_RECEIVED
    UPSTREAM_REVISION_REQUIRED
    UPSTREAM_PR_OPEN_UNMERGED
    COMMANDF_CRITICAL_PATH_UNCHANGED
    PRODUCTION_HL7_PIN_UNCHANGED
    

    This remains an external upstream gate and stays excluded from the commandF critical path. No commandF production pin/source change, PR #11 merge, or downstream qualification claim is authorized by this review alone.

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