qqq-print-agent/docs/RELEASE-READINESS.md

Release readiness — signed 1.0 closeout

Updated 2026-11-04. The requested product release and local installation are complete. The dated records below distinguish the stable release from earlier candidate evidence; they are not a claim that every printer or deployment has been qualified.

Current release

• Windows: signed 1.0.1 on Gitea and matching GitHub updater assets, from source 5bf00f3291923f7afb34c6cce61276e69649689b.
• Release gate: run 41532214447 passed the build/test/package, security, CodeQL, script, Azure OIDC signing and publication gates. Published checksums, signatures and provenance were verified. See release signing.
• Installed: the November 3 check recorded MSI exit 0, app/service/CLI product version 1.0.1+5bf00f3291923f7afb34c6cce61276e69649689b, file version 1.0.1.65535, and service Running. No new runtime check is implied by this documentation update.
• Server and contract: qbit-printing 1.0.0, protocol 1.0.0 and QQQ 4.1.0-RC.21 are released. Native printing API 2.3, payload v1 and WebSocket subprotocol qqq-printing.v1 retain their own version meanings.
• Physical scope: Letter and 4 × 6 output were accepted earlier; the operator confirmed 4 × 1 on RC3 with the same relevant print path. Physical 4 × 2 was deferred and is not a release blocker. Dated evidence remains in Acceptance.
• Documentation: current install, sign-in, settings, media, logs, branding and release instructions reflect the stable release. Ordinary source changes use Gitea; the authorized GitHub signing and updater bridge remains in place.

No product implementation, release publication or installation task remains in this closeout. Shared CI-platform migration is separate from the shipped product. Historical unchecked items below do not reopen cancelled work or authorize physical tests.

Historical scope decision — 2026-11-02

The post-print native UI layout exception (WinUI LayoutCycleException after a print, recovered by the existing mitigation) was removed from the 1.0 blocker list by the user on 2026-11-02 and is accepted as a known limitation. It is not fixed, and the history below is preserved unchanged. It is not carried as another mandatory qualification.

Historical cancelled work, not active

Repeat Auth/S3 qualification, the print-latency investigation, and the MSI comparison/install-free upgrade analysis were cancelled by the user. Their recorded evidence is historical; do not resume them.

---

Historical matrix — candidate 0.4.0-rc.1 (superseded 2026-11-02)

The sections below were the working matrix on 2026-11-01. They are kept as evidence only and are stale where they list work as pending; the current truth is above.

Observed gates

Historical (superseded heads, kept as evidence): Windows PR #16 head 35ed93c7… hosted run 41340470827 passed Core 154 and CLI/Windows 317 with 0 warnings; server PR #30 head 4eee28e9… passed 183 reported / 182 executed / 1 opt-in skip / Checkstyle 0. Protected main then advanced (server rendering at printing 2.2.0; client virtual-agent), so those heads are not the candidate.

Current candidate, local results only:

Tree	Gate	Result
Windows	First focused build	Rejected: PrintWorker.cs(96,29) MA0051 (method 67 lines, maximum 60), caused by the phase-timing instrumentation. Log preserved. An earlier sandbox build had stalled with orphaned build workers; that was environmental, not source.
Windows	After extracting SubmitJobAsync, RequireCurrentPrinter, RequireSubmitAllowed, SubmitNativeAsync and JobAttempt (no analyzer waiver)	Focused negotiation/runtime run 16 of 16 passed, no warnings, no analyzer errors. Core.Tests was not built.
Windows	Roslyn token compare against merge base c4c663a	Exactly the four intended semantic files differ (PrintWorker.cs, PrintWorkerTests.cs, QqqApiDiscovery.cs, AgentRealtimeMediaCapabilitiesTests.cs); no fifth file. The worker refactor was reviewed: same reservation, staged-file and submission disposal inside the same catch, same authority checks and state order.
Server	Full verify, sandbox	188 reported, 0 assertion failures, 11 loopback setup errors, 1 skip. Environmental; log preserved.
Server	Full verify, actual user, before fix	188 reported, 0 assertion failures, 1 error (PrintingRealtimeTest.mediaCapabilityIsReplacedOnEveryAuthenticatedHello, temp-directory cleanup). Cause: PrintingRealtimeRoutes.close() did not wait for in-flight channel drains, which recreated the stopped file database. Log preserved.
Server	After stopWorkers() fix and regression closeWaitsForChannelWorkersAlreadyRunning	PrintingRealtimeTest five consecutive runs, 4 tests each, no empty-database warnings. Full verify 189 reported / 188 executed / 1 intentional opt-in workload skip / 0 failures / 0 errors, Checkstyle 0. Source-header guard over 146 tracked files passed; staged whitespace checks passed.
Protocol	Check	76 definitions and 78 fixtures passed locally; compatibility docs published as a signed draft PR #15 (commit 1916324e6d2c0dac0fd48f2401b654303ab228f9).

Not yet observed: hosted CI on the final signed candidate heads (Windows Core.Tests, the new RunOnce_LogsOnePhaseTimingLinePerJobWithoutDocumentContent, Cli.Tests totals, MSI build at ProductVersion 0.4.0; server verify). Hosted passes do not clear the local Defender quarantine of the Core test assembly, which stays unbuilt and unrestored. No installer, installed-upgrade, physical-output or published-cohort native-callback result exists for this candidate.

Status words are strict. Proven names evidence that was observed for the stated scope. Partial means part of the requirement has evidence. Incomplete means no qualifying evidence exists yet, including where only source or a plan exists.

Requirement	Status	Evidence on record	Open gap / exact next evidence
AppKit parity (menu, About, settings, theme, logs, tray, update)	Partial	APPKIT-PARITY.md; real in-app 0.3.1 → 0.3.2 update accepted (ACCEPTANCE.md); 426-test Release gate at 0.3.2	Update download/installer failure, cancel and reboot-required paths; high contrast, Narrator, scaling and small-screen matrix. Not re-run for this candidate.
Realtime, reconnect, large-file control concurrency	Partial	Source tests: PauseAndCancelRemainResponsiveDuringLargeDownloadWithoutHttpPolling, NativeCookieAndDeviceProofUpgradeKeepsReceivingDuringStalledDocument, ordering/generation/bandwidth suites; installed 0.3.x reconnect observed (REALTIME.md)	TODO.md still lists the combined runtime review (stalled transfer, lost ACK, reconnect) as unchecked. Needs the hosted Core + Windows gate on the exact head, then an installed reconnect check on a compatible server.
Jobs table, details, sorting, 10k/day scale	Partial	JobSortingTests.TenThousandJobs_AllColumnsSortAcrossPagesAndInterleaveBoundedCloudQueue; AgentStoreScaleTests; 375,000-job store measurements in SCALABILITY.md; installed UI checks at 0.1.9/0.3.1	Measurements are local storage with fake printers. GUI rendering at scale, service scheduling and printer throughput are unmeasured. Not a throughput claim.
Printer media (this candidate)	Partial	Read-only real-driver/DC normalization probe 4×6 in / 300 DPI with no queue change (PRINTER-MEDIA.md); media is offered only to stable printing 2.3+ in major 2 (2.2 rendering stays a three-capability peer), covered by the 16 passing focused negotiation/runtime tests (HELLO per version, reconnect upgrade and downgrade)	The earlier hosted Core pass was on a superseded head. Final-head hosted Core is pending and does not clear the local Defender quarantine. No installed profile-selected print has occurred.
Print latency (measured 5–7 s ZPL test print)	Not pursued	Historical: source trace of the per-job path and a diagnostic plan (LATENCY-DIAGNOSTIC-PLAN.md in the coordination outbox); the candidate logs one phase-timing line per job	The user cancelled further latency work; the cause stays unproven and nothing is claimed. The log line remains available.
Branding and install	Partial	BRANDING.md; branding tests; MSI payload/resource/notice checks recorded for 0.3.2	Candidate MSI is built only by the hosted job on the final head (pending). Needs MSI resource comparison against the 0.3.2 baseline after CI, then the install-free plan in UPGRADE-QUALIFICATION.md.
Status and log rotation	Partial	AgentLogSettingsTests (10k jobs × 3 entries inside the default daily budget); daily-cap marker and clear-logs behavior in SCALABILITY.md	Installed rotation across a local-midnight boundary and cap visibility in the UI are not observed.
Native auth, permissions, recovery	Partial	Isolated real Authentik login, permission, logout, expiry/renewal and revocation against an unreleased API 1.9 candidate in local Docker; recovery regressions in QQQ-SIGN-IN.md	Installed browser/private-callback and long-lived grant need a compatible published API 1.9+ cohort. The local server 1.8 does not qualify. Incomplete until that cohort exists.
Retained-source and S3 retention boundaries	Partial	Server local full verify on the merged candidate: 189 reported / 188 executed / 1 intentional opt-in 10k skip / 0 failures, Checkstyle 0; callback authority and accepted-row history fixes independently reviewed; the server's docs/STORAGE-AND-RETENTION.md states the layout, outbox-copy and versioning boundaries	Final-head server hosted CI pending. Production IAM, bucket policy, noncurrent-version retention and cross-platform key migration are deployment work, not proven. Echo's experimental core tuple is unqualified; no silent repin.
Packaging, CI/CD, notices, About	Partial	Notice and Query About guards merged into both bases; release workflow pins actions and protocol; candidate version inputs (0.4.0 / 0.4.0-rc.1) and the reviewed workflow patch are applied by the coordinator, with action pins unchanged	The protocol pin in ci.yml is the final commit 42c0d249f81395f00408ec0f92638dcbbf5deb03 (Gitea PR 2); earlier pins are historical. Candidate identity is 0.4.0-rc.2 (ProductVersion 0.4.0). Hosted packaging steps on the final head are pending. MSI remains unsigned.
C# source standards (XML how/why headers, member layout)	Partial	License headers on all tracked C# (IDE0073, SOURCE-HEADER-COMPLIANCE.md); how/why headers and structural spacing completed in batches A to D, each token-compared as comment and blank-line only; Roslyn helper controls passed	Source cleanup is complete and merged. A final inventory over the merged tree has not been reported, and the merged UI build and 35 paper/media tests passed locally; hosted CI is pending.

Order of closure

1. Done locally: both trees pass their focused/full local gates (table above). Pending: signed functional commits and exact-head hosted CI for the Windows and server candidate heads.
2. Done: the Roslyn inventory/compare helper passed its positive and negative controls.
3. After the functional baseline is committed: documentation-only C# standards batches of at most 8 files, each with token-equivalence proof and never mixed with semantic fixes.
4. Version inputs (Directory.Build.props, build/Build-Installer.ps1, README, CHANGELOG) and the ci.yml/release.yml patch are applied for the candidate. No tag.
5. Normal Gitea Core/CLI/build CI on the integration head, then installation of the new label implementation and a small-stock paper test.
6. Historical, no longer active: the install-free upgrade plan in UPGRADE-QUALIFICATION.md and the latency diagnostic plan were cancelled by the user; their recorded evidence is kept.

Not claimed (historical, 2026-11-01)

Authenticode signing, physical-stock sensing, installed profile-selected printing, desktop/native OAuth callback or long-lived grant on a published cohort, remote cloud deployment, broad driver matrix, and any 1.0 readiness.