CICS coverage: file-I/O INTO / FROM as lineage edges (#3492)

[squid-protocol]

Part of epic #3489.

Problem

Field lineage (#3452) stopped at a file's FD record. READ … INTO, WRITE … FROM and the related verbs move whole records between the FD buffer and working storage, but they produced no data-move row. So a field couldn't be traced from an input file into the program and back out to an output file.

What it does

• Extraction (core/data_moves.py): new rows for

  • READ / RETURN … INTO t: source_kind file, the FD record → t. READ's NEXT / RECORD / KEY IS phrases are handled.
  • WRITE / REWRITE / RELEASE r FROM s: s → r.
  • ACCEPT t [FROM DATE YYYYMMDD | TIME | …]: source_kind special, and SYSIN when there's no FROM.
  • A READ without INTO, or a WRITE without FROM, moves no program data, so it gets no row.

• Reader (GalaxyIR):

  • A READ's file resolves to its FD record.
  • An FD's 01 records share one storage key, since they overlay the same buffer.
  • The whole-record I/O verbs carry a field's offset through, just as a group MOVE does.

  Example (CardDemo CBTRN02C): the backward trace from DALYTRAN-AMT now reaches the DALYTRAN-FILE FD record through its READ … INTO. The forward trace ends at the FD records the program WRITEs / REWRITEs.

• Unchanged: CICS / SQL / DL/I INTO and FROM areas were already lineage endpoints, because the area is program storage.

Ground truth

• An independent key reader (add-io-moves) feeds a new section io_moves and a new scored field "file I/O moves": CardDemo 146/146, zOpenEditor 63/63. CBSA has no native file I/O.
• "data moves" still compares only its original verbs, on both sides, so its sample_verified sign-off from Ground truth: IMS census + sampled data-move census (#3477, #3452) #3485 stands.
• Blind census: a new full io suite in cross_verify_sections.py, asked of every COBOL source with a FILE SECTION, an ACCEPT, or keyed rows.
  • The round trip gave 0 disagreements.
  • 6 reviewer batches agreed on 209/209 facts with 0 disagreements. CBSA's 17-file recall batch came back empty, as keyed.
  • Result: cross_verified.

Checks

• Golden masters: re-blessed. The only change is 118 added "Data Moves" rows per mode, all READ / WRITE / REWRITE / ACCEPT; verified additive by a diff against HEAD.
• The ledger is clean, and the ruff and mypy audits are clean.
• Full suite: 11274 passed.

Note: this is based on main without #3500 (symbolic maps), which was closed unmerged. The two touch different parts of galaxy_ir.

Closes #3492

🤖 Generated with Claude Code

https://claude.ai/code/session_017ZsVaAkb86P5r5JXDC2Y9g

[github-actions]

🐦‍⬛ Muninn Security Scan

✅ No security issues found.

🐦‍⬛ Powered by Muninn · Skald Lab