Skip to content

ElixirOffice 0.3.0

Pre-release
Pre-release

Choose a tag to compare

View all tags
@leag1234 leag1234 released this 21 Oct 09:11
· 4 commits to main since this release
v0.3.0
764e834

ElixirOffice desktop 0.3.0

ElixirOffice 0.3.0 adds document tabs and moving documents between windows, with clearer desktop dialogs and improved computer-font startup. It includes the saving and installer fixes from 0.2.4 and earlier releases.

Added

  • Document tabs. Keep several documents in one window and switch without losing edits, destinations, password state or undo history. Up to 16 document and Home tabs are supported across the app's windows.
  • Move documents between windows. Drag a tab out to detach it, drag it onto another window's tab strip, or use File > Move document, the tab-strip button or the focused tab's Shift+F10 menu. The editor stays alive while moving.
  • Arch Linux community package. Install with yay -S elixiroffice-bin or paru -S elixiroffice-bin. This community AUR package repackages the released DEB and is not maintained by the ElixirOffice project. Report packaging problems on its AUR page and application problems in the project's issue tracker.

Improved and fixed

  • Linux DEB runtime dependencies. Explicit ALSA, GBM and GL dispatch dependencies avoid relying on libraries installed by unrelated desktop or test packages.

  • Desktop dialogs. Options uses grouped controls; Rename accepts Enter; recovery choices explain whether a copy is available and preserve durable recovery copies. Save/Don't Save/Cancel remains scoped to the document that initiated it.

  • Computer fonts. Background font discovery waits for the document engine and retries within a bounded startup budget. A warm index no longer causes a false read-failure warning; genuine permission, index and read failures remain visible.

  • macOS artwork. Refined application and disk-image artwork includes Apple-generated small icon representations.

  • ODS charts and password saves. Preserve embedded chart caches and worksheet identities, including local-table name collisions. Protected ODF exports use corrected CBC padding. The matched Wasm pair and three public LibreOffice 24.8.2.1 load/XLSX-export controls preserve ordered chart values and worksheet identities.

Qualification and known limitations

  • Final DEB dependency qualification requires a normal apt/dpkg installation, fixed-point ldd, dlopen/LD_DEBUG checks and Arch/Electron dependency cross-checks against the final build identity. The minimal container cannot start Electron because namespace setup returns EPERM, so no clean native start is claimed; host runs use a different sandbox helper and are diagnostic only.

  • AppImageUpdate uses updater-tool bytes matching the exact recovered pins. A moving download URL is a plausible cause of the earlier mismatch, but that cause is not proven.

  • Native tab detach/move behavior is qualified on Linux X11 at display scale 1 without a window manager. The owner manually tested tabs on macOS with the ad-hoc 0.3.0-rc.191 test build on 2026-10-20: several documents and detach to a separate window, with no crash or unusual behavior observed. This manual owner test is not a physical qualification matrix. On 2026-10-21 the owner tested the build 16 Windows candidate: tabs opened, closed and detached to another window successfully. This is a manual functional check, not a complete physical qualification matrix; other Windows configurations, compositors and display scales need further testing. Automated tests do not establish physical OS behavior.

  • Pressing Escape during a drag outside a window can still detach the tab. The document stays open; use Move document to bring it back.

  • The modern LibreOffice 26.8 encrypted ODS envelope is unsupported. Support for older tested protected ODS inputs does not cover this envelope.

  • Strict ODF 1.3 schema checks fail on all 27 tested XML parts in both the accepted 0.2.2 runtime and the new complete exports; both declare ODF 1.2. No regression is shown by this comparison; normative ODF 1.2 validity remains unmeasured. Strict ODF 1.3 conformance is not claimed. Source chart-fixture schema and converter roundtrip checks have separate, narrower scope.

  • Native LibreOffice qualification used public synthetic fixtures, a temporary app copy, a fresh profile and loopback UNO without a sandbox; no network-isolation result is claimed. Longer startup waits changed timing only.

  • Converter archive checks retain three ignored path-member cases with unchanged sentinels and two accepted DOCX/DOCM late-CRC cases. The full runtime qualification records these raw results.

  • Windows installers are unsigned. The macOS package is Apple Silicon only, ad-hoc signed and not notarized. Final installer/native acceptance must be recorded against the final build; earlier runtime or source checks do not certify final package bytes.

  • Original PDF text editing, OCR, PDF-to-DOCX, offline digital signatures and real-time co-editing remain unavailable. Physical printing, OS associations and every Windows save environment are not certified by automated checks.

Final build verification

Build 16 from Product 650967856ce88022a785cdc422ebe3aa97d4a631 and DocumentServer 2831cb0984207071f025bfe300a22f3eaef3e69e. Linux/Windows packaging and static resource audits are recorded in BUILD-INFO.json. Final Linux/Windows independent payload/resource/runtime identity checks, 606/606 strict desktop tests on each build platform, clean DEB dependency/install checks and default-worker/public-ODS/PDF browser gates passed. The owner accepted the targeted Windows tab open/close/detach test on 2026-10-21. The final macOS arm64 DMG passed static identity, source, resource, image-integrity and ad-hoc signature audits. The final build 16 Mac-host desktop suite recorded 603 passes out of 606 tests, with exactly three known platform-fixture failures and no skipped tests; these failures remain deferred to 0.3.1. The suite is not green. Final native GUI and printing acceptance remain unverified.

Linux DEB installation qualification uses the fixed-point ELF dependency audit, dynamic-loader observations, Arch/Electron dependency cross-checks and a fresh dpkg/apt install test. A pristine-container native start is unavailable because Chromium namespace creation is denied by the container environment. Host fallback runs with a different configured sandbox helper are diagnostic only; they do not certify a clean native package start. Physical/native acceptance is recorded separately.

Follow-up observations and tester feedback

  • The Windows owner reports a 30–60 second interface freeze and high CPU use while creating or opening a document on a Core i3 computer with 8 GiB RAM. The owner considers this non-blocking for release. The timing and cause have not been independently measured.
  • Windows testers: report whether this occurs, the ElixirOffice version/build and Windows version, CPU/RAM, document type, first versus subsequent open/create timing, and Task Manager CPU/memory/disk observations. Prefer a blank or synthetic document and omit private document content and filenames from reports. Also report whether Computer fonts is enabled and whether only ElixirOffice or the whole desktop stalls. The existing %APPDATA%/ElixirOffice/desktop-safety.log, when present and non-empty, contains timestamped reason codes and may be attached after quitting; it does not record opening durations. Save diagnostics apply only if saving also fails.
  • A Windows owner screenshot shows garbled characters around the document name in the save-changes confirmation. A UI follow-up is deferred; occurrence on other platforms and the underlying cause are unconfirmed.
  • The floating equation toolbar has blank category icons in affected DOCX documents; its palettes still contain mathematical choices. This issue was also reproduced in 0.2.4 and is deferred to 0.3.1.

Shutdown and diagnostic scope

Build 16's focused Linux X11 test used a private app copy with a different, already configured sandbox helper. All three functional scenarios passed, with four native view-order samples and no page errors. The overall harness nevertheless exited 1 after its 300-second close timeout after both synthetic documents were modified; the exact wait cause was not instrumented. All owned processes were cleaned up. This is a disclosed alternate-helper diagnostic, not a pristine package or full-run PASS. The earlier forced-cleanup SIGSEGV remains unresolved; this timeout does not establish or dismiss its cause. Physical printing and broader native platform qualification remain unverified.

Assets 11
Loading