Files
ARCADIA-Studio/harness/STATE.md
T

78 lines
26 KiB
Markdown
Raw Normal View History

# Progress Ledger — first release
**Checkpoint publication — 2026-09-10:** the user requested committing all current changes and pushing `codex/release1-continuation` to both `origin` and `gitea`. This is a work-in-progress snapshot. The interrupted palette-focused compilation failed with E0659 (ambiguous `column!` in `app/architecture_authoring/palette.rs:168`) and E0507 (moving the view out of an Arc in `diagram.rs:624`); an unnecessary-pattern-parentheses warning also remains in `architecture_authoring.rs:346`. The prior 584-test pass does not cover this newer palette revision. No Cargo/rustc process or local subagent was active at the publication preflight. Fix these errors and complete automated/native acceptance before packaging the redesign. Gitea coordination starts at issue #1 and the project Wiki; older evidence below retains its original scope.
**Latest priority checkpoint — Capella consultation and redesign:** the user rejected the large creation buttons and explicitly requested consulting the installed Capella. Native control resumed for that consultation in a disposable Capella 7.1 workspace. Actual palette, F2 rename and activity allocation were exercised; saved XML confirms allocation distinct from ownership. The horizontal creation/connection rows are being replaced with a compact expandable palette and inline naming. This supersedes the Escape hold and toolbar-fix proposal below. New implementation, native acceptance and a fresh user build remain pending. [Observed benchmark, alternatives and acceptance criteria](release-capella-basic-editing-20260907.md). The first-release goal remains active.
**Latest checkpoint:** basic-editing v1 passed full desktop584/584, Windows binaries, format and strict all-target Clippy. Native v1 D8CFAA confirmed direct contextual named Entity creation at100%, and found a22px placement shift caused by the pending toolbar changing canvas height. The user pressed Escape at the attempted Undo; that action did not execute. **Desktop control is paused again.** Process9600 remains the isolated unsaved QA application; do not overwrite its binary or presume a clean close. Code-only placement fix is continuing. [Exact v1 evidence and remaining acceptance](release-basic-editing-reset-20260907.md). This supersedes the resumed-control state below; the first-release goal remains incomplete.
**Superseding priority checkpoint:** the user explicitly said “Resume desktop checks” and rejected the focus on advanced review before basic diagram editing. Isolated native checks resumed and reproduced the global creation catalog, placeholder creation/camera jump and noisy review. Contextual named creation and compatible layer-transition destinations are now being implemented and tested. Their new executable has not yet been accepted. [Current priority and scoped evidence](release-basic-editing-reset-20260907.md). The older Escape hold and next-step priorities below are superseded; historical binary/test results retain their original scope.
Current checkpoint: 2026-09-07. The requested Windows development preview is built and packaged with its runtime, documentation, notices and complete examples. Latest Fit6, two existing scroll/branch regressions, full desktop570/570, workspace bins, formatting and strict workspace/all-target Clippy pass. Corrected server56/56 retains its earlier unchanged-server scope. Packaged DC50 GUI launched and the OA graph Fit overview was observed at1444×932 and1102×932; minimum-height and pending-form native acceptance remain open. The user stopped Computer Use with Escape; do not resume native control without renewed authorization. Native Linux core436passed/2ignored remains verified; corrected Linux server and full Linux GUI build/runtime remain open. The fresh Linux desktop cross-build passed on Rust1.90 in2m20s using the preserved cache; ELF and all37 teaching files are verified. Three later guest boots failed the early IO-APIC timer check before SSH/tests; the exact-kernel workaround was then rejected by automatic approval review before launch. All task VM processes are stopped. **Goal active; release incomplete.** The authoritative objective is `C:\Users\User\.codex\attachments\10b0f701-1743-4de7-9ee1-225920e19649\goal-objective.md`: a usable, feature-complete OA/SA engineering tool for Windows and Linux with graphical/source/REST parity, artifact traceability, offline guidance and independent end-state acceptance.
## Workspace and ownership
- Current branch `codex/release1-continuation`, checkpoint **4b6b5fc1977523e8cf52fdd5503a0a71c87a3f63** commits290files of source6 work and evidence. [Bounded checkpoint audit](release-checkpoint-integration-review-20260906.md) verified65pinned manifest entries and145exact staged/working fixture/archive files; `.gitattributes` protects their measured line endings. Integrated ancestry starts at `c091f80`; [branch audit](release-branch-integration-20260906.md) prevents reimplementing already integrated work. New source7 implementation is authorized but not yet accepted.
- Historical6BAB native scope (superseded in rendering by BCC1 and Fit by DC50): acceptance binary `.local/native-qa/bin-source7-acceptance/arcadia-studio.exe` SHA256 `6BAB6CDE7D31326F3C3CF7342D9B25FB6ADA032864BBE8104625E5EC9736283B` follows the Windows433/55/561 and strict/build gates below. Reset and ReadStatus each passed exact complete model/source Save/Undo/Redo and reopening with the correct derived Reply allocation30/31. OA caller→First.B passed exact persistence/history/reopen; routing remains open. Process11060 closed cleanly; isolated old6 recovery process11240 also closed after its complete source/JSON/raw-draft pass. Old5 process5244 subsequently passed exact30-source/full-JSON recovery and raw-workspace comparison. Current7 process8224→10420 passed invalid nested-link draft crash recovery; completing the recovered SA X→[R3,Inner]/B link passed exact model/two-source Save/Undo/Redo and saved reopening. Those historical native processes closed; the later BCC1 routing and DC50 Fit evidence is recorded separately below. [Native evidence](release-source7-native-acceptance-20260907.md).
- Completed bounded90minute Linux gate: one-job Rust1.90,455-file immutable snapshot SHA256 `7104946c89b2be7a1a030fdff47cb4405bacf05357d68cf35d181e860f55c279`, launched07:02:43UTC, deadline08:32:43UTC. Server compilation passed17m10s; unit2/2 and HTTP50/51 passed, with one504-vs-412 test assertion reproduced in an isolated rerun. Its two-second read/PATCH checks passed, but later persisted-state assertions were not reached. This establishes a timing-dependent fixture/deadline mismatch, not a production locking defect. Native core completed436passed/0failed/2ignored across40targets, Rust1.90, compile23m03 and1992seconds total; all455 source pre/post checks passed. The two ignored cases are optional inheritance scaling and explicit strace fault injection (earlier separately scoped injection passes remain historical). All artifacts were copied before sync/poweroff; QEMU6280, launcher5000 and seed816 were absent08:13:56UTC. [Current Linux report](release-source7-linux-gate-20260907.md). Earlier Linux evidence retains its older snapshot scope. WSL remains pending restart; no host restart is authorized.
- Windows session5735 completed all six gates with exit0: focused prepared-HTTP and cancellation, full server56/56 in68.78s, format10.44s, strict workspace/all-target Clippy99.60s, bins102.22s. Status/logs: `.local/source7-post-routing-gates-status.json`. Distinct GUI BCC1FC6DB68123C9D3A43CE53AA8EF59E33C54080632C104C99ED13270D5333F process8568 passed the scoped corrected routing checks and closed cleanly. [HTTP evaluator correction](release-http-preparation-phase-regression-20260907.md), [routing correction](release-contextual-route-native-fix-20260907.md), [native details](release-source7-native-acceptance-20260907.md). These passes precede the now-tested [Fit/layout correction](release-process-fit-layout-audit-20260907.md).
- The requested Windows preview is `dist/ArcadiaStudio-preview-source7-20260907-x86_64-pc-windows-gnullvm/arcadia-studio.exe`; its565-file ZIP includes all37 teaching-folder files and has SHA256 `1fac10b30daf2f8e1dac29e9469485e4bea142e77ff5a13e4c9d1e41ff03abb2`. The adjacent provenance records actual source/binary hashes. This is a development profile, not an optimized final release. Latest gates are in `.local/source7-latest-preview-status.json`; full desktop570/570 completed in198.82s test time,200.63s command time, exit0 in `.local/source7-latest-full-desktop-status.json`. Initial Fit4/2 failures exposed real zero-size-slot remounting; corrected constructors pass unchanged expectations. [Fit review](release-process-fit-independent-review-20260907.md), [development summary](../DEVELOPMENT_SUMMARY.md).
- Computer Use is stopped. Read-only inspection initially found preview9148 and second GUI8044 live; later9148 exited without agent intervention, while8044 remained untouched. No GUI control/closure is authorized for background verification. The cross-compiler lease ended after the successful142.766s monotonic run; no cross processes remain. The Linux worker completed a bounded headless guest experiment and released the lease. Three boots (same configuration twice, then Normal boot priority) reached terminal IO-APIC timer panic before SSH/Cargo; all task VM trees stopped and exact evidence was preserved. Exact matching kernel/microcode extraction passed without changing base/overlay hashes; automatic approval review rejected preparation/launch of the transient no_timer_check experiment as “blocked by policy.” No fourth boot or corrected tests ran. No task VM is live; see [the preserved boot evidence](release-source7-linux-server-corrected-gate-20260907.md). [Cross-build report](release-source7-linux-gui-cross-build-20260907.md).
## Current source build and stopped runtime gate
The guide-only followup compiled successfully with locked Windows workspace bins in12.70s Cargo/12.82s command, exit0. Status `.local/source7-guide-current-build-status.json` records guide SHA256 `80A6797051FB28E767F2AB5DD2F335ABCD9F3E5F3A75C731CDE46D8B90D5C92E`; resulting GUI SHA256 `327FBDF734716A7F1AF2874183D46A531A8F429BA1A478E83A3E7F2EB1D6248B`. Formatting and independent actual-label review pass. This later debug artifact does not replace the user-facing packaged DC50 preview, its archive or earlier native evidence. No full570 rerun is claimed for the text-only followup.
Previous goal turn made progress by confirming full570 tests and handing off the package/summary. This continuation produced the successful Linux GUI build/ELF evidence, validated purpose oracles, corrected/built guide text, and concrete guest-boot evidence. Goal remains active/incomplete. The denied direct-boot action must not be split or retried to bypass automatic review; no new Linux job is running or scheduled.
## Latest Linux build and purpose-oracle evidence
Linux GUI snapshot458files `source7-20260907T102528493108Z` has archive SHA256 `9b10aa8457a5371367f7e24a0b2260b63c19fed69873d98a3972062a744503d7`. Run `20260907T1025486926611Z` passed locked/offline Rust1.90 zigbuild, both source checks and independent ELF/package readback. GUI72,953,296bytes SHA256 `330A3872D48EBF24AD21E6D4FA13CB6098BEBB3CB7FF82A2E082925CCF3DD346`; ELF64PIE, glibc requirements≤2.28, all37 lesson files exact. Actual Linux GUI runtime is still unverified. Subsequent guide-label text is excluded from both published Windows and this frozen Linux preview.
Extended native checker `source7-native-contract-incidence.exe` SHA256 `16994DC437F5CCC5E5F6FBB4212C03449F2389AE4615FA98B2930B3637468EE3` compiled with warnings denied in4.05s during a coordinated short lease. OA Exchange and SA CommonWork self-checks pass: exact contextual assembly, unchanged precedence, wrong direction/work rejection, caller-only exchange coverage and stable source update. Separate43-element/10-relation incidence fixture baseline passes; all original golden hashes and preserved OA-link/Reset accepted cases remain exact. Independent review found no material issue. [Cases and reproducible native steps](fixtures/source7-native/README.md). Native purpose authoring/history remains pending.
## Completed and verified at this checkpoint
| Scope | Evidence and limit |
|---|---|
| Historical source6 diagnostics and guide | Semantic filter **108 desktop + 2 core** passed; full desktop **527/527** passed; strict all-target Clippy and GUI/server build passed (`.local/semantic-diagnostics-final-tests.log`, `desktop-diagnostics-final-tests.log`, `diagnostics-final-clippy.log`, `diagnostics-final-build.log`). Subsequent guide **3/3** focused tests, strict Clippy and build passed (`.local/guide-navigation-{tests,clippy,build}.log`). The guide test increases compiled desktop inventory to528; no full528 run or final full-workspace total is claimed. |
| Current artifact implementation checkpoint | Full desktop **518/518**, corrected fresh artifact picker **6/6**, strict all-target Clippy passed. Logs `.local/source6-artifact-inspector-desktop-tests2.log`, `.local/source6-artifacts-final-tests2.log`, `.local/source6-artifacts-final-clippy3.log`. Four subsequent artifact recovery falsifiers and two core absence oracles passed separately. This predates latest display corrections. |
| Native artifact MOVE / CREATE / DELETE | All three complete source Apply/Save, one Undo/Save and one Redo/Save oracles passed; **nine saves succeeded first attempt**. MOVE invalid Owner and disabled optional text, and CREATE excluded row/all eight decisions/reserved UUID survived controlled-stop recovery exactly. DELETE removed destination from identified moved-source absence while retaining the source link. [Exact native report](release-artifacts-native-20260906.md), [fixture/generator instructions](fixtures/release1-native/artifacts/README.md). |
| Source-6 atomic batch review | Included valid and excluded invalid raw rows survived forced-stop recovery; repaired two-source Apply and complete Save/Undo/Redo source comparisons passed. [Native report](release-current-native-verification.md). |
| Scenario ordered-member variant | At1100×700, Take rejected ambiguous insertion; explicit final order retained the local self-test. Complete Apply/Save, Undo/Save and Redo/Save matched the independent source oracles with only the schema5→6 header token upgrade. All three saves succeeded first attempt. Branch steps/fragment branches and this variant's crash recovery remain separate cases. [Exact native report](release-current-native-verification.md). |
| Native guide navigation | On current guide binary at1100×700: related Scenarios-bottom→Exchanges opens the new heading; filtering `exchange` retains the current topic's bottom position; `Git` auto-selects Build first and opens its heading. No dirty model. Root's actual native observations close these bounded scroll cases; [implementation/tests](release-semantic-conflict-diagnostics-20260906.md) remain separately scoped. |
| Communication means | Drawing/create/reuse and legacy reverse full-model/source/history passes are retained. Corrected reverse Apply now opens the explicitly requested OA scenario in Behavior, preserving the original four-object Architecture. Distinct oracle predicts10UUIDs/zero placements; accepted model/all3sources, Undo baseline, Redo and reopen pass. Every Save succeeded first submission. Navigation5focused tests, strict Clippy and bins build pass; compiled desktop535 is not a full535 run. [Exact report](release-communication-native-20260906.md), [implementation](release-behavior-transition-context-20260906.md). |
| Teaching project and Home hook | Corrected94element/62relation/5view/68trace/30source example passes4expanded direct-rustc tests. Minimum-size Home/guide, raw-source guard, reviewed NameTake and exact source/history/reopen pass. Navigation fix is reverified. Additional data/interface/type drill-down, saved-view comparisons, invalid-view guard and typed Entry action selection pass. Fold/unfold and function rename/Undo/Redo each match complete model/all30source oracles, five first-submission saves; saved reopen retains the Entry target identity. Complete process/scenario/reuse/evidence lesson remains open. [Native report](release-teaching-native-20260906.md). |
| Child/Incident and source-5 recovery | Two-row attachment recovery and complete Apply/Save/Undo/Redo passed, every save first attempt. Exact old source-5 raw/base/include/path retention, manual repair and two-file Save passed. [Attachments](release-attachments-native-20260906.md), [source-5 native](release-source5-native-recovery-20260906.md), [independent review](release-source5-checkpoint-independent-review-20260906.md). |
| Dense saved views | Forty-label baseline and selected real relation identities inspected; filter, duplicate/hide, original10/copy9 persistence across restart and exact model oracle passed. Additional coverage/context intersections, empty selections and Names-and-items preview/Apply/Undo observed. Minimum-size overview requires working zoom/pan. Limits remain in [native evidence](release-current-native-verification.md). |
| Core/source/REST integration | Fresh `--all-targets` core run: **433 passed/1 ignored**,44 target summaries, exit0 in138.72s; fresh server run: **55/55**,4 summaries, exit0 in64.74s. Logs `.local/source7-core-all-targets.log`, `.local/source7-server-all-targets.log`; exits/timings in `.local/source7-current-gates-status.json`. These supersede the earlier combined core count and pre-budget-fix HTTP51 checkpoint. OpenAPI **72fixtures/204schemas/90leaves** and both strict reST renders pass. [Core checkpoint](release-source7-core-checkpoint-20260906.md), [API integration](release-source7-api-integration-20260906.md). |
| Source7 desktop authoring and recovery | Latest routing3/3 and full desktop **564/564 pass**; [routing checkpoint](release-contextual-route-native-fix-20260907.md) records exact logs. Earlier full desktop **561/561 passed**, zero failed/ignored,263.39s in `.local/source7-desktop-full-tests2.log`. The initial560/1 run exposed canvas scroll-state loss when a new conditional navigation control shifted its widget; a fixed navigation container corrected it, and the unchanged regression passed. Coverage includes qualified5, operation3 (selected Async request fields), semantic source7 facets/omission3, recovery8, legacy source5 13, artifact search2 and focus guard1. Exact old-version enum/cache-path intent rejection and raw text retention are covered. The complete-scope endpoint picker preserves access beyond512; canvas preview remains bounded. Guide now accurately describes attachment/artifact reconciliation and artifact filtering. [Desktop checkpoint](release-source7-desktop-authoring-20260906.md), [recovery](release-source7-recovery-20260906.md). |
| Windows strict/build checkpoint and native7 | Strict workspace/all-target Clippy passed2m50s; root observed final format exit0/empty log. Three affected semantic tests passed3.46s after equivalent lint edits; native build exit0 recorded in `.local/source7-current-gates-status.json`. These earlier results scope the6BAB binary, not the later routing test binary or pending new strict/build sequence. Reset/ReadStatus full model/all-source Save/Undo/Redo/reopen pass. OA First.B persistence passes but routing acceptance remains open. Old6 Recover preserved the complete30-file bundle/workspace/Ports DTO, exact spaces/Unicode/fingerprint/reserved identity; Cancel/Save as and JSON export passed full independent source/model equality. Original seed unchanged. Old5 complete accepted/raw recovery, exact repair/30-source Save and full-model export pass. Current7 invalid nested draft crash recovery plus completed link Save/Undo/Redo/reopen pass; rendering still uses the older binary. [Native report](release-source7-native-acceptance-20260907.md). |
| Current Linux source7 and cross-built probe | Frozen native server compilation passed17m10s; unit2/2 and HTTP50/51, with one timing-dependent504-vs-412 assertion reproduced twice. No complete server pass; later targets not claimed. Native core436passed/0failed/2ignored; VM cleanly off08:13:56UTC. Separately, the complete Windows macro/build-script cross-build probe's hash-verified ELF executed successfully on Linux; this is neither a full desktop build nor GUI acceptance. [Native Linux gate](release-source7-linux-gate-20260907.md), [cross-build probe](release-linux-cross-build-probe-20260906.md). |
| Historical Linux/MSRV core and server | Ubuntu24.04.4/x86_64/glibc2.39/ext4/Rust1.90: core38/38, patched journal4focused/42full and3explicit syscall fault cases pass on recorded snapshots. Actual server build passes34m42s including3m24s pause. Real HTTP modular import, invalid PUT unchanged source/ETag, clean SIGTERM and exact model/source/ETag restart pass on the older28-source lesson snapshot. Full server test command timed out while compiling; zero tests executed. That gate shut down cleanly21:53:56; the separate current gate has the scoped results above. Current desktop/Linux package unverified. [Linux report](release-linux-validation-options.md), [journal contract](release-linux-journal-publication.md). |
| Packaging foundation | Portable Windows QA staging, dependency/runtime/font/Rust notices, archive readback and clean-PATH smoke passed at earlier binaries. [Package report](release-windows-package-review.md). QA archives are not current final release artifacts. |
Historical full platform/native milestones retain their original revision scope; [complete prior ledger](history/2026-09-06-ledgers-before-consolidation/STATE.md) preserves all evidence links and previous environments. A test or native pass never transfers automatically to a changed executable.
## Active findings and checks
Native artifact review exposed same-row conflict wording and reserved identities shown as deleted. Corrections retain typed actual-slot equality, row-addressed reservations and accepted-identity precedence; independent re-review found no further material defect. The completed108-semantic/527-desktop checks supersede the earlier interrupted run and include reserved-label/partial-creation regressions. Revised artifact wording still needs its own native observation on the current binary. Guide-scroll implementation/tests/build and the bounded native cases above passed. [Diagnostic/guide report](release-semantic-conflict-diagnostics-20260906.md).
**F1** teaching onboarding/review/history has current-binary native evidence; remaining tutorial work is scoped in its report. **F2**, faithful behavior-canvas SVG export, is deferred by root's scope decision: export remains architecture-only and must be described that way; no faithful scenario/process/state SVG acceptance is claimed.
Reverse scenario's first-dependency focus/overlapping architecture placements and teaching NameTake's separate layer/view mismatch are fixed and natively reverified. Latest NameTake4focused tests, strict alltargetClippy28.26s and binsbuild50.50s pass; no full539desktop pass is claimed. [Navigation evidence](release-semantic-review-navigation-20260906.md). Minimum-size process working space is cramped; native lesson evidence records23% overview versus100% scrolling rather than claiming readability at Fit.
Independent scope reconciliation verified **S1** missing contextual assembly into repeated/nested process references and **S2** missing maintained interface-operation identity in parameterless messages. Isolated existing-core probes falsify the proposed workarounds. Root ratified the jointly critiqued [source7 integration decision](release-source7-integration-decision-20260906.md), including deterministic cache-independent admission, whole-link deletion, purpose-specific analysis, exact Call-derived operation/Reply identity and one frozen old5/6 checkpoint migration. These contracts now have implemented core/source/REST/desktop/recovery paths and the current automated gates above; remaining native7 and current Linux acceptance remain incomplete. [Scope reconciliation](release-scope-reconciliation-20260906.md).
Linux ordinary rename's empty-target replacement was independently reproduced and corrected using `renameat2(RENAME_NOREPLACE)` without fallback. Exact errno/no-write assertions now pass under actual syscall injection. This establishes the ext4 publication invariant, not power-loss durability, hostile ancestor protection or model-file compare-and-swap. Windows retains its no-replace API and bounded scanner retries.
## Remaining gates and next actions
1. Fit implementation/review and automated gates are complete, including full570 desktop tests. Finish its native minimum-height/Focus/pending-form checks when native control resumes. Preserve completed BCC1 routing and old5/old6/current7 recovery/source/history oracles. Complete remaining current-binary S1/S2 purpose workflows and teaching cases; operation allocation deletion remains subject to independent channel validation. Native-check revised reserved labels and same-row conflicts. The later guide-only control-label correction passed formatting, independent review and the fresh workspace-binary build; it awaits the next packaged/native release snapshot.
2. Close uncovered selective-view lifecycle cases; advanced ordered/process/scenario/fragment variants beyond the accepted Scenario case; typed state actions and pending Linux data/hosted/compact workflows. Implemented semantics must not be called missing merely because acceptance is outstanding; [coverage audit](release-current-coverage-20260906.md) is the baseline map, newer scoped reports supersede its old gap claims.
3. Verify complete current Linux integration suites and desktop/server build/runtime, then supported-size native workflows. The successful core/MSRV pilot is a reusable starting point, not application acceptance.
4. Preserve current strict Clippy/format/reST/example/source/OpenAPI results and rerun affected checks after changes. Complete optimized/MSRV/final platform gates, actual hosted transport variants and final migration/package checks. Rebuild portable distributions from the accepted exact tree, validate notices/runtime/content/checksums and clean-environment launch before replacing old bundles.
5. Conduct fresh independent requirements/domain/parity/architecture/runtime/UX/end-state reviews; resolve material findings. Commit/integrate authorized source with reviewable scope; remote publication is not requested. Only complete the goal when its applicable criteria are met.
Meaningful progress continues: native lifecycle and Linux no-replace uncertainty have been resolved with exact oracles. The highest-value work is current-binary validation and closing uncovered engineering workflows, not accumulating repeated historical headings or treating aggregate test counts as release completion.