Verified October 13, 2026 against merged main
102a341fdcffd73fbc1622e39894befbe1badf18.
The foundation (#211), API (#217) and server (#218) are completed. The remote
Eclipse, web and synchronization implementations from #248 and #251–#264 were
integrated by #266.
#267 reorganized that
implementation and documented local Docker/Tailscale access.
| Acceptance requirement | Implementation and verified behavior |
|---|---|
| Versioned Quarkus API for owned projects, tasks, activities and labels | API v1, the canonical OpenAPI contract, repository-backed application services and the actual packaged HTTP acceptance suite cover CRUD, relationships, validation, optimistic versions, atomic operations and precise timestamps. Server/core builds have no Eclipse or Mylyn dependency. |
| Explicit authentication and owner isolation | Hashed bearer-token registry, read/write permissions, revocation and owner-scoped resources/change feeds. HTTP tests reject cross-owner access; installed account/server switches and phase.owner.binding prevent an old draft from writing into a reassigned account. |
| Remote Eclipse selection and optional Mylyn tracking | phase.native and phase.restart cover native records and encrypted token reuse; phase.mylyn* covers real activation/deactivation and retained Task List identity. phase.backend* covers server, embedded and remote return transitions. Remote phases retain closed, unchanged local storage; selection does not transfer data. |
| Standalone web workflows, interoperable with Eclipse | Eight Chromium cases use the real server. phase.web.eclipse.{records,drafts,tracking,reports,cleanup} proves actual browser/installed-Eclipse interoperability, shared UUIDs, exact totals, owner time zone and stale-draft rejection. The frontend uses API v1 and never connects to the database. |
| Defined concurrent refresh, conflicts and reconnect | Owner-scoped committed long polling; phase.changes.{owner-scope,activity-report,live-draft,reconnect} verifies automatic refresh and recovery. Real database restoration is covered by phase.restore.* and phase.label.restore.*, including reused numeric versions, preserved drafts and reviewed saves. |
| Visible failures and safe interrupted operations | phase.task.* covers authentication, offline operation, conflicts and committed timeout/cancellation. Create review, four tracking outcomes and six native deletion outcomes require explicit reconciliation. Four phase.import.* markers cover a committed lost reply, retained command key and receipt replay after an actual installed restart. No implicit local fallback or ordinary-write replay is used. |
| Native preferences, reports and interchange | phase.native.{label-preferences,report-controls,folder-dialogs} covers label drafts, exact reports, actual GTK file/folder pickers, cancelled transfers and fresh-ID import. The installed base feature separately verifies import replay across restart. |
| Deployable server and operational instructions | Server guide, Docker setup and Tailscale, deployment and Eclipse connection. Container acceptance covers PostgreSQL, rolling updates, backup/restore and durable restart. |
All eight jobs in Build run 38368609913 passed: API-contract, Server, Web, Remote-Eclipse, Publication and the Linux, macOS and Windows builds. Server containers run 38368609937 also passed verification.
Downloaded evidence was inspected rather than inferring acceptance from a green workflow alone:
server-test-results: 61 Java cases, no failures/errors/skips; the HTTP receipt
confirms 12 contract cases, normal/forced restart, revocation, long polling,
readiness and hierarchy/rollback/precision checks.web-acceptance: eight Chromium cases, no failures/errors/skips, with
passed=true and source commit matching merged main.remote-eclipse-acceptance: all nine installed test reports contain one
executed case each, with no failures/errors/skips. The base receipt contains
16 passing phase markers; the optional Mylyn receipt contains 48. This includes
actual GTK pickers and browser/Eclipse interaction, not only adapter tests.The server, browser and both installed receipts use the same distribution SHA-256:
1f16fb8b7886f3365b30869613783678c577a200fab33acb37e8b9ba49251c3b
Its downloaded ZIP checksum and timekeeper-release.json were checked against
that digest and commit 102a341fdcffd73fbc1622e39894befbe1badf18.
Counts from overlapping Java reactors are not added as unique tests. Workflow
artifacts may expire; this report preserves the source/run identities and the
inspected conclusions. The existing CI continuously reruns these scenarios.
These receipts satisfy #187 and its client/synchronization children #220, #221 and #219. Earlier checkpoint lists remain historical evidence; installed preferences/pickers, backend switches, web interoperability, real H2 restore and interrupted-operation/restart checks now have the evidence above.
This is functional client/server acceptance, not a published 2.0 release. The optional Mylyn symbolic-color Error Log issue #243 was deferred beyond 2.0 by the maintainer on October 13 and remains open. Final tagged artifacts and public p2 installation remain separate release gates in the release process. H2 restoration with both live UIs and PostgreSQL Compose restoration are separate tested scenarios; this does not claim both live UIs were exercised during PostgreSQL restoration. Failure scenarios establish the behaviors listed above, not every possible failure/operation combination.
Encrypted-token restart tests use an isolated test master password. They do not certify the user’s macOS Keychain configuration. The client’s storage diagnostic now identifies the Equinox failure category and points to Eclipse secure-storage preferences; the connection guide describes provider/password/file recovery without deleting shared credentials or storing plaintext. The earlier generic screenshot alone cannot identify which provider failed.
The completion change also initializes configured server accounts with the eight embedded defaults, using one shared domain catalog. Existing nonempty label sets retain their records, colors and versions. Initialization uses the normal owner transaction and service ports, including change notifications. Accounts added at runtime receive the same defaults on first use; an empty existing account is initialized at the next server startup.
Four additional real-database tests cover startup before client requests, concurrent first use/owner isolation, customized and empty existing accounts, and rollback without partial defaults. Packaged process acceptance additionally checks initial names/colors, distinct owner UUIDs and unchanged data across restart. The earlier merged-main receipt above predates this follow-up; its source/checksum is retained as recorded rather than presented as this change’s validation.
The corrected follow-up head 08f956a3e2c540fe9c385cb4cf653e9c4ba008c9 passed
all eight Build jobs in run 38382631394
and container verification 38382631454.
Downloaded reports contain 65 server cases, eight browser cases and all nine
installed cases, without failures/errors/skips. All 64 installed phase markers
passed. The packaged server receipt confirms defaultLabelsVerified=true and
both restart modes. The shared server/browser/installed checksum is
a3b021dfc388b72c05cd828bc9c5addcc5f09a15fbacb9706ff829afd4f95adb.
The tested PR merge source is c6d91a96dfdb401fa0dad453c07fd87d809820f9, rather
than the head alone. The first follow-up exposed a fixture opening label
preferences before direct cleanup deletions had reached its ready snapshot;
the corrected fixture waits for canonical projects/tasks/labels before opening
that page.
Pages now follows a successful main Build workflow, including the complete
platform matrix and Installed client tests (the renamed Remote-Eclipse job).
Its separate /updates/testing/ channel uses only that verified main p2
artifact; stable and preview remain release-only. The website explains the
installation steps and channel choice. The publisher checks source/run identity,
GitHub’s artifact SHA-256, p2 metadata and every referenced JAR. PR/fork runs,
incomplete/skipped client tests, wrong source/checksum and expired artifacts
cannot populate testing.
The publication follow-up passes 26 script/documentation tests and actionlint.
A complete site was also assembled from the actual main artifact for run
38368609913, with feature 2.0.0.202610130820, source 102a341, and artifact
SHA-256 98cf4ee7dfd8c3080cc1ea8d825a7f068259a65ed9e8e2b9c093a8202ae3a0b1.
Its testing composite contains a real child repository with all declared files.
Publication of this channel occurs after the workflow changes merge and Pages
deploys; local assembly alone is not a live installation check.
The later workflow correction puts Build website and update site after the complete
matrix and acceptance jobs. Its site uses the current run’s downloaded p2
artifact, while Packaging and channel tests runs separately in parallel.
Pages has no immediate main-push deployment; it follows successful Build.
The builder prepares the entire documentation/p2 site and uploads it on main.
Pages verifies the prepared artifact’s checksum and source/run identity before
deploying those exact bytes. Tests compare all prepared/deployed files and
refuse PR/fork/incomplete runs, changed source, corrupt archives and missing JARs.