Known issue closure platform bound

The check is known_issue_closure_platform_bound. It is VERIFY-stage. It errors. It only fires once a known issue declares platforms:. No record in the corpus with that field is a skip. For each declared record it reads the only current verdict. A live reproduction is not a false closure. A non-reproduction from a reachable platform is a witnessed closure. A non-reproduction from outside the declared set is uncreditable_closure. A declared constraint with no platform stamp is unstamped_closure. A reproduction from outside the declared set is platform_declaration_refuted. A linux token without os: is malformed_platform_declaration. It does not write the YAML. It does not prove the Go. It does not run the reproducer.

This hop is the closure companion to build matrix complete. That page is whether a declared build target's config tokens showed up on some result. This hop is whether a not-reproduced verdict was observed on a platform the record said could reach the subject. The same superset rule applies: an observation covers a declaration when its config: tokens include every token on one alternative. A Salesforce “known issue” page is not that floor. Ads known issue is that status page. A green Darwin job is not that floor. //go:build, #ifdef, and sys.platform make a skip look like a fix. The finding names the record whose only closure was unexercised.

The resolutions are alternatives, not steps. If the subject is Linux-only, refresh the reproducer on a declared platform and keep both rows. If the declaration is wrong, widen platforms:. If the stamp is missing, run proof evidence refresh so the host axes land. Opting the check out reports skip, never pass. Do not delete platforms: to silence the finding if the subject is still gated. That is the finding.

# KI-116 platforms: [os:linux | os:android]
# proof/evidence/ki-116-reproducer.yaml
# verdict: known_issue_not_reproduced
# config: [os:darwin, arch:arm64]
# proof audit --check known_issue_closure_platform_bound
# [VERIFY] known_issue_closure_platform_bound
# KI-116 uncreditable_closure on [os:darwin,arch:arm64]
# ERROR
# proof help known_issue_closure_platform_bound

Read both sides before you edit anything. The declared reachable set, and the host that recorded the absence. A corpus with no platforms: skips this hop. That skip is not a proof that every closure was observed where the subject can run:

proof audit --check known_issue_closure_platform_bound
proof help known_issue_closure_platform_bound
proof evidence refresh KI-116
proof known-issue edit KI-116 --add-platform 'os:linux'
proof config set project.checks.known_issue_closure_platform_bound.enabled false

The teaching graph is still one declared reachable set whose only closure was observed elsewhere. Close it by refreshing on a declared platform and keeping both rows, or by correcting platforms: if the declaration is wrong. Do not strip platforms: to make the skip look like a pass if the subject is still gated. The hop does not run that refresh for you. The hop does not prove the Go.

# refresh on linux-x86_64, keep the darwin row
# proof audit --check known_issue_closure_platform_bound
# 0 uncreditable closures
# VERIFY may move on. pass sits on a reachable absence, not a skipped compile

The suite-level hop stays on build matrix complete. The quality floor stays on known issue complete. Jama still authors. Proof vs Jama.

Error when a known issue that declares platforms: offers a closure the observing host cannot support. Default severity is error. It fails the merge when the audit treats error as blocking. No platforms: on any record is a skip, and that skip is not a proof that every closure was observed where the subject can run. Opt-out reports skip, never pass. The hop does not write the evidence file. It does not SSH to linux-x86_64. It does not compile the gated subject. It does not prove the Go. A quiet hop is not a proof that Jama's shall still holds on every kernel, only that every in-scope declared record currently has a creditable verdict, or that nothing asked. We have not scored this floor against a frozen Jama pack, a VectorCAST corpus, or a Salesforce bulletin. The loss is named, not scored. known_issue_platform_coverage is the coverage hop: which records have only been looked at from one platform. See known issue platform coverage. This hop is whether the only closure is even creditable.

The suite-level hop stays on build matrix complete. The quality floor stays on known issue complete. The engagement stays on software correctness audit. Jama still authors.

What is known issue closure platform bound? Same question. Same URL.

Is this known issue complete? No. That hop is whether the record has evidence and an origin. This hop is whether the only closure was observed where the subject can run. See known issue complete.

Is this build matrix complete? No. That hop is whether a declared target's config tokens showed up on some result. This hop applies the same superset rule to a KI closure. See build matrix complete.

Is this a Salesforce known issue? No. Ads known issue is that status page. A hosted vendor bulletin is not this cell.

Does a Darwin skip clear a Linux-only KI? No. A non-reproduction is a universal claim. One unreachable host cannot settle it.

Does this finding fail the merge? Yes by default. The check keeps error severity. A skip because nothing declared platforms: is not that error.

Does a quiet hop prove every closure shipped? No. A corpus with no platforms: skips. The hop observes declared reachability versus the observation host. It does not prove the Go.

Is this known issue platform coverage? No. That hop warns when a creditable closure exists on only one of several declared build targets. This hop fails when the only closure is not creditable at all. See known issue platform coverage.

Is Proof an LDRA alternative for the instrument? No. LDRA still owns the avionics toolchain. Proof vs LDRA.

Is Proof a Jama alternative for the shall? No. Jama still authors. Proof vs Jama.