---
name: v050-published
description: "v0.5.0 (M9 feature milestone) PUBLIC 2026-06-13 (counter 10); todlando drove merge+tag, deployah signed+published; first chance at Linux seamless no-bounce proof on fixed v0.4.2 brokers"
metadata: 
  node_type: memory
  type: project
  originSessionId: 1fb9fc52-b258-4b44-9c4c-7cae9345098f
---

v0.5.0 SIGNED + PUBLISHED 2026-06-13 (release counter 10; sequence v0.3.2=6, v0.4.0=7, v0.4.1=8, v0.4.2=9, v0.5.0=10 — counter != semver). M9 feature milestone SHIPPED. **HANDOFF SPLIT**: todlando drove merge+tag (all M9 → main @ 22a9b6f GREEN, version 0.4.2→0.5.0, tagged v0.5.0); **deployah** did sign+publish only (release keys never enter CI = signer-incumbent step). Differs from v0.4.x where deployah drove the full solo sequence — both modes work.

Minor bump (not patch): feature milestone + BREAKING `[pty_digest]` removal, pre-1.0 → minor. PRs #7(W1) #9(W2+gateway-wan) #10(W3 digest); #8 closed, its W2 content rode in via #9.

Acceptance hashes: linux `4c079100…84217ae`, win `984298ba…264f2ff7`. signed `54188f12…05e3ec9`. update-set.json v10. https://github.com/SaberMage/spt-releases/releases/tag/v0.5.0

**Why it matters:** fleet is now on the FIXED v0.4.2 brokers (canonical-path respawn, see [[v042-published]]), so v0.4.2→v0.5.0 apply on kitsubito is the FIRST genuine Linux SEAMLESS no-bounce field-proof — the proof deferred since [[v041-linux-respawn-blocker]]. Bytes-assert→auto-rollback now armed (wrong-bytes respawn FAILS apply + rolls back, no false-success).

**LINUX SEAMLESS PROVEN 2026-06-13** (operator greenlit roll-all). kitsubito v0.4.2→v0.5.0 apply: broker pid 1357210 HELD (no restart), brain pid 1357219→2135176 (rolled LIVE), gen 0→1, brain.ready exe_hash = 4c079100 == deployah's signed linux hash (char-for-char, deployah verified), applied-state phase:applied version:10 PROMOTED (bytes-assert passed → no rollback), ZERO manual bounce, pump live throughout, daemon never dropped. The current_exe()-follows-rename bug that killed v0.4.1 on Linux is DEAD — v0.4.2 canonical-path-respawn broker delivered genuine seamless on prod hardware. Closes the whole respawn-path arc. hfenduleam (Win) ALSO seamless: broker 17232 HELD, brain 39820→56744 live, gen 2→3, exe_hash 984298ba == signed win (deployah verified), promoted v10 no rollback no bounce — note hfenduleam pump had already peer-propagated+staged counter 10 from kitsubito/origin, applied staged set directly (correct; not the v0.3.2 single-binary re-stage floor trap). Both runner nodes seamless. ENLYZEAM (3rd node, Windows, target 984298ba): rides NATURAL pump propagation (option b) — deliberate-proof already banked on kitsubito; natural ride validates steady-state auto-cadence path; neither deployah nor todlando has enlyzeam direct access (operator does, drove it via at-logon scheduled task in restoration). Async-confirm enlyzeam exe_hash==984298ba next cadence to fully close. M9 fleet effectively SHIPPED.

**How to apply:** deployah recommended the DELIBERATE kitsubito apply to capture the proof on purpose; it worked. deployah on-call for remaining hash checks. Publisher path proven 4× (counter 7/8/9/10) — seed `SPT_RELEASE_SEED` resident in ENV (64-hex), inherited by fresh pwsh children; Monitor runs bash not pwsh. Next track: spt-claude-code. Backlog: 2 DEFERRED.md gaps (CI-flake applyhost socket-bind kitsubito; REQ-UPD-8 single→set same-version re-fetch rollback-floor) + hfenduleam Win-CI disk-bloat gotcha (LNK1318/os-error-112 = disk full, cargo clean, not code) — all mint REQ before fixing.