RC4.6 publication ledger¶

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

Published identity¶

• VIIPERRC4.6: Release Candidate 4.6 Hotfix — Clearer Audio, Reliable Feedback & Disconnects.
• Published 2026-09-14T15:26:47Z, release ID 388498525, as an unsigned named-RC prerelease under the existing draft-first policy. No signing-policy exception was added or broadened.
• Final tested source: 96b59e37bb7649d0f00004f5e58c57d988959dfb, pushed to main before tagging. Annotated tag object: 01c0de4ed34008b9ba0e6327a3ed86773f91ec85.
• Application, MSI and Burn version: 5.0.6.0; display/update-channel identity: VIIPERRC4.6. Both changelogs, installer defaults and upgrade fixture agree.
• Detailed changes since RC4.5.9 are in the tagged release notes.
• USB/IP remains 0.9.7.7. DS4Updater remains the published 2.0.6, source 173c27b4eb31f369280abce173e31264b317231e; no new updater release was needed.

Companion broker publication and source binding¶

VIIPER v0.1.5-rc4.6 was published first at 2026-09-14T14:57:52Z, release ID 388483169, from source c2fc304bc21f829ddb35b9e3e3d6b6ac0a9198a7. The Windows PE version is 0.1.5.0; product version is 0.1.5-rc4.6.

• Main CI 34856384005 and tagged build 34857881208 passed. Infrastructure-only download/artifact failures required retries; source, workflow and checksum protections were not weakened. The successful Windows AMD64 artifact remained the first-attempt bytes.
• All 12 published broker assets were downloaded and checked. The exact corresponding-source ZIP has 808 entries and its Git comment identifies the full broker commit. It was not reconstructed from an unrelated checkout.
• DS4Windows was repinned to the actual tagged CI executable, its notices, full build/provenance tuple and corresponding-source digest before DS4 CI. Both bundled executable aliases match the published CI executable.
• The dedicated synthetic process-test broker is not a shipped artifact.

Broker executable SHA-256: 9A334912E26FAC09C6DF17BA75A272A41D871D9BBE14934D06F983E101F3DF3C. Broker source ZIP SHA-256: 19C59D787B04F9597B47D40D7F6BA18F8837A8F590397A41F5E456F4A80A5787.

Tests and installer lifecycle¶

• The initial post-pin full run failed an exact-zero allocation assertion. Publication was held. A native allocation-profiler investigation reproduced counter growth without object allocations during GC allocation-context accounting; the same real writer's intentional report-clone positive control was detected. The existing bounded strict measurement scope was applied to this test, without raising the zero-byte limit, adding retries, changing production GC behavior, or attributing the original untraced sample with certainty. See the allocation evidence.
• Final rebuilt local x64 suite: 5,717 passed, zero failed, three live-audio gated skips, including eight opt-in isolated C#/Go lifecycle process tests. Focused allocation regression/positive-control coverage: 48 passed.
• Exact-source main CI 34859888276 passed 5,709 tests, zero failures, 11 opt-in skips. The eight process tests are opt-in there; this accounts for the difference from the local total.
• CI also passed setup diagnostics, startup ownership/registration tests, complete packaging, offline layout, MSI install/repair/uninstall, and upgrade from published RC4.5.5 to 5.0.6.0 with two profiles preserved. The upgrade fixture found zero installed registrations after uninstall.
• These installation lifecycle tests ran on the disposable hosted Windows runner. The user's installed applications, drivers, profiles and controller sessions were not changed for publication validation.

Tagged artifacts and package completeness¶

Tagged draft build 34861077780 passed on its first attempt at the exact tested source and uploaded all 13 workflow-owned assets. No local build was substituted, and no asset was overwritten. The release body exactly matched the tagged release notes.

Verification bound the Actions uploader, run/source identity, release ID, GitHub sizes/digests, checksum manifest, corresponding-source ZIP comments and RELEASE-BUILD.json to the downloaded bytes. Changelog source files matched their committed Git blobs with only pure CRLF-to-LF archive-export normalization where needed. Independent broker evidence came from its already verified CI publication, not from the DS4Windows candidate's own declarations.

The portable ZIP and extracted folder contain 553 files, 297 dependency assets, 23 language satellites, the self-contained .NET/WindowsDesktop 8.0.30 runtime, the required unchanged authorized Xbox persona, broker and notices. Baseline comparison verified 503 immutable byte hashes and 14 text assets; reviewed app/broker replacements were pinned independently.

Passive WiX extraction checked eight embedded Burn payloads, 551 MSI files, 549 shared file hashes equal to the portable ZIP, 549 ownership entries and 550 package-manifest hashes. It did not execute setup, MSI or extracted helpers on the user's PC. Hosted lifecycle tests used the CI package from the same source; the final release package received these byte-level composition checks rather than a second host installation.

Real updater transaction and public checks¶

The immutable DS4Updater 2.0.6 DLL has SHA-256 04B8E564F25CBEC14C65192D9E10408C01218B95B31EAFD1F0511B799E7873F1. Its actual filesystem transaction applied the final release ZIP to a fresh synthetic RC4.5.9 fixture and verified all 553 installed files. It preserved eight user-data sentinels (profiles, actions, linked profiles, auto profiles, Joy-Con pair data, key placeholder, plugin placeholder and log), removed an obsolete owned DLL and all five old VIIPER-0.1.4-rc4.5.6 extras files, and left no staged transaction. Neither application was launched.

The actual draft receipt/API passed four checks including draft rejection. After publication, 23 actual-receipt policy cases passed: earlier RCs advance to 4.6, equal/newer versions do not, the receipt resolves 5.0.6.0 without a hardcoded release map, drafts and altered receipts are rejected, and binary downgrade protection remains enforced. The existing bootstrap obtains the published updater; no stale updater executable was added to the ZIP.

All 13 published asset sizes/digests still match the downloaded, verified draft. The published-event audit 34862199728 passed at the exact tagged source. It downloaded and verified the public bytes against the successful draft-build receipt without rebuilding or replacing assets; the release-build job was correctly skipped. Its byte-verification step completed at 2026-09-14T15:27:34Z.

Exact downloadable bytes¶

Application DLL SHA-256: DC573CC6C9214B968A7382E6F39AA51705F34AD307548B28624A0F89DA5ACB20. Application EXE SHA-256: 9C2383FE0A911B053243AD25D9AC21DCBC7EAABFB778C0CA44F1D7398BE8A5D6. Remaining standalone notices/checksums are bound by the published receipt.

Local downloads and extraction evidence are under Desktop DS4Windows-RC4.6-Release; test/updater proofs are under isolated_results/rc46-publication/. Initial failed-test evidence is retained.

Acceptance limits¶

The fixes cover audio source recovery/rate conversion/retirement, bounded and ordered cross-controller feedback delivery, explicit manual Bluetooth removal, and a stale Game Bar profile-eligibility guard. They do not retune Switch 2 rumble gain/encoding/cadence or replace RC4.5.9's faster DualSense feedback path.

Final listening, physical disconnect behavior and gameplay feel still require tester confirmation. The rare once-in-a-day movement interruption remains unattributed; the Game Bar guard is not proof that incident is solved. Genuine disconnect/shutdown neutral-report safety remains intact. No unlimited physical output rate, zero end-to-end latency, HidHide repair or completed issue #81 gameplay acceptance is claimed. No issue was closed in this release publication step.