aboutsummaryrefslogtreecommitdiffstats
Commit message (Collapse)AuthorAgeFilesLines
* feat(android): send a manifest of what each backup run sentChristophe Besson9 hours13-40/+315
| | | | | | | | Photos, videos and files record, per file, their path on the node, their path and album on the phone, dates, size and SHA-256, uploaded as meshbay-manifest/manifest-<date>.jsonl once the run is done. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* feat(android): back up the files of folders the person choosesChristophe Besson9 hours36-234/+1396
| | | | | | | | A Files section takes folders through the system picker (no storage permission), refuses DCIM, Pictures and Movies, skips what the photo backup sends, and runs on the photo backup's own runner into <folder>/<account>-drive. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* feat(photos): show a phone's clips in their albumsChristophe Besson10 hours18-15/+464
| | | | | | | | Videos in a photo folder are thumbnailed by the node, counted and marked in the album, and played in the group's video player from the lightbox; the slideshow passes them by. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* feat(android): back the videos of the chosen albums up beside the photosChristophe Besson10 hours26-82/+379
| | | | | | | | A 'Videos too' option, off by default, sends them into the same YYYY/YYYY-MM folders. Files are read from the phone a ranged chunk at a time, so a large video is never whole in the page. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* feat(android): back the calendars up as an iCalendar fileChristophe Besson10 hours25-21/+495
| | | | | | | A Calendar section sends every calendar the person can edit, as a dated .ics, into <folder>/<account>-calendar once a day when it changed. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* feat(android): back up the personal profile onlyChristophe Besson11 hours6-41/+103
| | | | | | | A copy of the application inside a work profile offers no backup, and no source reads another profile. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* feat(android): back text messages up, in a build Play does not getChristophe Besson11 hours33-40/+437
| | | | | | | | A Messages section sends the SMS added since the last copy, as restorable <smses> XML, into <folder>/<account>-messages/YYYY. A play flavor has neither READ_SMS nor the code that reads messages; full is the default. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* feat(android): one backup destination, a group the account owns aloneChristophe Besson11 hours36-666/+943
| | | | | | | | Chosen once at the top of Android Sync for every kind; photos and contacts go into <folder>/<account>-photos and -contacts. Owner and sole member are checked at set-up and before every run. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* feat(hub): delete a photo from the Photos appChristophe Besson12 hours5-9/+400
| | | | | | | On a right-clicked tile and in the lightbox bar, after a confirmation, for the node's operator or the photo's uploader, as in Files. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* feat(android): back the phone's contacts up to a group of one's ownChristophe Besson12 hours31-38/+1492
| | | | | | | | A Contacts backup section on the Android Sync page sends a dated .vcf into <folder>/<account>-contacts once a day when the address book changed, only to a group the account is alone in, checked again at every run. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* feat(hub): photo backup moves to its own Android Sync pageChristophe Besson13 hours22-22/+77
| | | | | | | A Phone section of the side menu leads to it, on the Android application only; Settings no longer holds it. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(node): stop aiortc's SCTP sender leaking bytes in flightChristophe Besson14 hours3-0/+92
| | | | | | | | A chunk gap-acked then resent after T3 stayed counted for good; enough of them and the node sent nothing more on that connection. The photo backup stalled on it after a few dozen photos. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(hub): let the cast relay script end by itselfChristophe Besson14 hours1-1/+3
| | | | | | | | | | | test_cast_subtitles.py's node script called process.exit(0) right after relay.stop(), with fetch's sockets still closing. Node 24 on Windows aborted there on a libuv assertion (UV_HANDLE_CLOSING, src\win\async.c) two runs in three, after printing its results, so the module fixture failed and seven tests errored at setup. Nothing is left once the relay stops: the script ends by itself, in the same 0.1 s. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(android): back photos up under YYYY/YYYY-MM, not YYYY/MMChristophe Besson15 hours6-14/+14
| | | | | | A month folder named 08 alone read like an album number. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(hub): wait for a resumed transfer's state, not for 120 msChristophe Besson15 hours1-4/+15
| | | | | | | | | | | | | The pausable run in test_transfers.py moves in setTimeout(5) chunks, and the resume test gave it a fixed 120 ms to finish. A setTimeout(5) lasts about 15 ms on Windows' default timer, so the resumed run took 156 ms there and the test failed with nothing wrong; the 20 ms wait before pausing had the same thin margin. The run now records how far it got (state.at), and the test pauses after two chunks and waits for "done", bounded at 2 s. The assertions are unchanged, and a resume that restarts from zero is still caught (checked by introducing one in transfers.js). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(node): a name an LGPL file declares itself is its own bindingChristophe Besson15 hours1-0/+7
| | | | | | | | | | | The LGPL closure check excluded only names declared at the top level of the LGPL files, so keyring.js's `const state = () => ...`, local to createKeyring, read as a call to photo-sync.js's top-level `state` once that file was added. A name the file declares at any depth now shadows an AGPL global for that file. A real call into an AGPL module is still caught (checked by adding one to keyring.js). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(client): say when the account's node is elsewhere, and let go of itChristophe Besson15 hours17-7/+297
| | | | | | | | | | | | | | | | | | | | | | Signing in on a desktop links this machine's node to the account, but never over a key already linked (that would cut off the user's other machine) nor a node set up for another account. On a real install both left the node reading "Running" while the hub refused it in a loop, and nothing said why. And the only way to unlink was the linked machine's own Node page, of no use once that machine is gone. - The Node page works out, from the node and the hub each time it looks, whether this node can serve the signed-in account (nodeLinkProblem), and says why not: "Link this node instead" (asked first) puts this node's key on the account; "Use this node for my account" switches a node set up for another account through node:start, which takeOver now lets past a node that answers "running". The first version went through node:start for both, and clicking it on a real install did nothing: a node signed in before the account was linked elsewhere answers "running". - The Profile page unlinks the account's node (DELETE /v1/users/me/node_key), from any machine. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* feat(packaging): the Store package declares what the NSIS scripts doChristophe Besson15 hours26-152/+728
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | A test-signed install of the MSIX build, in the WindowsApps folder a Store install uses, showed that every script-made piece of the NSIS model breaks there, because each names the install folder and every update deletes it: the firewall rules went stale, the PATH entries piled up pointing at deleted folders, and the Startup-folder .vbs was refused ("Permission denied") right after sign-in. The network capabilities the manifest declared covered nothing: they make rules for sandboxed apps only, and a listener in the package still got the Windows firewall prompt. The package's own startup task was on by default, started the node whatever mode the Node page said, and ran the console executable, whose window stopped the node when closed. The package now declares what Windows then creates at install, carries across updates and removes with the app, all without an administrator prompt (each measured on the real install, through an update and a reboot): - firewall rules for the node, in a custom manifest template, since only a package-level element can hold them; - the startup task, off by default, running meshbay-nodew.exe, a new build of the daemon without a console; - an execution alias for meshbay-node.exe, so the app adds no PATH entry. The node's CLI switches the startup task (platform.startup_task, ctypes over the WinRT ABI): Windows gives the package's identity to the executables in it, not to a powershell.exe the app starts, which got "Element not found". `meshbay-node autostart install | remove | status` therefore works in the Store package from the app and a terminal alike; the app caches the answer, since the Node page polls. Starting at boot stays the .exe installer's: the Store package offers no service mode, and the CLI refuses `service install` there. Process listings count both image names. The Node page's status poll cleared the message of a refused action within five seconds; the two errors are kept apart now (all Windows builds). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* feat: back up the phone's photos to a group, once a day on Wi-FiChristophe Besson31 hours38-14/+2800
| | | | | | | | | | | | | | | | | | | | | | | | | The Android application sends the photos taken on the phone to one folder of one group chosen by the member (docs/MESHBAY_DESIGN.md §9.12). The phone lists MediaStore, keeps a ledger of what was sent and hands each photo's bytes to the page by an opaque token on the packaged origin; the page decides when a run is due and uploads through the existing path, one photo at a time under a slot. - Once a day from the last finished run, on an unmetered network only; "Back up now" asks first on mobile data. Leaving Wi-Fi stops after the file in flight. - Photos already on the phone are sent by default, newest first, under <folder>/YYYY/MM; edits are sent beside the original as -edited-<date>. - Additive by construction: nothing is ever deleted, renamed or replaced on the node, and a photo deleted on the node is not sent again. - A confirmation names the group, owner, members, folder and size when the destination or starting point changes; a lasting refusal (disk full, folder read-only or gone, no longer a member) is said once and retried a day later. - No ACCESS_MEDIA_LOCATION, so the platform redacts photo locations. - A dataSync foreground service keeps a run going with the screen off. HEIC/HEIF photos are sent but not shown in Photos yet (§15.2). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* feat(node): refuse an upload on a full disk with a stated reasonChristophe Besson31 hours18-7/+220
| | | | | | | | | | | | Nothing on the upload path knew about ENOSPC: a write that found no room raised out of the handler, the catch-all answered "Request failed", and the .part stayed behind holding the space that had run out. The node now refuses with `disk_full` at chunk 0 when the announced size would leave less than 1 GiB free, and at any write that fails with ENOSPC/EDQUOT, dropping the partial. The client carries the code on the error and the transfers panel says "The node's disk is full" in every catalogue. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(packaging): ship QE/default.env as is, the one TMDB token sourceChristophe Besson31 hours3-92/+34
| | | | | | | Both builds copy QE/default.env beside the node and stop without it. No token parsing, no other source, no MESHBAY_ALLOW_NO_TMDB. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(hub): an open application looks for new notifications when you come backChristophe Besson35 hours3-0/+80
| | | | | | | | | | | | | | The page asked for the notification list at sign-in and at a token renewal, and at no other time. A notification created after the application started, such as a chat line the hub wrote 20 ms after the message, stayed unseen until the next launch. The hub has no channel to the page and is not polled on a timer, so the list is asked for again on a gesture: the application returning to the foreground (an Android phone included) and the home page, where the list is shown. Two requests less than 30 s apart count as one. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* feat: let the operator purge a group's chat (MNP 6.1)Christophe Besson37 hours31-46/+713
| | | | | | | | Signed chat_purge from the Chat settings deletes every stored message; epoch keys and attachments stay. The ack is broadcast so open chat panels empty. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* feat: notifications on Android while closed, with nothing to installChristophe Besson37 hours41-11/+1945
| | | | | | | | | | | | | | | | | | | | The phone fetches what is new every fifteen minutes with a poll secret (POST /v1/push/poll) that reads notification lines and nothing else. When a UnifiedPush distributor is already installed, the hub also pushes at once, encrypted to the phone (RFC 8291); losing the distributor falls back to fetching. The hub now honours "disable all notifications" itself: create_notification creates nothing for that account, as it already did for a muted group, so neither switch lets anything reach a phone. The interface used to be the only reader of the account-wide switch. Push endpoints are member-supplied URLs: a send refuses non-public addresses, connects to the address it checked, and follows no redirect. Android build untested here (no SDK on this machine). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* feat: sign Android releases with the release keyChristophe Besson3 days6-12/+63
| | | | | | | assembleRelease reads the key from ~/.gradle/gradle.properties and fails without it instead of falling back to the debug key. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* perf(node): sample 9 MB with the size above 9 MB, keep known ids0.19Christophe Besson3 days8-69/+106
| | | | | | | | hash_version 3: size + first 4 MB + last 4 MB + 1 MB at the middle, 5.5x faster cold on a USB disk than the 45 MB sample. The cache now serves a hit under whatever version it holds, so no existing id moves. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(hub): re-read the roster before saying a key changedChristophe Besson3 days4-20/+194
| | | | | | | | A member invited after the roster was read showed "key changed" on each message until a reload. Read it again once per account and device on the connection, shared by concurrent messages, and pin only the final verdict. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(node): name a root after its drive when its basename is takenChristophe Besson3 days6-8/+90
| | | | | | | | | Two drives with a folder of the same name made the second add fail, and no screen could supply another name. add_root now names it "Name (H)" or "Name (parent)"; a name the operator typed is still refused on a clash. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* chore: bump version to 0.19.0Christophe Besson3 days9-9/+9
| | | | Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(client): stop the node at Quit in "only while open", whoever started itChristophe Besson3 days3-12/+81
| | | | | | | | | | | | | | | | | | | Switching from "at sign-in" to "only while MeshBay is open" left the node the sign-in launcher had started running after Quit: only a node this process had started was stopped. In that mode the app owns the node, so Quit stops the one that is there. The start with the app and the sign-in's own start (ensureNode) also both ran `autostart start` at launch -- three meshbay-node.exe were seen racing for the port. The sign-in's start and node:start now wait for the launch's. The end-to-end test covers the mode: Quit leaves no node, opening the app starts one. It launches the app with the environment it was imported with: the suite's conftest points HOME, USERPROFILE, LOCALAPPDATA and APPDATA at a throwaway directory per test, and the app started under that crashed at once (0x80000003), which first looked like a crash of the app itself. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(node): check the startup mode in the Windows end-to-end testChristophe Besson3 days1-5/+22
| | | | | | | | The boot task or the sign-in launcher, as chosen, and the node in session 0 for the service's S4U logon or in the signed-in session otherwise. Passed in service and sign-in modes on the installed 0.18.0 build. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix: set the Windows node up at sign-in, and stop it for realChristophe Besson3 days24-59/+885
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Found by the first Windows beta tester, then reproduced on a clean install. After a service-mode install nothing set the node up for the account that signed in: the boot task started a node that quit ("hub.username not set"), and the sidebar showed Node / Create group only once the hub held a node key. The only way to the wizard that provisions was the home page's welcome card, which an account already in a group never sees. The way out was `meshbay-node init` and the key pasted on the profile page -- which is also what PACKAGING-GUIDE.md told people to do. - main.js `node:ensure`, called by app.js at sign-in: provisions, starts and links the node this build ships (Windows, bundled node only). A node set up for another account, or an account linked to another node, is left alone. node:start waits for it, so the two never race. - The sidebar shows the Node section when a node exists on this machine. - The Node page's status is the node's: its control API and the process list, not the service task's state (a node started from a terminal ran while the page said Stopped). Stop says Stopped only once no meshbay-node.exe is left, and stays offered for a process that answers nothing. - CLI stop kills the pid that answered when a graceful stop does not finish, and fails with the reason when a node process is still there. - The daemon ends its process 3s after _shutdown(): Python's exit waited for a busy indexer thread, with the control API already closed. Armed by main() only, never by a daemon run inside a test. - node.toml is read as utf-8-sig (PowerShell 5.1 writes a BOM), and a config that cannot be read is logged instead of dying silently in service mode. - "Pair this browser" queues the code for the next group of this node to open instead of saying "Paired successfully"; no banner before a group. - test_e2e_windows_app.py (opt-in, MESHBAY_WIN_E2E=1) drives the installed app against a throwaway hub: fresh account to linked node, Stop, Start, Restart, checked against the real processes. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(node): drop a directory moved out of the root on Windows tooChristophe Besson3 days2-0/+19
| | | | | | | | | On Windows watchdog cannot tell what a deleted path was -- it is gone -- and reports a directory moved out of the root as a file deleted, so the fix in 3d5a168 never ran there and test_the_watcher_reports_a_directory_moved_out failed. A deleted path the index still holds entries below was a directory. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test: read UTF-8 explicitly in the licensing and keyring-vector testsChristophe Besson3 days2-26/+42
| | | | | | | | | | On Windows a bare `read_text()` or `subprocess.run(text=True)` decodes with the locale's code page (cp1252), not UTF-8. `test_licensing.py` then failed on a byte of the vendored LICENSES.txt, and `test_keyring_vectors.py` decoded the generator's output, which carries CJK test strings, into something that no longer matched keyring.json. Both files are UTF-8; say so. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(node): name members admitted without an invitation nameChristophe Besson3 days11-9/+95
| | | | | | | | | | | A member who joined by link, by a new device or into an open group was pinned in the roster with no name, so the audit log showed only the first characters of their id. The hub's MNP token now carries the account's username, and after the handshake the node writes it into the roster for an account whose name is empty. An invitation's name is never overwritten; the name is a label, authority stays on `sub`. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* perf(node): list a group's directories off the event loopChristophe Besson3 days8-121/+167
| | | | | | | | Every full index walked all roots on the loop, and a node with several large roots stopped answering for seconds. Walk directories only, on the roots' disk thread; index_sync is spawned and still answers on failure. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(hub): keep "add a directory" when one node check is missedChristophe Besson3 days2-13/+125
| | | | | | | | A node busy indexing could miss the single 3 s check after a root was added, and the button stayed hidden until a reload. Ask up to four times before deciding there is no node on this machine. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(client): remove firewall rules and boot task on uninstall, unaskedChristophe Besson3 days4-31/+54
| | | | | | | The Yes/No before it defaulted to No, so they stayed behind. Read both unelevated and raise the UAC prompt only when one is left. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(node): drop a directory moved out of the root from the indexChristophe Besson4 days3-8/+152
| | | | | | | | Watchdog reports such a move as one "directory deleted" event and nothing for the files, which the indexer ignored until the next reconcile. The freeze rules still apply: root live, directory gone, parent present. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(hub): let a download wait out a reconnect instead of failingChristophe Besson4 days4-7/+131
| | | | | | | | Chunks sent while the reconnect's connect() runs throw at once, and six retries 1.5 s apart ran out before the reconnect landed. Wait for it, up to two minutes, without spending retries. Follow-ups parked in §15.3. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(hub): move "clear finished" into the finished group's headChristophe Besson4 days3-16/+46
| | | | | | | | 9269374 let the transfers header wrap, which broke the one-line header test_layout_measured enforces. The button now sits beside what it clears, and the header is title and summary only. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(hub): stop the transfers panel scrolling sidewaysChristophe Besson4 days1-1/+7
| | | | | | | The header (title, summary, clear button) overflowed the 330 px panel in French. Let it wrap, with a gap, and hide horizontal overflow. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(node): prefer the bundled ffmpeg over PATH in the frozen buildChristophe Besson4 days2-2/+59
| | | | | | | | shutil.which never looks beside meshbay-node.exe, and node-runtime is appended to the user PATH, so any earlier ffmpeg.exe on PATH ran instead of the pinned copy. Log the resolved media tools at startup. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(hub): play video on iPhone through ManagedMediaSourceChristophe Besson4 days12-2/+25
| | | | | | | An iPhone has no MediaSource; every film was refused as an unsupported codec. A browser with neither now says so. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* feat(hub): say "N groups unreachable" on Search for a few seconds, not for good0.18Christophe Besson5 days7-34/+86
| | | | | | | | | | | | The line sat above the results for as long as the page was open. It is now said once a cross-group pass is over, for five seconds, in the same passing note as "Link copied" — moved out of copy-link.js into note.js (`say(text, ms)`), one note at a time for the whole page. The `.search-unreachable` rule goes with the line it styled. The copy-link probe now also checks in Chrome that the note goes by itself. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* test(hub): join from an invitation link and land under the group's handleChristophe Besson6 days2-5/+35
| | | | | | | | | | | The invitation probe listed no groups, so after Join the address stayed the #/group/<id> the button navigated to and the group links' rewrite to #/name@owner was never exercised on that path. A third case has the hub list the group once joined, as it does: the group page opens, the address shows the handle without a history entry of its own, and the code still never reaches the hub. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* feat: copy a file's or folder's #/name@owner link from Files, Music, Photos ↵Christophe Besson6 days23-20/+567
| | | | | | | | | | | | | | | | | | | | | | | and Search "Copy link" puts the address group-link.js resolves on the clipboard, on the hub's origin rather than the page's, so a link copied in the desktop application is not app://meshbay. Files offers it for one row, from the right-click menu or the toolbar with one row ticked (a phone's way in); Music on one track's menu, whose dots a phone has; Photos on a right-clicked tile and in the lightbox's bar. The video player and the file preview carry a link button next to Download. Applications get a `linkFor(entry | folderPath)` prop (MESHBAY_DESIGN.md §9.2) and offer the action only when it names a link. The group page builds it from the hub's row; Search from each result's own group and its path before the merged views prefixed it, and names no link for a folder of the merged tree, which a group name alone does not identify. harness/copy_link_probe.py mounts the three applications in Chrome and reads what reached the clipboard. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* feat: open a group, a folder or a file from a #/name@owner linkChristophe Besson6 days20-12/+650
| | | | | | | | | | | | | | | | | | | A group can now be reached by the handle shown under its name, and a path after it points inside the group: #/name@owner/root/dir/file downloads the file and opens Files on its folder; a folder opens Files there. The handle is resolved in the client against the account's own /v1/groups/mine, so no hub route answers for a name and nobody can probe for one. While a group is open the address shows the handle (replace, no history entry); a linked path is taken out of the address once acted on, so a reload does not download twice. Signing in no longer sends everyone home: the form stood in for the page the address named, and that is where a link opened signed out was going. group-link.js holds the parsing and lookups, executed whole by test_group_link.py; harness/group_link_probe.py drives the router in Chrome. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix: list as members only the accounts the node has admittedChristophe Besson6 days14-7/+60
| | | | | | | | | | | | | | | | The Members list showed the hub's membership, which an account gains when it accepts the invitation or redeems a link, before it has presented its code to the node. The node's roster is the authority (MESHBAY_DESIGN.md §3.4), so the list now crosses the hub's members with the sealed group roster the node already sends every connected member. An account the node has not admitted yet is shown to the owner alone, as waiting for its code, with the Remove button; other members do not see it. When the roster cannot be read, the hub's list is shown as before. groupRoster() takes { fresh: true } so the page sees who joined since the connection opened. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* fix(node): plug an auto-ejected removable root back once its files returnChristophe Besson6 days8-10/+247
| | | | | | | | | | | | | | | | | A node started with the desktop session runs before the session has mounted its USB drives. The safety net then auto-ejected every removable root and persisted it exactly like an operator's eject, so after each reboot those roots stayed ejected until someone plugged them by hand (seen on a node whose /media drives were mounted a minute after it started). An auto-eject is now stored as such ("auto" in roster.db). At startup and at every reconcile, an auto-ejected root whose path is readable again is checked against a few files the hash cache knows under it, at the same path with the same size and mtime; one found and the root is plugged back and rescanned. An empty mount point or another drive in its place is not recognised and stays ejected. An operator's eject is never undone automatically. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
generated by cgit v1.2.3 (git 2.51.0) at 2026-10-10 23:13:14 +0000