[squid-protocol]

Part of #4270.

What

Register X30 in oracle_assumptions.md (IBM's pages cited there). Chosen by programs made whole (cics_census.py blockers, baseline origin/main 82918f0, 111/179 whole).

• DEFINE COUNTER (COUNTER / POOL / VALUE): the counter is created with VALUE, or "an initial value of zero" when it is omitted (IBM); a counter of that name and pool that exists is INVREQ RESP2 202 "Duplicate counter name" (IBM's page lists no DUPREC); a pool or counter name outside IBM's characters is INVREQ RESP2 403 / 404. No POOL is 8 blanks, a different pool from HCPOOL.
• DELETE COUNTER: removes it; a counter that is not there is INVREQ RESP2 201 "Named counter not found" (IBM lists no NOTFND); a bad pool 403 (the page lists no 404).
• Bug (GET COUNTER): a missing counter answered NOTFND (RESP 13). IBM's GET COUNTER page lists INVREQ RESP2 201 and no NOTFND. Fixed in both runtimes (CicsTask.getCounter now returns {RESP, RESP2, value}; the det translator writes both; the stub's GGCGCNT); GET and QUERY COUNTER also answer 403 / 404 for a pool / name outside IBM's characters. GET COUNTER now accepts RESP2.
• Refused by name (what IBM leaves open): MINIMUM / MAXIMUM (the defaults are "low-values" / "high values"; IBM does not say how they read signed), NOSUSPEND (BUSY needs a coupling facility), DCOUNTER (unsigned doublewords); at run time on both runtimes a VALUE below zero (406 depends on the unstated minimum) and a counter name of blanks only.
• The stub's GET-COUNTER event line is gone (the Java side never had one; the crucible converter rejects it).

GET COUNTER fix: impact on proven cases

No verdict moved. The only committed cases that run GET COUNTER are the equivalence cases genapp-lgacdb01 and genapp-lgacus01 (scenario add-without-counter): LGACDB01 tests WS-RESP Not = DFHRESP(NORMAL), so NOTFND and INVREQ take the same branch. Their why text said "(NOTFND)"; corrected per IBM. det_port.py check vs origin/main: 77/79 ports unchanged, those two changed (the GET COUNTER codegen) and det_port.py run genapp-lgacdb01 genapp-lgacus01 re-proves both (PROVED, 151/151 and 185/185 statements). No cics-crucible case, committed port or other equivalence case used GET COUNTER.

Proof

New cics-crucible case condition-handling/hc-named-counters (companion below; 2 scenarios, 24 breadcrumbs, traced by hand from IBM's four counter pages, check-hand-derived OK, validate.py 24/24, no SPEC change: the case defines its own counters): cobol-stub 2/2 and java-ported (det port HCCNTR 90/90 statements, 0 holes) 2/2, with --unpinned; a mutated log (L1 R=13) fails both sides. Unit tests: translator, spec entries, harness, CicsTask and the stub C (compiled), including the 403 / 404 rules and the refusals. pr_gates --ratchets 7/7 (against a v0.6.0 checkout), --fast 5/5.

Census (before -> after, translation only, burned + census corpora, baseline 82918f0)

• Programs translated whole: 111 -> 112 of 179 (non-burned 12 -> 12). Made whole: lgsetup (GenApp, 17/91 -> 91/91 statements, 0 holes).
• Programs using DEFINE / DELETE / GET: whole 20 -> 21; holes naming those verbs 10 -> 8 (left: zECS's GET DCOUNTER, refused by name).
• Next blockers (programs whose only gap): LayoutError fixed-form literal past column (8, non-burned), missing copybook BOOKFUNC (7), BAQRINFO (2), CMQGMOV (2), ISPF screens (2), InqacccuDfhcommarea: data after a POINTER (2, CBSA), <name>: undeclared item (1 / 4 touched).

Cross-repo

Companion: squid-protocol/cics-crucible#16 (the case). Merge it first; this PR's CI cics-crucible job does not see the case until a release + pin bump (not done here: no tag, the pin untouched).

🤖 Generated with Claude Code

https://claude.ai/code/session_01MQNoe4wPJgr7dbUvs3DG6D

[github-actions]

🐦‍⬛ Muninn Security Scan

✅ No security issues found.

🐦‍⬛ Powered by Muninn · Skald Lab