Active user-request ledger — 2026-09-09¶

Historical record. This preserves the source, tests and limitations recorded at the time; it is not current setup guidance. See the documentation index.

This is the working checklist for the current follow-ups. Source implementation, automated verification, physical acceptance, and public release are separate milestones. Nothing in this ledger implies that an in-progress build is installed or that the published RC4.5.1 assets were replaced.

Requests and acceptance criteria¶

Final automated verification for this source batch¶

Release publication is authorized as VIIPERRC4.5.2 — Joy-Con 1 & Nintendo Options. The final source corrections are implemented and verified below; complete package/hardware acceptance and the exact-source GitHub release workflow remain separate delivery gates.

• Local DS4Windows Release x64 complete suite: 4,797 passed, 0 failed, 11 opt-in skipped (4,808 total). Result: DS4WindowsTests/TestResults/2026-09-09-nintendo-updater-actions-full-final.trx.
• Final suite after the rumble fixes, universal impulse routing, and disabled-state styling: 4,877 passed, 0 failed, 11 opt-in skipped (4,888 total). Result: DS4WindowsTests/TestResults/2026-09-09-rc452-release-source-final.trx. The immediately preceding full run also passed 4,877 tests. Actual dark/light WPF renders were inspected; the route option fits narrow layouts and visibly dims when unavailable. All allocation assertions remained enabled.
• The preceding run found one stale upright-right stick-assist test expectation. The producer routes upright-right to RX/RY and sideways-right to LX/LY; the fixture now tests both orientations, fractional precision, unused-axis isolation, and orientation-change baselining. No production change was made to satisfy that stale expectation.
• The 11 opt-in cases require live process audio capture or real Go/DS4 retirement end-to-end setup. Allocation assertions remained enabled and passed.
• Updater's full 245-test result and actual release ZIP fixture are documented separately in DS4Updater/docs/portable-update-2.0.5-validation.md in the sibling repository. Updater publication is recorded below. The DS4Windows batch has not been installed into Program Files or publicly released; the later portable handoff is recorded below.

Additional release gates¶

• Original Joy-Con logical Disconnect now claims both exact paired halves without recursively disconnecting unrelated/reconnected controllers. Unexpected single-half physical loss still retains the survivor. Session-only pairs can unlink despite a damaged pair store; actual saved-pair removal remains persistence-checked. Focused validation: 58 passed, DS4WindowsTests/TestResults/2026-09-09-original-joycon-disconnect-release-gate.trx.
• Updater now recognizes VIIPERRC4.5.2 as binary 5.0.5.2. Full updater validation: 245 passed, no skips, including the immutable released ZIP. Both compressed x64/x86 publish checks passed. Version 2.0.5 is published at https://github.com/hbashton/DS4Updater/releases/tag/v2.0.5; the latest-release API returns v2.0.5.
• Updater source is committed and pushed to master at d2e9b3ae316c52b8320055e2acc063da42d05905. Exact-source GitHub Actions run 34374372375 passed the full test job and both platform builds. This is CI evidence, not publication or launched-worker end-to-end acceptance.
• Updater release run 34376111898 succeeded from that exact tagged commit. Both public assets were downloaded and matched the pre-release CI artifacts, PE version 2.0.5, and ProductVersion commit. SHA-256: x64 F1A38ED7C958A7837259812094B9E816B6478EC76E8FEC2B1D998756A44093DB; x86 B8369CDB38519845A94E89B6B56C1F361EFA1278C5605CEF85AD9CAA836D806D. The DS4Windows safe bootstrap still needs delivery in RC4.5.2; this does not retroactively enable old portable callers.
• Hades II hands-off pulsing, Switch 2 Pro Bluetooth / virtual DualSense: the installed RC4.5.1 binary running at dump capture matched the published release. User-authorized heap dump captured at 10:56:55 local time shows retained compact motors [43,0], native report ID 0x02, flags 0x0c/0x57/0x00, zero native motors, both physical trigger gates false, and an active canonical BodyLow value 11051 with sustained refresh. The full dump shows no PCM carrier or media callbacks, not an active silent PCM stream. Four consumed ordered-control slots retained the same stale compact motor value. No backlog, write failure, or observed write past TTL was found. This is direct evidence of stale compact rumble being admitted by the Nintendo translation. Fresh native source-selector admission now suppresses that value and prevents accumulated media snapshots from restoring it; 48 focused tests passed. The later 86- and 126-test integration batches below supersede that initial verification. Dump and private diagnostic logs remain local on Desktop and must not be included in release/source assets.
• The bounded loopback packet-monitor attempt produced zero packets, so it is not evidence about sender behavior or rumble contents. Its trace/filter were stopped and removed; the memory dump supplied the useful evidence instead.
• Follow-up review identified two additional replay cases before release: stale media must not restore a nonzero motor after a fresh zero-motor stop that retains compatibility mode; mixed PCM/adaptive output must not repeat old PCM samples as a sustained adaptive effect. Both production fixes passed 126 focused tests, including the actual original Joy-Con handler and physical maintenance writer. Original Joy-Con source selection integration previously passed 86 focused checks. Existing compatibility hold semantics remain documented and unchanged: media can maintain source liveness, but cannot create new motor intent or undo a stop.

Independent CI and portable handoff¶

• Source checkpoint 554914381794c467afcf8ea63a52a8ce78317100 was pushed to main. Independent CI run 34378088514 found three failing variants of RealOutputWorkerDrainsNeutralBeforeDeviceStopReturns. Publication was held. Their stop-delivered assertion passed, but the expected second packet did not: the fixture had no current Joy-Con registration, so the production authority guard had already substituted a neutral first packet. A new explicit first-packet-active assertion reproduced the invalid precondition for all four Joy-Con variants. The fixture now uses a real current coordinator registration and scoped output settings; the original active-drain assertions remain. Four separate tests cover unregistered-output neutralization. No production code was changed for this correction. Initial active publication is prepared before starting the worker, excluding the legitimate takeover-neutral setup from the blocked-active-write race. The already-idle test uses deterministic manual scheduling, since shutdown may legitimately enqueue another neutral. Focused lifecycle/feedback checks: 128 passed. The real active-drain test passed 100 repeated cases (20 runs across five transport variants). Final reviewed, unfiltered local full suite: 4,881 passed, 0 failed, 11 existing opt-in skips (4,892 total), including allocation assertions. Result: isolated_results/rc452-ci-rumble-fixture/rc452-ci-rumble-fixture-reviewed-full.trx. The final source commit still requires its own successful GitHub CI before tagging; these local results are not a substitute for that gate.
• The complete portable candidate built from that checkpoint was verified before launch: 551 files, 296 dependency assets, 23 language satellites, matching VIIPER aliases and the authorized Xbox persona. It started from its Desktop folder with separate lab data at 11:46 local time. The backend became ready; this alone does not establish controller or Hades II acceptance. Installed files and released RC4.5.1 assets were not replaced.
• The candidate's compiled updater bootstrap accepted live v2.0.5 metadata and the verified public updater executable, then formed the safe RC4.5.2 launch request. Its injected process boundary did not start a worker. This extends bootstrap verification, not launched-worker end-to-end update acceptance.
• A separate initially updater-free Desktop fixture also exercised the production bootstrap's actual fresh download of public 2.0.5. Origin, size, hash, PE identity, staging cleanup and safe launch arguments passed. An extra PowerShell post-check initially needed a typed-string correction; it was completed against the same downloaded bytes without a second asset download. No updater worker or controller app was launched by this check.
• The next exact-source CI run 34380417498 at 6e99c44bef78dddd0c1ba440e62a1e9066f71643 passed shutdown tests but failed the first registered Joy-Con allocation case: 4,752 bytes versus zero. Publication remained held. A new native-profiler capture of the same warmed writer reproduced counter-only jumps during background-GC preparation with zero objects and a valid deliberate-allocation control. The test-only measurement correction and limits of attribution are documented in Nintendo allocation release gate. No production GC policy, writer code or zero-allocation tolerance was changed.
• After that correction, the ordinary no-profiler/no-diagnostic run passed 144 focused tests and 4,886 unfiltered full-suite tests, with zero failures and the same 11 opt-in skips (4,897 total). Same-writer positive controls prove real allocations still fail the zero-byte gate; the separate native pressure control covered 1,280 strictly zero measurement windows. Final source must still pass its independent CI before release tagging.

Post-release follow-ups¶

• Reddit RC4.5 Stop/read-failure-995 hang: a matching fallback-worker/service-lock deadlock was reproduced with four failing-first tests. Source fix implemented; 152 focused and 4,915 full-suite tests passed, zero failures and 11 existing full-suite opt-in skips. Added 29 regressions covering removal/cancellation, reconnect ownership and hotplug publication. Not installed or published; reporter hardware confirmation remains open. See Stop/removal regression.
• Consistent small Joy-Con L vibration-test stutters remain open. Earlier traces did not establish an active sustained-rumble window; they are not evidence that the stutter is fixed. The Stop report takes priority in this pass; running apps and controllers are left untouched.

RC4.5.3 publication and corrected report scope¶

This update supersedes the earlier pending-publication status snapshots above.

• VIIPERRC4.5.3 is published as an unsigned prerelease: release ID 385874559, 13 assets, source d2e743a2d871910d2e4733c6875ab87d47096286, Windows version 5.0.5.3. The complete portable package retains VIIPER 0.1.3-rc4.5. Post-publication workflow 34412155943 completed successfully from that exact source.
• DS4Updater 2.0.6 is stable and published. The existing compiled RC4.5.2 bootstrap successfully downloaded it into an initially updater-free isolated fixture, verified its origin, size, SHA-256 and PE identity, and reached a capture-only launch boundary with the safe RC4.5.3 arguments. The compiled 2.0.6 resolver also accepted the unmodified public RC4.5.3 metadata and verified the actual release receipt, ZIP, EXE/DLL identity and package marker; no simulated publication envelope was used. Inputs were unchanged and no updater worker, application or broker was launched. Local proof files in the sibling updater repository are isolated_results/updater-2.0.6/PUBLIC-RELEASE-METADATA.json, PUBLIC-RC453-UPDATER-VERIFICATION.json in that same directory, and bootstrap/PUBLIC-BOOTSTRAP-VERIFICATION.json. This is bootstrap/package acceptance, not launched-worker end-to-end or controller hardware acceptance.
• The reporter clarified that read failure 995 occurred before pressing Stop and that RC4.3 works while RC4.5 fails. The earlier explanation treating the initial failure as cancellation caused by Stop was not established and does not fit that order. RC4.5.3 fixes the reproduced Stop/removal deadlock and exact reconnect-ownership faults; it does not establish or claim a fix for the reporter's initiating read failure. Controller model/transport and the initial cancellation source remain unconfirmed. The user explicitly authorized this release while that separate RC4.3-to-RC4.5 investigation continues; a further correction belongs in a later release if reproduced. Joy-Con rumble stutter also remains open, without a hardware-fix claim.

RC4.5.4 recovery correction¶

• The corrected failure-before-Stop chronology now has a failing-first integration regression through the actual reader, DualSense lifecycle logger, removal handler and public service Stop boundary. The old synchronous handler deadlocks; the queued RC4.5.3 handler passes. This does not attribute the initial 995 to Stop.
• A separate DS4 Bluetooth terminal-failure defect was reproduced: a new forced output could block removal after the failure log. RC4.5.4 removes that terminal send, while retaining ordinary issue #84 control-pipe routing and all unexpected error diagnostics. Five new tests include blocked-output and audio-lane cases.
• Versioned Release/x64 validation: 4,927 passed, zero failed, 11 existing opt-in skips, unfiltered and including allocation assertions. Exact-source CI and complete release artifact checks are still required before publication.
• Windows binary version is 5.0.5.4, release tag VIIPERRC4.5.4. Published RC4.5.3 and updater 2.0.6 remain immutable; no broker change is needed. The initiating native cancellation and Joy-Con rumble stutter remain unconfirmed. See corrected chronology and recovery evidence.

RC4.5.4 publication completed¶

This supersedes the pending CI/publication gates in the preceding snapshot. VIIPERRC4.5.4 is published from f15c3001ccb5ba577a20c9174e252e2835383631, with all 13 verified release assets. Exact-source CI 34413872665, draft build 34414663236 and post-publication verification 34415309349 all passed. The complete installer and ZIP are also in Desktop/DS4Windows-RC4.5.4-Release; neither was run.

The unchanged updater 2.0.6 resolver accepted untouched public RC4.5.4 metadata and the real receipt/ZIP/PE/marker identities. Twenty further repetitions of the six new recovery tests passed (120 cases). Source, test scope, hashes and limits are recorded in the published recovery evidence. The initiating 995 and Joy-Con rumble stutter are still unconfirmed; publication does not turn those open questions into hardware-fix claims.

RC4.5.5 setup and startup correction¶

• The user's failed RC4.5.3 setup was a stale unmarked RunVIIPER task conflict, not a demonstrated USB/IP driver failure. The exact old task/settings were backed up and repaired, and the approved Retry completed successfully.
• Permanent fixes separate portable runtime preference from installed startup, preserve task ownership, align priority, update owned tasks in place, and recover narrowly verified legacy registrations only after durable backups. Unrelated tasks and unsafe backup paths remain protected.
• Final upgrade review also reproduced and corrected a logon repair loop: with installed startup enabled, runtime selection now matches the verified canonical startup broker and normalizes the stale portable cache. Four failures reproduced the mismatch before 123 focused cases passed with the fix.
• Versioned Release/x64 suite: 4,958 passed, zero failed, 11 opt-in skips, including allocation assertions. Startup recovery, real backup-file and 22 current-attempt diagnostic cases passed; the bootstrapper build is clean. These three script gates are now required in CI.
• Prepared tag VIIPERRC4.5.5, Windows version 5.0.5.5, title Reliable Setup & Startup. Exact-source CI, release artifacts and updater acceptance remain publication gates. VIIPER and updater 2.0.6 are unchanged.
• See root cause, safety boundaries and evidence. The initial 995 cause and Joy-Con rumble stutter remain open; this installer correction does not change input or haptic transport.

RC4.5.5 publication completed¶

The preceding publication gates are complete. VIIPERRC4.5.5 is published as Release Candidate 4.5.5 Hotfix — Reliable Setup & Startup, from immutable source 2b24e0994819ca939fa00e44e15e0509ae7f0053. Exact-source CI 34432374338, draft build 34433022241, and post-publication verification 34433548343 passed. All 13 public assets were checked against the successful draft's byte identities. The untouched public release metadata and actual ZIP/receipt/PE identities were accepted by the compiled, unchanged updater 2.0.6. This was package acceptance, not execution of the updater worker. Complete downloads are retained in Desktop/DS4Windows-RC4.5.5-Release. Published assets are not to be overwritten.

New haptics reports: investigation in progress¶

• Joy-Con / Switch 2 Test Heavy and Test Light: user reports consistent slight stutter and crackling, over USB and Bluetooth. Live inspection found installed RC4.5.4, a Bluetooth Joy-Con 2 (L), and profile ds. The test-preview lane bypasses the native DualSense feedback queues. No physical fix is claimed. Two bounded event-263 observer sessions returned no events (one idle, one during Test Heavy); those captures do not establish absence or timing of physical writes. The subsequent event-402 capture recorded 1,438 active writes, one unchanged waveform, continuous counters, median 15.5632 ms and maximum 17.5856 ms, with no ETW loss. The user confirmed this captured run still stuttered. This is host-write evidence, not motor/radio delivery proof.
• GTA V Enhanced rapid-fire haptics: third-party report affects physical DualSense emulating DualSense over either USB or Bluetooth. Failing-first fixtures reproduce loss of closely spaced exact native commands in VIIPER's latest-state latch, and eviction/expiry in DS4Windows' ordered dispatch queue. Broker bounded admission and downstream retained-head fixes now pass focused regression tests and independent review. Full combined validation and a complete private hardware candidate remain required before delivery. These are confirmed source defects, not yet a captured attribution of the reporter's GTA session. They are not included in RC4.5.5.
• Preserve finite PCM/one-shot behavior, immediate explicit stops, independent input processing, bounded queues, and lifecycle cancellation. Do not solve command loss by silently overwriting, growing unbounded queues, or routinely disconnecting controllers. Validate any candidate portably, not in Program Files.

Current execution boundaries¶

• Preserve the earlier source and live-preview changes, now checkpointed in the release source commit; later test corrections do not replace unrelated work.
• Source and fixture tests did not restart the user's apps. The user then closed DS4Windows cleanly at 11:33:46 local time and quit VIIPER for the separately requested portable hardware handoff. Neither process nor its 3241/3242 listeners remained at the handoff preflight.
• Do not install tests into Program Files, overwrite released assets, or claim physical validation from synthetic tests.
• Original Joy-Con HD rumble has different temporal/packet limits from Joy-Con 2; document approximations, preserve sides/bands/peaks where representable, and never send guessed Switch 2 commands to original hardware.
• Updater 2.0.5's public publication and the next DS4Windows delivery remain distinct from local builds and validation.