CICS coverage: symbolic maps generated from BMS source (#3490)

[squid-protocol]

Part of epic #3489 (CICS estate coverage gaps).

Problem

A CICS program COPYs its screen's symbolic map, which is the copybook the BMS assembler (DFHMAPS) generates at build time. Repositories rarely ship it: CBSA has none of its ten. So every screen field (CUSTNAMI, MESSAGEO, …) stayed unresolved, and field lineage (#3452) stopped at the terminal.

What it does

• core/bms_symbolic.py generates the COBOL symbolic map as copybook text from the screen fields we already extract (Engine: extract BMS symbolic-map field layouts as a fact channel #3347). That covers:

  • the TIOA prefix;
  • L/F/A/I items per named field, plus the O record REDEFINES;
  • extended-attribute bytes (DSATTS / EXTATT=YES, in C P H V U M T order);
  • PICIN / PICOUT;
  • OCCURS, in IBM's …D / DFHMSn form.

  It matches all 21 generated copybooks CardDemo checks in, line for line. That comparison is a test whenever the corpus is present.

• GalaxyIR parses the text with the engine's own record parser. It attaches the result (EngineFile.symbolic_copies, never added to files) to any program whose COPY <mapset> has no real COBOL copybook. Copybook resolution, storage spans and qualified lookups all see it. A real copybook always wins. A mapset defined in several BMS sources resolves to the file named after it: CBSA's BNK1TFM is in both BNK1B2M.bms and BNK1TFM.bms.

• GalaxyIR.symbolic_map_layouts() lists every generated map as NAME @offset+bytes.

Effect (CBSA)

• Unresolved data-move targets: 779 → 0. Unresolved sources: 105 → 1.
• field_lineage(UPDCUST, COMM-NAME, backward) now reaches BNK1DCS's CUSTNAMI, tagged with the endpoint cics MAP BNK1DC read.

Ground truth

• The key computes its own layout by arithmetic from the BMS source: 3 + k + LENGTH bytes per field, after the prefix. It writes no COBOL and uses no record parser.
• New subcommand add-symbolic-maps; new scored field "symbolic maps"; ledger flag symbolic_validated.

	CardDemo	CBSA
symbolic-map items	5307/5307	1320/1320

A blind census of the new key section is still owed. CardDemo's layouts are also checked against the real generated copybooks, which serve as an external oracle.

Checks

• Tests: an exact-text unit test, the CardDemo oracle, IR tests (resolution, and a real copybook winning over the generated one) and a key test.
• Golden masters are unchanged: this is a reader derivation with no new table.
• The ledger is clean, and the ruff, mypy and dead-key audits are clean.
• Full suite: 11273 passed.

Closes #3490

🤖 Generated with Claude Code

https://claude.ai/code/session_017ZsVaAkb86P5r5JXDC2Y9g

[github-actions]

🐦‍⬛ Muninn Security Scan

✅ No security issues found.

🐦‍⬛ Powered by Muninn · Skald Lab