soft3/radio/audit/2026-10-10-memstore-parent-pairs/comparison.md

π 0.01% 282 words · 2 min

Exact-parent comparison limits

Radio0d11468503c052aa5dcefb0dabc4602bd8efdc60, Hemera23f3bbcff910ea6d504ceb505680a539260869da; final candidate manifest d7c6cc1560eeeadc1b0c880cca79a3e2b09a2379489bac37706b0603db7069ca. ruby compare-diagnostics.rb compares complete error blocks after closure/target path normalization and build-progress removal; full results are in paired-diagnostics.json.

  • Final formatter output matches parent: ordinary340 hunks, named460 hunks after additionally normalizing hunk start lines. The initial named candidate had two new import-order hunks; these were fixed, final source/tests re-frozen and rerun.
  • Workspace/consumer compiler, MSRV owning example/workspace compiler, named nextest compilation default/none/all, named Clippy default/none/all and nightly docs error blocks match by this method. Owning strict docs exit0 on both.
  • Owning Clippy is masked, with unequal observed error sets: parent additionally emits cyber-radio unnecessary_sort_by; candidate owning output stops at11 cyber-bao errors. Candidate named all-feature Clippy does emit that same cyber-radio error. No new-test Clippy certification or aggregate lint parity.
  • Deny exits13 on both. Complete blocks differ only at one synthesized manifest's path-dependent column113/116 and corresponding marker padding; the same low-confidence license diagnostic is emitted. This is tool output, no legal claim.
  • Native nextest outcomes differ. See timeout-diagnostic.md for exact names, error causes, serial reruns and the separate60s observation result; no aggregate suite parity or green broad-gate claim.
  • Semver0.51.0 reports202 passed/58 skipped checks, exit0, against both exact parent and v0.1.0. Its fresh451-package scratch resolution differs from the committed gate lock. Current/parent/tag source manifests/locks remain retained; the surviving generated helper manifest/lock is preserved with an explicit final-state-only note. Tag uses the pinned current Hemera; no historical release closure or behavioral-compatibility proof follows from semver.

Raw command receipts remain authoritative. Formatting/source comments/tests introduce no public API change; repaired ephemeral bytes match the existing 64-byte parent layout. Empty-root identity, persisted checksum/recovery, observation performance, bounded proof budgets and particle/root authority remain separate obligations.

Graph