Skip to content

RC4.3 to RC4.4 native DualSense haptics regression audit

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

Evidence checkpoint: September 5, 2026. The user reports that Hades II's advanced feedback worked with a physical DualSense emulating a DualSense in RC4.3, but not in the released RC4.4. This audit has reproduced and corrected a native-media fallback defect in software. It has not reproduced the user's game session or established that this is its only cause. The physical connection type and whether ordinary rumble/adaptive triggers survived remain unconfirmed.

Exact release boundary

  • VIIPERRC4.3: 3579450dc8f50a74d9532e711249589c732c460b.
  • VIIPERRC4.4: 4d41a613c771588e05558a386c2ca5591b6949c3.
  • Bundled VIIPER changed from v0.1.0 to v0.1.2. Its V5 framing, media generation, output writer and audio conversion diffs were inspected as well; this checkpoint does not exonerate every broker/USB-IP change.

Reproduced failure

ApplyAtomicAudioHapticsFeedback calls ApplyFeedback with freshNativeOutput: false: the frame is an audio interval, not a new game HID command. Its compatibility motor bytes can be zero while its waveform is nonzero.

  1. TryApplyBluetoothCombinedHapticsOutputReport can return false. USB targets cannot consume this Bluetooth carrier. On Bluetooth, an occupied bluetoothCombinedTemplateUpdateClaimed flag rejects immediate template publication even though the controller identity and media payload are valid.
  2. 5fcc5ab864a25536179bda5ca7efef29c101bb5a, included in RC4.4, changed raw native forwarding to require caller-owned scratch storage. The media callback supplies none. The old fallback prepared a raw native report itself; the new fallback returns false at this point.
  3. Execution then reached SetDevRumble, treating the media frame's zero motor bytes as an explicit local rumble command. This advances the physical local rumble generation. f2807747609962566fd73b85e7874e5ef20e679c, also in RC4.4, introduced the template claim that supplies a deterministic BT rejection scenario for this path.
  4. The BT compositor explicitly permits a new local rumble generation to override native motor state. Its MergeLocalRumbleIntoV5AudioSnapshot sets common flag0 bits 0x03 and copies the compatibility motor levels. This is not an innocuous zero in an audio sample: it selects the classic rumble path. The Sony-authored Linux PlayStation driver independently identifies HAPTICS_SELECT as the classic-rumble selection in dualsense_output_worker (checked September 5, 2026).

The physical consequence is strongest on the BT route above. The USB tests prove the erroneous local-command publication too, not the complete USB audio/rendering failure. Game-level causality still requires an actual trace and replay on the user's transport.

Narrow correction

After native media transport attempts, a compatible physical DualSense's non-fresh audio frame is consumed without falling through to raw-state replay or compatibility rumble. Genuine fresh HID/control reports retain their current route. Other physical controller families retain their existing waveform-to- rumble translation. No packet format, timer, driver, broker or profile setting was changed. This prevents a rejected media interval from changing feedback mode; it does not claim that rejected intervals were physically delivered.

Validation

DualSenseNativeMediaFallbackTests invokes the production media callback using an unopened HID object, a cached test identity, no physical workers, and no audio capture. The BT case holds the real template-claim flag to model contention deterministically. Both primary virtual pads and audio sidecars are covered. Global test state is restored in finally.

  • Before correction: all four USB/BT media cases fail, each publishing one unintended compatibility rumble command. Genuine compact control passes.
  • After correction: all five cases pass.
  • Related DualSense/audio/VIIPER regression suite: 546 passed, 3 existing audio tests skipped, 0 failed.
  • Full DS4Windows suite: 3,610 passed, 3 existing audio tests skipped, 0 failed (63 seconds; isolated C#/Go Xbox interop peer enabled). This green run does not attribute or resolve the previously recorded intermittent allocation-test failure from the earlier Switch 2 work.
  • Evidence under workspace _results: dualsense-native-media-fallback-red-20260905.trx, dualsense-native-media-fallback-green-20260905.trx, and dualsense-native-media-regression-focused-20260905.trx, and dualsense-native-media-regression-full-20260905.trx.

The running Desktop b65 portable session has not been replaced and does not contain this new source patch. Program Files, installed configuration and drivers were not modified. No release or installer was published.

Required game acceptance

  1. Confirm physical USB versus Bluetooth, and distinguish waveform haptics, ordinary rumble and adaptive triggers.
  2. On a separate Desktop portable candidate, compare the same Hades II action and profile. Verify nonzero virtual rear-channel media and physical media presentation, while checking for unintended compatibility-mode updates.
  3. Confirm perceived detailed effects and their stop behavior. If the game remains silent, inspect endpoint selection/PCM ingress and native transport health; passing unit tests is not grounds to close the reported regression.