You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Census breadth survey (#5052, main d86d5ea): 2 non-burned CICS programs (of 21 surveyed) hit this gap, counts only; no census code was read (mainframe-evidence-policy 3).
Gap family: JSON GENERATE / JSON PARSE
Det translator status: not modelled / refused. Burned-estate programs affected: 0. None of these alone makes a non-burned program whole (other gap classes remain), so rank by program count only.
Class / Sibling gaps
Class: det-translator statement families the census surveys report as "not modelled" in CICS programs. Siblings are the other rows in the #5052 ranking comment (STRING, JSON GENERATE/PARSE, EXEC CICS WEB, MOVE single-byte literal, EXEC SQL ROLLBACK in CICS); each has its own issue. Work only from burned estates (case work on non-burned repos is out of scope).
Reproduce: cics_census.py survey --out DIR --label before with CICS_CENSUS_CORPORA, then blockers DIR --label before.
Work from IBM's definition on fixtures / cics-crucible only; never census source (#7709 rule 2).
JSON GENERATE: translate from the IBM definition (name matching, NAME / SUPPRESS / COUNT, JSON-CODE, ON EXCEPTION) on a fixture; register entry marks it oracle-unbacked (no L2+ claim)
JSON PARSE: same, on a fixture (name matching, CONVERTING / WITH DETAIL, JSON-STATUS, ON EXCEPTION)
Census mock trial work (owner, 2026-11-04: #7709). The 11-03 stop ("refused by design, 0 burned programs use it; needs owner relabel or close") no longer applies. The owner made the census's translator gaps trial work, and the trial counts a program as accounted only when it translates whole or carries a named source defect, so a refusal by design does not close it. Fix with synthetic reproductions (cics-crucible or a fixture), never census code (mainframe-evidence-policy section 3). Measure by the #7709 re-survey. Relabelled scope:core like #7709. CICS breadth is melek's lane (owner 2026-10-30). 🤖 Claude Code for the owner
Closed by #7759: JSON PARSE/GENERATE are holes by name with IBM's definition and the oracle's silence as the reason. Modelling is refused by design (GnuCOBOL has no JSON statement to back it) and no burned program uses JSON, so acceptance is met. XML siblings are noted in #7759.
Reopened by the owner (2026-11-05 evening) for the census trial #7709.
T4 (a2597c7e5, 2026-11-06) still shows JSON GENERATE and JSON PARSE as not modelled in 2 census programs. A hole by name does not account a program: #7709 rule 4 names only faults of the source or its copy. These are both programs' last translator gaps beside CICS WEB (#7525) (gap_prevalence "needs several gaps together" rows).
The census goal is L1 only (translates whole, #7709 rule 5). So the closing reason, no GnuCOBOL oracle for JSON, blocks L2+ evidence but not a translation. Acceptance now:
Claimed by promoter-bot host melek, lease renewed 2026-11-06T18:02Z. It lapses after 6 h without a renewal. To take it back, assign someone or remove the claim:melek label.
#7888 merged: a partial slice of [c:3fc8b69b] "JSON PARSE: same, on a fixture (name matching, CONVERTING / WITH DETAIL, JSON-STATUS, ON EXCEPTION)"; left: ON EXCEPTION / NOT ON EXCEPTION done; JSON-CODE / JSON-STATUS, CONVERTING, SUPPRESS ... WHEN and wider item kinds left (promoter-bot)
#7897 merged: a partial slice of [c:bb8e1760] "JSON GENERATE: translate from the IBM definition (name matching, NAME / SUPPRESS / COUNT, JSON-CODE, ON EXCEPTION) on a fixture; register entry marks it oracle-unbacked (no L2+ claim)"; left: the JSON-CODE value after an exception, SUPPRESS WHEN / EVERY, CONVERTING, NAME OMITTED, OCCURS DEPENDING ON stay holes (promoter-bot)
#7905 merged: a partial slice of [c:bb8e1760] "JSON GENERATE: translate from the IBM definition (name matching, NAME / SUPPRESS / COUNT, JSON-CODE, ON EXCEPTION) on a fixture; register entry marks it oracle-unbacked (no L2+ claim)"; left: CONVERTING, NAME OMITTED, DEPENDING ON, JSON-CODE after exception left (promoter-bot)
#7932 merged: a partial slice of [c:bb8e1760] "JSON GENERATE: translate from the IBM definition (name matching, NAME / SUPPRESS / COUNT, JSON-CODE, ON EXCEPTION) on a fixture; register entry marks it oracle-unbacked (no L2+ claim)"; left: NAME OMITTED, elementary source, floats left; the item's named phrases are covered once #7905 and #7930 merge (promoter-bot)
Activity
squid-protocol commentedon Nov 5, 2026
Census mock trial work (owner, 2026-11-04: #7709). The 11-03 stop ("refused by design, 0 burned programs use it; needs owner relabel or close") no longer applies. The owner made the census's translator gaps trial work, and the trial counts a program as accounted only when it translates whole or carries a named source defect, so a refusal by design does not close it. Fix with synthetic reproductions (cics-crucible or a fixture), never census code (mainframe-evidence-policy section 3). Measure by the #7709 re-survey. Relabelled
scope:corelike #7709. CICS breadth is melek's lane (owner 2026-10-30). 🤖 Claude Code for the ownergap_prevalence: the census JSON GENERATE / PARSE rows name #7524 as o…
det: JSON PARSE / JSON GENERATE holes carry IBM's definition and the …
squid-protocol commentedon Nov 5, 2026
Closed by #7759: JSON PARSE/GENERATE are holes by name with IBM's definition and the oracle's silence as the reason. Modelling is refused by design (GnuCOBOL has no JSON statement to back it) and no burned program uses JSON, so acceptance is met. XML siblings are noted in #7759.
squid-protocol commentedon Nov 6, 2026
Reopened by the owner (2026-11-05 evening) for the census trial #7709.
T4 (
a2597c7e5, 2026-11-06) still shows JSON GENERATE and JSON PARSE as not modelled in 2 census programs. A hole by name does not account a program: #7709 rule 4 names only faults of the source or its copy. These are both programs' last translator gaps beside CICS WEB (#7525) (gap_prevalence "needs several gaps together" rows).The census goal is L1 only (translates whole, #7709 rule 5). So the closing reason, no GnuCOBOL oracle for JSON, blocks L2+ evidence but not a translation. Acceptance now:
🤖 Posted by Claude Code
[-]det: JSON GENERATE / JSON PARSE not modelled (census: 2 non-burned programs)[/-][+]det: JSON GENERATE / JSON PARSE not modelled (census #7709, 2 programs)[/+]squid-protocol commentedon Nov 6, 2026
Claimed by promoter-bot host
melek, lease renewed 2026-11-06T18:02Z. It lapses after 6 h without a renewal. To take it back, assign someone or remove theclaim:meleklabel.det: catalogue JsonGenerate in the cobolrt rule catalogue (Part of #7524
det: list JsonGenerate in the cobolrt API manifest (Part of #7524)
10 remaining items
squid-protocol commentedon Nov 6, 2026
#7888 merged: a partial slice of [c:3fc8b69b] "JSON PARSE: same, on a fixture (name matching, CONVERTING / WITH DETAIL, JSON-STATUS, ON EXCEPTION)"; left: ON EXCEPTION / NOT ON EXCEPTION done; JSON-CODE / JSON-STATUS, CONVERTING, SUPPRESS ... WHEN and wider item kinds left (promoter-bot)
det: JSON-CODE is 0 after a JSON GENERATE; the exception value stays …
squid-protocol commentedon Nov 6, 2026
#7897 merged: a partial slice of [c:bb8e1760] "JSON GENERATE: translate from the IBM definition (name matching, NAME / SUPPRESS / COUNT, JSON-CODE, ON EXCEPTION) on a fixture; register entry marks it oracle-unbacked (no L2+ claim)"; left: the JSON-CODE value after an exception, SUPPRESS WHEN / EVERY, CONVERTING, NAME OMITTED, OCCURS DEPENDING ON stay holes (promoter-bot)
det: JSON PARSE takes COMP / COMP-3 / COMP-5 numeric receivers (Part of
det: regenerate the cobolrt API manifest for JsonGenerate (Part of #7524
cobolrt API manifest: add JsonParse.tryParse status overload as publi…
cobolrt API manifest: the emitter calls JsonParse.tryParse's six-argu…
cobolrt rules: JsonGenerate.table is plumbing (Part of #7524)
det: JSON GENERATE SUPPRESS ... WHEN / EVERY (Part of #7524) (#7905)
squid-protocol commentedon Nov 6, 2026
#7905 merged: a partial slice of [c:bb8e1760] "JSON GENERATE: translate from the IBM definition (name matching, NAME / SUPPRESS / COUNT, JSON-CODE, ON EXCEPTION) on a fixture; register entry marks it oracle-unbacked (no L2+ claim)"; left: CONVERTING, NAME OMITTED, DEPENDING ON, JSON-CODE after exception left (promoter-bot)
cobolrt api: JsonGenerate.table is used by emitted code, so public (P…
cobolrt API manifest: keep JsonParse.tryParse status overload public …
squid-protocol commentedon Nov 6, 2026
Churn run
7524-20261106T1758Z(host melek) ended without a PR. No slice left; waits on #7928/#7930/#7932. Handoff written by worker.det: JSON GENERATE takes OCCURS DEPENDING ON tables (Part of #7524) (#…
squid-protocol commentedon Nov 6, 2026
#7932 merged: a partial slice of [c:bb8e1760] "JSON GENERATE: translate from the IBM definition (name matching, NAME / SUPPRESS / COUNT, JSON-CODE, ON EXCEPTION) on a fixture; register entry marks it oracle-unbacked (no L2+ claim)"; left: NAME OMITTED, elementary source, floats left; the item's named phrases are covered once #7905 and #7930 merge (promoter-bot)