carlos / platform Public

Commits

• 7caf2c22b Merge remote-tracking branch 'origin/main' into region-ha-two-box-build Paul Campbell 1 week ago
• 8ef2c8a23 Merge remote-tracking branch 'origin/main' into region-ha-two-box-build Paul Campbell 1 week ago
• 3b895b42f sim: a dirty return ends the source's files protection, and its install parks the pre-move generations, only when the target admitted the move (Astra joint round 4 item 2) Paul Campbell 1 week ago
• eaa08175d sim: a rollback keeps the source region the files home until the source's return, and a dirty return's install park retires the pre-move generations (moves a5e8c13f, D6) Paul Campbell 1 week ago
• 372f7a4b4 sim: the probe stop stops the writers, flushes and records the positions before root's stop (ruling 87), so the cold first serving wake loses nothing the probe session acknowledged Paul Campbell 1 week ago
• 86294cb85 sim: expected outcomes for reloc-admits-probe-only (missed fidelity guard, appended) and reloc-stuck-refuses-admission at two boxes (redundant alone since joint round 2) Paul Campbell 1 week ago
• 917771bbf sim: the first serving wake after a completion is cold, and a release closes the session's launches at the stop and admission only at Collect Paul Campbell 1 week ago
• ded693148 sim: a delete or purge over a completion box whose termination fence verifies passes the barrier (ruling 86), and the operator's declare-lost crutch is gone Paul Campbell 1 week ago
• d2e97bf13 sim: expected-outcomes.txt records the C1 and C11 fallback rules as missed fidelity guards, with their direct tests Paul Campbell 1 week ago
• 363055fee sim: witnesses and direct tests for the completion protocol, the catch-up hold and the C1/C11 fallbacks, and the residual and purge-deadlock scenarios Paul Campbell 1 week ago
• 52ab8507d sim: a relocation ends as 1b-next builds it (probe stop, completion pass, barrier, N6), a delayed source catches up its row hold, and the C1/C11 fallbacks are today's behaviour Paul Campbell 1 week ago
• 295f31a63 Merge origin/main into region-ha-two-box-build (before the Task 11 joint round) Paul Campbell 1 week ago
• e814ece38 sim: a committed landing's failover keeps the landing's op id Paul Campbell 2 weeks ago
• cef63683c hostlife: files-home-move-phases text names the commit boundary and both witnesses (joint round 3 P3) Paul Campbell 2 weeks ago
• 54c0b6305 Astra joint round 3 (J1 fix round 2): unedited Paul Campbell 2 weeks ago
• cc0d678e7 Joint round 3 brief (J1 fix round 2) Paul Campbell 2 weeks ago
• 00bd0abe0 sim: the source region keeps the files live set until the move commits, and INV-30 checks every servable region's next pin survives Paul Campbell 2 weeks ago
• 25c7c9e14 Astra joint round 2 (J1 fix): unedited Paul Campbell 2 weeks ago
• 2e956f4e6 Joint round 2 brief (J1 fix only) Paul Campbell 2 weeks ago
• dfab457ca sim: the files home follows moves §5.1's authoritative region, so a park queued during a move keeps the live manifest in the source region until the flip Paul Campbell 2 weeks ago
• 55f03a6c7 sim: the files live set is region-qualified, so a manifest PUT that lands in a former region after a move's cleanup is retired by the sweep there Paul Campbell 2 weeks ago
• a2ef02034 Astra joint round 1 (two-box model + 1b-next stage A): unedited Paul Campbell 2 weeks ago
• 10ac836d5 docs: the two-box brief names model commit db4b2a09 and states the final review's fix wave: the 151261fe mirror by region, park-evidence-every-box and its witness, and the one release-and-journal helper Paul Campbell 2 weeks ago
• db4b2a090 sim: expected-outcomes.txt records reloc-name-drops-move-target as missed at one box, with its direct test; park-evidence-every-box needs no line, since its witness catches it in either shape Paul Campbell 2 weeks ago
• cdbfa1f73 sim: reloc-name-drops-move-target is a liveness guard: its witness is dropped and expected-outcomes.two-box.txt records it missed, with TestConvertedMoveDropsTheMoveTargetFromTheBarrier as its direct test Paul Campbell 2 weeks ago
• 819a76c41 sim: stale comments corrected: reloc-pin-evidenced-empty is caught by its witness, relocRelease's steps name the flush and receipt rules, startFailover names its target at the start, and mutate.sh's STALE verdict cites the merged expectations at two boxes (final review minor 7) Paul Campbell 2 weeks ago
• 6b52cf73a sim: exec-writer and reloc-replacement-table are rare 30000 at two boxes, caught 1 and 2 times in 3,000, so a 3,000-seed sweep that misses them by chance does not fail (final review minor 6) Paul Campbell 2 weeks ago
• 0c3deba48 sim: the one-box identity tests pin boxesPerRegion to 1, so their traces are the one-box shape's whatever HOSTLIFE_BOXES says (final review minor 5) Paul Campbell 2 weeks ago
• 028150d03 sim: the operator's replay of a stuck relocation excludes a released target, never a released source, which ruling 13 makes eligible again (final review minor 4) Paul Campbell 2 weeks ago
• 02ba19b40 sim: the region-wide stop-evidence wait before a two-box park-replica or park-superseded is the named rule park-evidence-every-box, with a witness of seed 819's shape that violates INV-10 under its mutation Paul Campbell 2 weeks ago
• 66379fb31 sim: naming a converted move's same-region target takes the move's own target off the participants in the same CAS (rule reloc-name-drops-move-target, mirroring 1b-next 151261fe); a stopped move target no longer holds the relocation at pinning Paul Campbell 2 weeks ago
• 04c9487da sim: relocReleasing and relocProbeFailed journal their release through relocJournalRelease, the one copy of the release-and-journal CAS (final review Important 3); two-box COUNTERS and traces unchanged Paul Campbell 2 weeks ago
• 3f86896b3 docs: the two-box brief names the model commit after the expectations fix Paul Campbell 2 weeks ago
• 11837acab docs: the two-box brief states the raw sweep result and its reconciliation, points at the 1b-next ledger with Task 7's two open concerns, gives the worktree paths, and cites the committed diagnosis Paul Campbell 2 weeks ago
• 8a1e8508c docs: the diagnosis of the unmutated two-box sweep failures, for the joint round Paul Campbell 2 weeks ago
• 9ef58e4b0 sim: the one-box expectations file is restored; create-choice-once and reloc-replacement-table are dropped for two boxes in expected-outcomes.two-box.txt, where the sweep catches them Paul Campbell 2 weeks ago
• 08c77007e docs: the joint-round brief for the two-box model (L18) against plan 1b-next rev 19, round 1 Paul Campbell 2 weeks ago
• 5766a0e1e sim: expected-outcomes.two-box.txt overrides the one-box expectations at HOSTLIFE_BOXES=2 (timer-canonical and source-readiness missed at two boxes, exec-writer caught); mutate.sh merges it, one-box reads the base file unchanged Paul Campbell 2 weeks ago
• df1a5f20a sim: expected-outcomes.txt drops create-choice-once and reloc-replacement-table: the two-box sweep catches both once the baseline is clean (seeds 9, 250, 1817 and 2854) Paul Campbell 2 weeks ago
• cf566b41e sim: the two-box baseline's three causes: INV-10 exempts files keys a park kept live (the at-rest sweep owns them once obsolete), a region's park-replica waits for every box's stop evidence, and an operator-named source that is also the target is replaced when it is fenced Paul Campbell 2 weeks ago
• cf1a2bd52 sim: mutate.sh reports the shape (HOSTLIFE_BOXES), runs the region HA rules as a group and lists without running; expected-outcomes.txt lists the region HA rules caught only in combination or bounded by liveness, loss, determinism or the operator, with the test that shows each Paul Campbell 2 weeks ago
• 2ea08a48e sim: Astra's round 8 on the two-box model: root's stop receipt is tracked apart from the recorded positions, with the flushing controller's identity; only that controller finishes from the positions, and after a restart the receipt is needed too Paul Campbell 2 weeks ago
• 33f684bbe sim: Astra's round 7 on the two-box model: only a terminated box's volume evidence discharges its local cleanup, so a stopped box keeps it, and INV-26 checks the operation's discharges; the operator stops a target only when its release is unrecoverable and otherwise resumes it Paul Campbell 2 weeks ago
• a2064475d sim: Astra's round 6 on the two-box model: a release needs the session's own replicator and files, or the positions recorded for its stop, and is otherwise refused until the deadline; the operator stops a target that can never release Paul Campbell 2 weeks ago
• 2b8f5c1e9 sim: Astra's round 5 on the two-box model: a relocation participant's fence that cannot be consumed yet holds the regional phase, as Worker.Step returns before its phase switch, and the worker fences the source, then the target, then any other box Paul Campbell 2 weeks ago
• 6b991069d sim: Astra's round 4 on the two-box model: a target named before any pin on every path, PinFailover as the phase predicates' starting, and a failover started only through Worker.Start's guards; the drain start in Worker.Start's order Paul Campbell 2 weeks ago
• 843991013 sim: Astra's rounds 2 and 3 on the two-box model: an initial failover's target acknowledges before its pin is selected (PinFailover), while converted moves, nested failovers and replays pin first, a crash between the published marker and the open replication, the adverse volume under a failover of the current executor, and the pinned manifest's generation parked at completion Paul Campbell 2 weeks ago
• 8b673c998 sim: Astra's round 1 on the two-box model: stuck refuses admission, region HA checks a sole box, durable acknowledgements gate every transition, deadlines from the fence record, the exact phase matrix, a discriminating test per guard, files objects with origins and region binding, moves' parks keep the files live set, committed-landing and record-race fences, the controller record that blocks updates, the forced open boundaries, and the tie-break over the real host Paul Campbell 2 weeks ago
• b652b8ff0 sim: plan 1b-next rev 19's fixes for the model's findings: a nonempty database pins only files that carry instance.key (the L21 window fails safe), probe-failed and a dead target with no spare go stuck after the deadline with the fence consumed, and every unopened P is recorded on every path and parked at completion Paul Campbell 2 weeks ago
• 3cb7f53c8 sim: region HA scenarios: every L18 case in a one-box and a two-box region, claim §11's extras, twenty witnesses for the relocation, fence and files guards, direct evidence for every bound, and reproductions of findings F1 to F3 Paul Campbell 2 weeks ago
• 25982b0f1 sim: moveTo's doc comment sits on moveTo, not before createExecutor (alignment for Task 5's patch) Paul Campbell 2 weeks ago
• 55ab9c258 sim: the files publisher from sequence 1, P's prepared manifest at 0, the release's final manifest, the files pin by exact version over the admitted chain, the files live set with the chain's newest manifest, timers declared only when no landed manifest carries them (INV-28), and no dead box's disk read in the two-box shape Paul Campbell 2 weeks ago
• 3f51773b5 sim: termination by the cloud, automatic enrollment, the regional worker's fence (lost, timers declared, regional chores passed to the region, local chores discharged only on qualifying volume evidence), failover from serving, source loss, the replacement table, the phase-sensitive exhausted fence and no replacement owing a park-local Paul Campbell 2 weeks ago
• 9fe763438 sim: relocation by drain through region HA §5.8's phase table with plan 1b-next rev 16's guards: the release and its deadline, the barrier on every phase change, a pin bound to its claim version and to evidenced emptiness, EnsureLocal and VerifyReceipt around P's first admission, prepared_opened journaled, three probe attempts counted first, released coverage, the row cases and the replacement cap Paul Campbell 2 weeks ago
• beebe873e sim: match Task 1's tree to the dry run's pre-image Paul Campbell 2 weeks ago
• 177192087 sim: one or two boxes per region (region HA L18): HOSTLIFE_BOXES and TestHostLifecycleTwoBox, choose() for creates and move targets recorded once, adoption's one proven owner, regional chores for any box of the region, replacements to the region's minimum; the one-box sweep is byte-identical Paul Campbell 2 weeks ago
• 45d97e011 Region HA two-box model: plan rev 11 and its Astra reviews (from region-ha-hostlife-two-box 8f95c4e0) Paul Campbell 2 weeks ago