Pinky PPPP: one glance, one truth — the stuck-lanes window fix (v1.99.0, #58) #134

Merged

ClaudiuSchuster merged 1 commit from oss-oo/pinky-pppp-window into main

2026-10-16 23:40:12 +00:00

ClaudiuSchuster commented

2026-10-16 23:39:26 +00:00

Owner

Owner finding #58 (2026-10-17 via gen-4): the header STUCK-LANES indicator showed the all-time stock while the tables listed only the selected window — the header contradicted the tables for the same glance.

The cure — ein Blick, eine Wahrheit:

One authority: the tables own window machinery (WIN_MS/winSel/windowAnchor/windowKeeps). The new lane twin laneKeepsWindow cites the SAME windowKeeps — a lane counts iff the tables would list at least one of its member rows (lane.memberIds ride the payload; the server family grouping stays the only grouping).
Straddle reading (matches the tables, documented in the dossier): every stuck-class evidence is a LIVE state, so a straddling presumed-dead row (reads running) counts at EVERY window (running rows never leave the tables); a stuck state on rows the window aged out (a stuck bell at a long-idle door) drops out exactly when its rows drop out of the tables. An evidence-age reading would be a no-op (every stuck state is live now) — row-visibility is the tables semantics and the surfaces may not disagree.
One derivation: windowedFleet (fleet module, pure; spec installed once via setFleetWindow) — chip label+tooltip, slim-line alert+stat, and the vitals panel render the windowed census; windowDropped names the fold; windowLabel stamps which window the numbers describe; all preserves the all-time stock verbatim. Bells tally (live machinery) and verdict alarms (starvation, idle fleet, master door) stay window-free by design; the server census itself is untouched (W14).

Evidence: vault rig tests/rigs/pppp_window_cdp.py 15/15 legs (firing/switching/negative/straddle/tables-agreement/server-invariant) + make rig-pppp-window; suite tests/test_round_pppp.py (1703 total incl. node harnesses + tripwires + opt-in E2E); six Super-Linter lanes green; LLLL specimen re-proven 16/16 on the shared tests/rigs/rigkit.py kit.

**Owner finding #58** (2026-10-17 via gen-4): the header STUCK-LANES indicator showed the all-time stock while the tables listed only the selected window — the header contradicted the tables for the same glance. **The cure — ein Blick, eine Wahrheit:** - **One authority**: the tables own window machinery (`WIN_MS`/`winSel`/`windowAnchor`/`windowKeeps`). The new lane twin `laneKeepsWindow` cites the SAME `windowKeeps` — a lane counts iff the tables would list at least one of its member rows (`lane.memberIds` ride the payload; the server family grouping stays the only grouping). - **Straddle reading** (matches the tables, documented in the dossier): every stuck-class evidence is a LIVE state, so a straddling presumed-dead row (reads running) counts at EVERY window (running rows never leave the tables); a stuck state on rows the window aged out (a stuck bell at a long-idle door) drops out exactly when its rows drop out of the tables. An evidence-age reading would be a no-op (every stuck state is live now) — row-visibility is the tables semantics and the surfaces may not disagree. - **One derivation**: `windowedFleet` (fleet module, pure; spec installed once via `setFleetWindow`) — chip label+tooltip, slim-line alert+stat, and the vitals panel render the windowed census; `windowDropped` names the fold; `windowLabel` stamps which window the numbers describe; `all` preserves the all-time stock verbatim. Bells tally (live machinery) and verdict alarms (starvation, idle fleet, master door) stay window-free by design; the server census itself is untouched (W14). **Evidence**: vault rig `tests/rigs/pppp_window_cdp.py` 15/15 legs (firing/switching/negative/straddle/tables-agreement/server-invariant) + `make rig-pppp-window`; suite `tests/test_round_pppp.py` (1703 total incl. node harnesses + tripwires + opt-in E2E); six Super-Linter lanes green; LLLL specimen re-proven 16/16 on the shared `tests/rigs/rigkit.py` kit.

ClaudiuSchuster added 1 commit

2026-10-16 23:39:26 +00:00

Pinky PPPP: one glance, one truth — the stuck-lanes window fix (#58 )

Checks / Functional checks (pull_request) Successful in 19s

Details Explain this check

a7ba44f02b

The header's stuck-lanes census showed the all-time stock while the
tables listed only the selected window (24h/3d/7d) — the header
contradicted the tables for the same glance. The cure:

- ONE authority: the tables' own window machinery (WIN_MS / winSel /
  windowAnchor / windowKeeps). The new lane twin laneKeepsWindow
  cites the SAME windowKeeps — a fleet lane counts iff the tables
  would list at least one of its member rows (member ids ride the
  payload as lane.memberIds, the server's one family grouping).
- Straddle reading (matches the tables): every stuck-class evidence
  is a LIVE state, so a straddling presumed-dead row (reads running)
  counts at EVERY window; a stuck state on rows the window aged out
  drops out exactly when its rows drop out of the tables.
- ONE derivation: windowedFleet (fleet module, pure; the window spec
  installed via setFleetWindow) — chip, slim alert and vitals panel
  render the windowed census; windowDropped names the fold; 'all'
  preserves the all-time stock verbatim; the server census itself is
  untouched (display fold, never a redefinition of stuck-ness).
- Vault: tests/rigs/pppp_window_cdp.py (15 legs, make
  rig-pppp-window) + the shared tests/rigs/rigkit.py kit both
  specimens import (LLLL re-proven 16/16 on it).
- Suite: tests/test_round_pppp.py (server memberIds legs, two node
  harnesses, wiring tripwires, opt-in vault E2E); v1.99.0.

ClaudiuSchuster merged commit 36921b1cf4 into main

2026-10-16 23:40:12 +00:00

ClaudiuSchuster deleted branch oss-oo/pinky-pppp-window

2026-10-16 23:40:12 +00:00

ClaudiuSchuster referenced this pull request from a commit

2026-10-16 23:40:13 +00:00

Pinky PPPP: one glance, one truth — the stuck-lanes window fix (v1.99.0, #58) (#134)