RC4.6.3 mouse concurrency validation¶

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

Scope and baseline¶

Started from fetched origin/main at 62f206ab6a7d4e83d50387ed52cc5e967a076e48 in a separate worktree. Earlier unshipped haptics, Game Bar and Device Options edits are not included. VIIPER, USB/IP, Xbox persona and updater dependencies are unchanged.

The initial 14 real MapCustom/Commit simulations produced 10 passes and 4 failures on the unchanged mapper. Ordinary R2/face-button left-click plus controller movement, W/Space/Ctrl and mouse movement worked in both press orders. An unrelated toggled Button action suppressed left-down or left-up. This is a confirmed conditional bug, not proof that all reporters used toggles.

Source inspection also found that macro mouse events bypassed mapped-button ownership: an overlapping macro could release another binding's held click. FakerInput used toggle-style events, so its extra down could itself release the button. Its extended-button path additionally ignored X1/X2 selection.

Implementation and regression coverage¶

Separate toggle state per physical binding/controller; preserve configured two-stage trigger toggles and retire removed/shifted bindings.
Publish edges of the union of mapped, touchpad, toggle and macro owners for each of the five mouse buttons. Preserve explicitly retained macro state without accumulating unreachable owners.
Flush short macro mouse edges through buffered output immediately, with no timer, sleep, synthetic re-click loop or movement suppression added.
Capture macro epochs before scheduling; disconnected/stopped slots reject both delayed and queued old macro mouse events. Stop retires all slots.
Serialize backend replacement with mouse publication and its final flush; replay held buttons once to the reset/new backend. Release the publication lock even when a backend flush throws.
Select the bundled FakerInput wrapper's actual XButton1/XButton2 masks.

Tests use the real mapper/macro entry points, memory output sinks, real managed FakerInput report state without connecting the driver, explicit concurrent barriers, and warmed allocation assertions. They do not inject Windows input, restart applications, or alter profiles/touchpad settings.

Mouse-only local result¶

dotnet test DS4WindowsTests/DS4WindowsTests.csproj -c Release -p:Platform=x64

6,642 passed, 0 failed; 12 opt-in hardware cases skipped. This includes ConcurrentMouseMappingTests, MacroMouseOwnershipTests and MousePublicationFenceTests. Existing compile warnings remain; they were not silenced or treated as new evidence.

The 276 mouse-focused tests also passed five consecutive runs. Initial CI run 35549434528 passed its full test job, package build, offline layout and MSI install/repair/uninstall tests. Its previous-release upgrade fixture rejected the correctly versioned 5.0.9.0 MSI because the fixture still hardcoded 5.0.8.0. The fixture now receives the expected product version explicitly from CI; identity and hosted-runner safety checks remain mandatory. A fresh successful final-source CI run is required before tagging/publication.

Publication additionally requires the final-source GitHub CI and tagged draft workflow, installer lifecycle/upgrade tests, and final downloaded asset checks. Their receipts are separate from this local test result.

The subsequent temporary-profile fix requested before publication was validated together with these mouse changes: 6,681 passed, 0 failed, 12 opt-in hardware cases skipped. See the combined profile-switch ledger.

Limitations¶

The reporter's profile, machine and game were not available for reproduction. No claim is made that these conditional fixes prove the cause of their report or override games that reject mixed controller/mouse input. Microsoft documents touchpad suppression after typing, but that is only a conditional troubleshooting lead; no Windows setting is changed automatically. See mouse-click troubleshooting.