Repository navigation
det: refuse a non-digit in a TALLYING counter or STRING/UNSTRING POINTER (Part of #4692) - #5327
Conversation
…RING/UNSTRING POINTER (Part of #4692) Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
|
No description provided. |
|
PR #5327 @ 2941009: 1 failed check(s) == crucible (failure) -- real |
… (Part of #4692) Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
|
CI shepherd: the head of this PR (65e2b7a) is not the one a maintainer labelled |
|
No description provided. |
|
No description provided. |
|
PR #5327 @ 65e2b7a: 2 failed check(s) == ground-truth (failure) -- real == ground-truth part 1/2 (failure) -- real |
…Part of #4692) test_every_committed_record_is_current failed: both ports changed since their proofs. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
|
CI shepherd: the head of this PR (7049d87) is not the one a maintainer labelled |
|
No description provided. |
|
No description provided. |
Part of #4692 (follows #4889: the receiver of ADD / SUBTRACT / MULTIPLY / DIVIDE).
What
The class is "a statement reads a zoned item's current value before it stores, and the item holds a non-digit". ADD-family receivers already refuse by name; the same read-before-store sites did not. They now go through
Cobol.counterNum(refuse by name,Zoned.refuseCounter) on the same IBM ground (no documented result; ZONEDATA(PFD) assumes valid zone bits):A digit-only counter counts as ever. Tests: 5 receiver shapes x 5 operations refused, digit-only starts, plain-item path (
test_det_nondigit_zoned.py); register C11 andcobolrt_rules.json(RT-CBL-COUNTER-NONDIGIT-REFUSED) updated; the PCCONF overlay runtime re-synced.Evidence
Burned estates: every INSPECT TALLYING counter in CardDemo/CBSA/GenApp/zECS/DBB is COMP/BINARY and zeroed first, so no burned port moves (translator untouched).
pr_gates --fast6/6,--ratchets8/8 pass.Cross-repo
None.
🤖 Generated with Claude Code