[squid-protocol]

Part of #4706 (follow-up to #4704; policy #4702: IBM is the reference, GnuCOBOL the instrument).

What

Estate PARM / installation-default / CBL-card values for result-changing options outside SEMANTIC_OPTIONS were recorded and then ignored. They now go through the one resolver and are refused by name or declared.

• compiler_options.OPTION_TABLE: one row per option (default, honoured values, refuse/declare policy, reason); unhonoured().
• EffectiveOptions.unhonoured() in estate_options.py; SEMANTIC_OPTIONS and the evidence option fingerprint are unchanged.
• equivalence_common.compile_options raises UnsupportedOption for refuse rows; option_differences declares the declare rows next to TRUNC(OPT); data_ccsid().
• register/C5.md: the option table (ZONEDATA, INVDATA, CODEPAGE, NSYMBOL, DISPSIGN, AFP/FLOAT, SSRANGE, NUMCHECK/INITCHECK, OPTIMIZE, plus ARITH / INTDATE / NUMPROC rows), IBM Programming Guide "Compiler options" cited.
• tests/cobol_mainframe/test_option_table.py (52 tests): PARM and card, NO- forms, cancellation, CODEPAGE, estate files, table/doc drift.

class member	result
ZONEDATA(NOPFD/MIG), INVDATA, NSYMBOL(DBCS), DISPSIGN(SEP), FLOAT(NATIVE)	refused by name
SSRANGE, NUMCHECK, CODEPAGE not matching the data	declared difference
OPTIMIZE, AFP, INITCHECK, defaults	recorded, no effect
ARITH(EXTEND), INTDATE(LILIAN)	unchanged: still refused via COBC_OPTIONS

Owner decisions needed

I took these provisionally; please confirm or overrule:

1. SSRANGE / NUMCHECK / CODEPAGE: declared, not refused. Proven as NOSSRANGE / NONUMCHECK / the case's data code page, listed as declared differences. Alternative: refuse.
2. INVDATA refused wholesale. Sub-option semantics (INVD(FNC|NOFNC,CS|NOCS)) are not read.
3. CODEPAGE 1140 and 37 honoured (IBM default; every proof already assumes it). Any other CCSID not matching the case's data_encoding is a declared difference: GenApp's CODEPAGE(285) is the one estate value affected.
4. FLOAT default HEX. The IBM 6.4 options page states no default; HEX is what register C6 models.
   Also unverified: the release in which NUMCHECK / INITCHECK / DISPSIGN / INVDATA appeared.

Gates

pr_gates --fast 6/6; --ratchets 7 pass, 0 fail, 0 skipped. tests/tools/test_evidence_report.py::test_the_committed_reports_are_current fails identically on clean origin/main (rendered report awaits the refresh bot; not committed here). Evidence records show harness/oracle/generator staleness only (scheduled re-proof).

Sibling gaps

ARITH(EXTEND), INTDATE(LILIAN) engine support; SSRANGE/NUMCHECK checks in the det port.

🤖 Generated with Claude Code

[github-actions]

🐦‍⬛ Muninn Security Scan

✅ No security issues found.

🐦‍⬛ Powered by Muninn · Skald Lab

[squid-protocol]

Owner decisions (2026-10-14), all four provisional calls confirmed:

1. SSRANGE / NUMCHECK / CODEPAGE not matching the data: declared differences (proven as NOSSRANGE / NONUMCHECK / the case's data code page), not refused.
2. INVDATA refused wholesale; sub-options not read.
3. CODEPAGE 1140 and 37 honoured; any other CCSID not matching the case's data is a declared difference (GenApp CODEPAGE(285)).
4. FLOAT defaults to HEX (register C6).
   Still open, for the sibling issue: the release each of NUMCHECK / INITCHECK / DISPSIGN / INVDATA appeared in; ARITH(EXTEND) / INTDATE(LILIAN) engine support; SSRANGE / NUMCHECK checks in the det port. Labelled for merge at ef56649fb338.

🤖 Generated with Claude Code

[squid-protocol]

No description provided.