Repository navigation
equivalence: BNK1UAC on generated Db2 rows (Part of #4507) - #5569
Conversation
cbsa-bnk1uac-generated: 30 generated ACCOUNT rows, 50 generated screen scenarios (ENTER inquiries, PF5 updates with type, rate, overdraft and statement-date fields drawn from the hand-written case's values), compare_sql on. Every event and statement equal on 61 scenarios, 31/31 paragraphs, 68/88 branches. The report is not "proven": the java-facade leg refuses BNK1UAC (as for the hand-written cbsa-bnk1uac). Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
|
No description provided. |
|
PR #5569 @ 958ca14: 1 failed check(s) == det (failure) -- real |
|
CI shepherd: the head of this PR (aebec07) is not the one a maintainer labelled |
|
No description provided. |
|
No description provided. |
|
PR #5569 @ aebec07: 2 failed check(s) == ground-truth (failure) -- real == ground-truth part 3/4 (failure) -- real |
|
No description provided. |
|
No description provided. |
|
PR #5569 @ 858dd40: 2 failed check(s) == full-suite (failure) -- real == full-suite part 2/3 (failure) -- real |
|
No description provided. |
|
No description provided. |
|
No description provided. |
|
No description provided. |
Part of #4507 (slice 2: cbsa-bnk1uac-generated, same shape as #5558).
Checklist: partial -- the update screen runs on generated rows with every event and SQL statement equal, but the java-facade leg of the report refuses BNK1UAC, so the case is not "proven" (see below)
What
cbsa-bnk1uac-generated(port_from: cbsa-bnk1uac): 30 ACCOUNT rows generated from the CBSA DDL, 50 generated scenarios (ENTER inquiries, PF5 updates, a few PF3/PA1) plus the 7 stateless stated ones. ACCNO drawn from the rows (15% miss); type, rate, overdraft and last/next statement dates drawn fromvalueslists lifted from the hand-written scenarios (valid, boundary, refused). No four-decimal rate (bounds_refused: BNK1UAC reference-modifies INTRTI past its 7 bytes).compare_sqlon; ACCOUNT and PROCTRAN compared after every task.Evidence (EQUIVALENCE_E2E=1, Docker Db2 + GnuCOBOL, JDK 17)
proven: false: the java-facade leg refuses every scenario ("the task RETURNed a [B, not Bnk1uacDfhcommarea (a LENGTH past its DTO ...)"). The hand-writtencbsa-bnk1uacgives the same refusal (re-run here,DIFFERS, and it has nodet_sweep_coverageentry), so it is not caused by generated rows. The test therefore asserts events, SQL and the UPDATE, notproven; no proof was loosened. No coverage ledger entry is added (the ledger holds proven cases only). No mutation run (needs a proven case).test_the_inqacccu_det_port_holds_on_generated_db2_rowsfails in the same E2E run (assert at line 302); not touched here.pr_gates --fast6/6; tree-wide guards 178 passed.--ratchetsnot run (no engine/golden/ledger change).Class
Class: a CBSA BMS screen case on a hand-written seed.sql and fixed inputs -> generated rows and scenarios. This PR: BNK1UAC. Not here: bnk1cac, bnk1cca, bnk1ccs, bnk1dcs, bnk1tfn (multi-table CUSTOMER/ACCOUNT FKs for the creating ones).
Cross-repo
None.
🤖 Generated with Claude Code