Repository navigation
equivalence: GenApp lgapvs01 and lgupvs01 error paths with LGSTSQ and injected RESPs (Part of #5252) - #5403
Conversation
… and injected RESPs (Part of #5252) Neither program ASSIGNs STARTCODE, so there is no D path to add. Their uncovered branches were the WRITE-ERROR-MESSAGE paths the case notes called refused; as in #5324 the case now runs LGSTSQ in the same task (X6) with DUPREC, NOSPACE, IOERR (lgapvs01) and NOTFND, READ IOERR, REWRITE IOERR / NOSPACE (lgupvs01). The four EIBCALEN branches stay uncovered: the 101-byte ERROR-MSG exceeds LGSTSQ's WRITE-MSG, which X6 refuses at the first LINK. Coverage ledgers and lgapvs01 evidence re-proved. Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
|
No description provided. |
|
PR #5403 @ 1a7730f: 2 failed check(s) == full-suite (failure) -- real == full-suite part 2/3 (failure) -- real |
… now that lgapvs01 proves its LINK (Part of #5252) Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
|
CI shepherd: the head of this PR (2323be0) is not the one a maintainer labelled |
|
No description provided. |
Part of #5252 (follow-up to #5324).
What
lgapvs01andlgupvs01neverASSIGN STARTCODE, so the issue's D-path premise does not apply to them. Their uncovered branches were theWRITE-ERROR-MESSAGEpaths, which the case notes called refused. As in #5324, each case now listsLGSTSQinprograms(X6) and has aregion, with these scenarios:Evidence (levels, not "proven")
IF EIBCALENarms inWRITE-ERROR-MESSAGE.ERROR-MSGis 101 bytes, longer than LGSTSQ's 95-byteWRITE-MSG, so X6 refuses the WRITEQ TD at the first LINK, before those IFs. A short-commarea scenario would stop at the same refusal, so none is stated. Closing it needs X6 to model LENGTH > FROM (oracle change, not done).Class
GenApp
lg*vs01programs whose uncovered branches are the error paths. Handled here: lgapvs01, lgupvs01. Done in #5324: lgacvs01, lgdpvs01, lgucvs01, lgicvs01 (D path).Gates
pr_gates --fast6/6 and--ratchets8/8 (ports, port-surface, fact-crosscheck, completeness, ground-truth, time, estate, cics-spec);test_evidence.pyandtest_det_coverage_ledger.pypass. No engine code changed, so no golden bless. The lgapvs01evidence.jsonis re-proved (unapproved, as the tool writes it).🤖 Generated with Claude Code