Repository navigation
X39: model-port XCTL value cases and burned-estate reach (Part of #5060) - #5478
Conversation
…ull) and burned-estate reach (Part of #5060) Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
|
No description provided. |
|
PR #5478 @ 218f752: 2 failed check(s) == full-suite (failure) -- real == full-suite part 1/3 (failure) -- real |
|
No description provided. |
Part of #5060 (first checkbox: model-port scalar fields). The runtime refusal landed in #5332 and #5381; this PR pins the rest of the input class and records burned-estate reach.
Class: X39, XCTL of storage a LINK caller owns, model-port value COMMAREA traced by equality. Handled here: String[] element, derived values (suffixed String, sum: not traced, pinned as NORMAL), null field. Already handled: nested/list/map values, primitive tables, cross-width numbers.
Burned-estate reach: no burned program is both LINKed and an XCTL issuer (CardDemo XCTLers are transaction-started; ZECS001 XCTLs but LINKs ZECS002, which does not; COPAUS1C's LINK target is not in the corpus). Unreached, so no runtime change.
Tests: test_model_link_xctl_4956 (JDK 17), the 4816/4957 XCTL tests, refusals-registered, tree-wide guards: 76 passed, 3 skipped. pr_gates --fast 6/6. No golden bless (no parsing change); no template change.
Sibling gaps: copy vs same storage needs a real z/OS run (stays REFUSED); value derived from a DTO field cannot be traced (no burned program needs it, frontier). SET ADDRESS alias done in #4957.
Cross-repo: none.
🤖 Generated with Claude Code