meshbay.git/packages/meshbay-client, branch 0.14

fix(win): a service-mode daemon can be replaced, and the Node page can link one

2026-09-14T21:51:36Z

Two live-reproduced bugs in Windows node start/stop, found sideloading the 0.14.0 build: - node:start's crash-recovery step killed a service-mode daemon with taskkill/CTRL_BREAK, both of which fail with "Access is denied" against a process running under the Scheduled Task's own S4U logon session (a different session from the Electron app's). The daemon it was meant to replace just kept running, unreplaced, and schtasks /run on a task Windows still considered Running was then a silent no-op too. Route through winServiceTaskEnd() (schtasks /end) first, the way nodeServiceStop/ nodeServiceRestart already correctly do. service-mode.ps1 also now starts the task right after registering it -- Register-ScheduledTask's own AtStartup trigger does not run it immediately, so nothing was listening until the next reboot. - The Node page's Start button called node.start() with no arguments, so an unlinked node (a fresh install, or one whose hub-side link was lost) could never link on Start alone -- only create-group-page.js's own call passed {hubUrl, username, token}. Reproduced on a fresh non-service install signed in to the real hub: Start hung for ~105s and failed with "could not link", pointing at a "Link Node" control that lives on Settings, not the Node page (that message is fixed too). Co-Authored-By: Claude Sonnet 5

chore: bump version to 0.14.0

2026-09-14T12:45:19Z

Co-Authored-By: Claude Opus 5 Claude-Session: https://claude.ai/code/session_01T6jPTeocXA1BePekdsgPya

fix(packaging): three MSIX first-run regressions found by a real sideload

2026-09-12T13:48:22Z

A second-machine sideload of the MSIX target surfaced three things the earlier verification round (which only proved the package installs and runs) had missed: 1. meshbay-node missing from PATH. installer.nsh's customInstall adds node-runtime\ to HKCU\Environment at install time -- an unelevated per-user write, never blocked by MSIX's no-elevation rule, only by the more basic fact that an AppX/MSIX install runs no custom code at all. packaging/win/ensure-node-path.ps1 (idempotent, no admin verb) plus main.js's winEnsureNodeOnPath() do it from the app itself instead, once per launch, shipped to Full and MSIX (not Light, nothing to add there). Verified live via the Node inspector protocol: the entry was in HKCU\Environment\Path after a launch, absent before. 2. A daemon that crashes on startup failed silently. spawnNodeDetached() used stdio: 'ignore', so a real crash reproduced live (a second instance colliding with the first on 127.0.0.1:18000) left waitForNode()'s generic 60s timeout as the only failure ever shown. spawnNodeDetachedWatched() pipes stdio and watches ~2.5s, rejecting immediately with the daemon's own stderr on an early exit; a survivor has its streams released and runs fully detached exactly as before. First version bounded the captured text by line count and a live test showed that cut the actual OSError line -- two uvicorn/asyncio tracebacks followed it in the real capture -- so it is bounded by characters instead. 3. No hint that a startup-mode choice exists. The install-time radio page was the only place this was ever offered, and nothing replaces it now that no install-time page can exist at all. SetupWelcome (the existing first-run banner) grew a conditional hint, shown only while a bundled node is present and neither autostart nor service mode is configured yet. Considered and rejected: linking straight to the Node page -- its route is gated on a linked hub node key, false on the exact fresh-install screen this hint targets, so the link would have been dead on arrival. New key setup.node_startup_hint, added to all ten locale catalogues. test_packaging_win.py gained six tests pinning all three (69 total). Full plan and verification detail: C:\Users\admin\devel\msix-installer.md section 13 (out of repo). Co-Authored-By: Claude Sonnet 5

feat(packaging): an MSIX target for Microsoft Store submission

2026-09-11T15:51:51Z

Store certification of the NSIS "MSI/EXE" submission failed on three checks (silent-install verification, Add/Remove Programs entry, bundleware check) -- traced and reproduced live to one cause: SmartScreen blocks an unsigned, internet-downloaded installer at the shell layer before Microsoft's own unattended validation bot ever gets to run it. MSIX sidesteps this class of failure entirely: submitted through the Store's native pipeline, there is no browser-download-then-launch step for SmartScreen to intercept, and Microsoft signs the package itself at publish time -- free, and specific to this submission type (Trusted Signing remains a paid service for the MSI/EXE path). Full plan and findings: C:\Users\admin\devel\msix-installer.md (out of repo). electron-builder.msix.yml carries the same bundle as Full (node runtime, ffmpeg, both service scripts) -- an AppX/MSIX install never elevates, by design, but that changes only *when* the two elevated operations can run, not whether the daemon ships. No main.js changes were needed: the on-demand elevation path for service-mode (winElevateServiceMode(), driven from the Node page) already existed for a different reason and depends only on service-mode.ps1 being present as an extraResource, true for any packaged Windows target. identityName/publisher/publisherDisplayName are the real values from Partner Center's app-identity reservation, not placeholders. build-win-msix.ps1 points electron-builder at the system Windows 10 SDK (auto-detected) instead of letting it download its own bundled copy -- that download's 7z extraction creates symlinks this target never uses and fails without SeCreateSymbolicLinkPrivilege, reproduced on this machine. build/appx/ carries the four tile images the AppX target requires regardless of showNameOnTiles, generated once from the existing app icon (see that directory's README) since the system-SDK redirect has no vendor samples to fall back to. build/appx-extensions.xml declares windows.startupTask by hand rather than via electron-builder's addAutoLaunchExtension, which always targets the Electron shell -- this points at the bundled node binary instead, matching what "starts at sign in" already means for Full. Verified live via a signed sideload install (self-signed test cert, cleaned up after): the package installs and the app runs correctly. One finding worth carrying forward -- the declared network capabilities (internetClientServer, privateNetworkClientServer) do not create any firewall exemption for this app, most likely because automatic capability-based exemption is an AppContainer-sandbox property and this app deliberately runs full-trust, outside any sandbox. Not a regression: no install-time elevation was possible either way, so the cost is the same one-time OS firewall prompt firewall.ps1's own header already documents as its fallback today. Co-Authored-By: Claude Sonnet 5

feat: gate the create-group wizard on whether a node is bundled

2026-09-11T15:51:51Z

MeshBay Light has no bundled meshbay-node.exe, so the create-group wizard (which assumes it can start a local node) needs its own signal, not just platform.node.available. main.js exposes it over IPC (node:bundled) by checking the packaged resources directory rather than trusting a build-time constant; preload.js and platform.js carry it through the usual contextBridge/wrapper path. winCanElevateServiceMode() replaces the two prior 'app.isPackaged' checks for whether the app can offer service-mode elevation -- Light is packaged but has no service-mode.ps1 to elevate into, so packaged alone was already the wrong test even before this target existed. create-group-page.js gates the wizard step that starts a node on the new capability instead of hiding the whole feature; node-page.js's comment fix is unrelated cosmetic drift caught in the same pass. Co-Authored-By: Claude Sonnet 5

feat(packaging): a Light installer target with no bundled node runtime

2026-09-11T15:51:51Z

MeshBay Light ships the Electron client + UI only -- no PyInstaller node freeze, no ffmpeg, no service install/autostart. Two standalone electron-builder configs (Full via package.json's build field, Light via electron-builder.light.yml passed with --config, which reads only that file -- confirmed against app-builder-lib's own config loader) rather than one config branching on a flag. build-win-common.ps1 holds the steps both orchestrators share (Node check, npm ci, Electron bump, sync-ui) so build-win.ps1 (Full) and the new build-win-light.ps1 cannot drift apart; build-win.ps1 is refactored to dot-source it with no behavior change (rebuilt and diffed byte-identical output). installer-light.nsh keeps the one thing Light still needs -- an unconditional firewall rule, since the client listens too -- and none of the service-mode/autostart machinery installer.nsh carries, which has nothing to gate without a bundled node. dist-light/ (Light's own electron-builder output dir) gets its own .gitignore line since the bare dist/ rule does not match it. Co-Authored-By: Claude Sonnet 5

feat(packaging): one radio page for Windows autostart, firewall every mode

2026-09-10T23:33:05Z

Operator feedback on the 0.13.0 installer: - The all-users / current-user page (electron-builder's PAGE_INSTALL_MODE) only ever showed "anyone who uses this computer" disabled -- MeshBay is per-user only (account-bound keystore/DPAPI, MESHBAY_DESIGN.md 11.2) and build.nsis forbids elevation. customInstallMode forces $isForceCurrentInstall so the page is skipped. - The two nested Yes/No MessageBoxes are one nsDialogs radio page (customPageAfterChangeDir): only-while-open / at-sign-in / background service, default background service. customInit seeds MB_AutoMode "2" for silent installs where the page never runs. "At sign-in" now writes the Startup .vbs from the installer (meshbay-node autostart install, unelevated); the old per-user branch set up nothing. - The firewall rules go in for every mode, not behind a second opt-in -- a node that accepts no connections is the failure mode MESHBAY_DESIGN.md 7.5 names. Folded into the service elevation for mode 2; their own single elevation for 0/1. Unelevated short-circuit kept but narrower: firewall.ps1 check AND service.ps1 status must both pass to skip mode 2's UAC. Var MB_AutoMode lives inside customPageAfterChangeDir, not at file scope: the uninstaller compile pass inserts none of the macros that read it and makensis -WX turns "unused Var" (6001) into a hard error. Not yet exercised on a real machine -- the NSIS UI cannot be driven from the build env. test_packaging_win.py pins the script shape. Co-Authored-By: Claude Sonnet 5

feat(hub): pin the app controls while a library scrolls

2026-09-10T08:29:12Z

Files, Videos, Music and Photos are read by scrolling, and everything that steers that reading left with the first screenful. Three bands now pin under the navigation bar, in a group and on the Search page alike: the tab bar (the search field, on Search), the application's own toolbar, and the file table's column heads. The group's name and description still scroll — they say nothing a reader needs while walking a directory, and the height they would cost is height the list does not get. A band's offset is the heights of the bands above it, and those are not constants: the toolbar wraps to three rows on a phone, grows a field while a folder is being named, and loses its filter on Search. So each band measures itself and publishes `--chrome-h` / `--toolbar-h` (static/sticky.js) and the stylesheet does the arithmetic in calc(), rather than a number written down twice — the fault CLAUDE.md already records against this layout twice over. A band publishes height *plus its own bottom margin*, and paints that margin as a ring of page colour, so the pinned layout is pixel-identical to the flow layout and nothing shifts at the moment a band pins. Three overflow faults came out of it, all of the same class and all of them what "the header does not stay" actually meant on Android — a document wider than the screen leaves everything pinned attached to a viewport the reader can no longer see, the navigation bar included: - a directory's name cell was a bare , so an unbreakable folder name (`Rage_Against_The_Machine_Discography_1992-2000_FLAC`) set the column's minimum: a 527px table in a 390px window - Search's group column did the same at 442px with an underscored group name. It also goes entirely below 768px, where there is no room for it and the breadcrumb already names the group - the shared-directories table has four columns of controls with a combined minimum near 440px, none of it compressible. On a phone the row stops being a row: the name and its eject/remove pair on one line, the two switches — each carrying the column head's own string as a label — on the next - and, found by measuring at 360px, the tab bar itself was 19px too wide `.file-table` moves to separated borders: a collapsed border belongs to the table rather than to the cell, so the column heads lost their rule the moment they pinned. Measured, not read. tests/harness/sticky_header_probe.py drives the shipped GroupPage and SearchPage against a stub node, walks to each application, scrolls to the end and reports every rectangle — 11 views x 4 widths x 2 engines. Its fixture says what real data says: the first version used `note-007.txt` and `un groupe`, which fit any screen, and found none of the above. A fixture narrower than real data tests the fixture. Also: `test_desktop_shell` no longer looks for the CSP after the first `-->`, which made it fail on correct markup as soon as a comment was added above it, and `search-page.js` joins test_hook_ordering's file list. Co-Authored-By: Claude Opus 5 Claude-Session: https://claude.ai/code/session_01Tx16FhyD2BUdpooGb5jcyN

fix: the PDF preview needs object-src and frame-src, in both policies

2026-09-09T13:48:53Z

A PDF preview showed the "this browser will not display the PDF inline" fallback everywhere — in the desktop client since its first launch, and in the browser since the hub started sending a CSP on 2026-09-01. It read as a missing native feature because before that commit the hub sent no policy at all, so Chrome had once worked and the application never had. Two directives govern one feature. `files-app.js` decrypts the file in the page and hands it to `` from a Blob; Chromium loads that as plugin data (`object-src`, absent and therefore falling back to `default-src 'none'`) and then renders it in an internal frame (`frame-src`). Opening either alone changes nothing visible — the second refusal produces the same fallback. `'self'` covers neither: a same-origin `blob:` URL is not matched by it in either directive, measured in Chrome 152 against the deployed page and in Electron 44 against the client's own policy. `plugins` stays at its default `false`: the built-in viewer is not behind that flag on Electron 44, verified by rendering one. Widening `object-src` from `'none'` to `blob:` admits only what page script minted itself, at a type this code sets — PDFium parsing bytes that came from a node, which is what any browser does with the same file once downloaded. Tests: each policy is pinned to carry `blob:` in both directives (each fails if either token is removed), and the two policies are now held identical directive by directive apart from the two deliberate differences — the comment claiming they were the same had already drifted and nothing checked it. The CSP source parser in test_desktop_shell.py read `//` comment lines as directives, which is the "parse directives, not text" mistake this file already records; it skips them now. Co-Authored-By: Claude Opus 5 Claude-Session: https://claude.ai/code/session_01XauykfBvRrpy6RYbF6F7Wu

feat: MNP 3.0 — a transfer needs a lease

2026-09-09T12:00:22Z

Stage 4 of ~/next/improve-downloads.md, the flag day. Leases become compulsory and a 2.x peer is refused at the handshake. **The bound on leaseless reads (§3.4.1) did not exist, and it is what makes the rest mean anything.** Browsing a group is never subject to a transfer slot — that is an operator decision and a requirement: a member must be able to browse a group at capacity exactly as they browse an idle one. But "not leased" cannot mean "unbounded", or a client that simply omits `tr` transfers outside every cap and the caps are decoration. A session may now read two distinct files at once without a lease: one because a viewer looks at one file, two so that prefetching the next photo stays possible. A count of files and not a byte budget, because a RAW photo is 60-80 MB and is browsing while a 40 MB archive is a download, and no size threshold separates them. Thumbnails, posters and cover art never reach this check at all — they resolve out of the node's own cache. It is a fairness control among cooperating clients, in the company of `max_concurrent_streams`, and is not a defence against a member determined to saturate a node's disk. That member is a member, and the answer to them is `member revoke`. **MNP_VERSION and MNP_MIN_SUPPORTED both move to 3.0**, on both sides. The messages are additive; the requirement is not. An opt-in switch would leave a leaseless branch reachable on every node, which is finding C6's lesson — a transport that accepted a bare JWT — one feature later. **The desktop client now checks before it connects.** The SPA is served by the hub and picks up a new client on reload; the application ships its own interface, so an un-updated one would sign in, list groups, and fail every connection with `version_too_old` — a refusal in a protocol vocabulary with nothing anyone can act on. It asks `/v1/hub/version` for `client.minimum` and says so plainly instead. An unreachable hub is deliberately *not* "too old": a captive portal or a closed laptop must not make starting the application impossible. **Every package is aligned on 0.13.0.** `meshbay-client/package.json` had drifted to 1.0.0 while the Python packages were on 0.12.0 — invisible until something compared those numbers, and then load-bearing: an installed client announcing 1.0.0 sorts above a 0.13.0 minimum and walks through the gate meant to stop it. That is stated in the code rather than left to be rediscovered; it is acceptable exactly once, because the operator is updating every client, node and hub by hand for this flag day. A new test fails if two packages ever disagree again, and another fails if the hub would refuse the client the tree builds. Node suite 1209 passed, hub suite 861 passed. Co-Authored-By: Claude Opus 5 Claude-Session: https://claude.ai/code/session_01HCGdheDLxGReuKHga3BtST