Repository navigation
Releases
OMARemote 0.9.17
What's new
- Faster sessions. The session view now uploads only the parts of the picture that changed to the
GPU, not the whole desktop on every update. Dragging a window over a Retina-sized desktop went from
20% to 3% CPU in the view, and a blinking text caret from 30 MB to under 1 KB per update. - VNC Low is 16-bit colour. Low quality asks the server for 16-bit colour on top of strong
compression: a third less data from any server. It matters most for a Mac, whose Screen Sharing has
no Tight or JPEG. Over a 100 Mbit link a Retina Mac went from 5 to 10 updates a second. - No more stale patches in RDP. Dragging a window around fast could leave bits of the old picture
on screen until something redrew them. Two threads wrote to the window at the same time and lines got
lost; that cannot happen any more, and a large burst of changes ends with one full redraw. - Clipboard, Gliff: nothing moves at connect. Each machine keeps its own clipboard until something
is copied during the session, so a clipboard history on either side never picks up what the other
held. A new per-connection switch, Copied passwords, also shares what a password manager copies (off
by default, never synced, not set by groups). - Clipboard, RDP and VNC: when the clipboard here is cleared (a password manager clearing a copied
password, for example), or you copy something else while the session is not focused, the server loses
what it was offered too. Switching Clipboard off mid-session does the same. A copy made on the server
itself stays there. VNC now behaves like Gliff at connect. - Hotkey windows hang from the top edge of the screen, like Omarchy's console, and are in place
before they slide in. - RDP at 125% or 150%: the picture starts on a whole screen pixel, so fonts no longer look
pixelated in a tab of an odd width. - Measuring: VNC and RDP sessions write a
stats:line to their log for each busy second (updates,
data, decoding time), to see where a slow session's time goes.
Security review
The changes since 0.9.16 were reviewed against docs/security-review.md, and the Reviewer reviewed every
commit and ran the omarchy-plugin-security skill on them; nothing open. Checked in particular:
- Copied passwords is this computer's own: it is in
LOCAL_KEYSandGROUP_LOCAL_SETTINGS, so no
sync, import or group file can switch it on, and a connection with it on counts as signing in here, so
a synced change of its address waits for you. - Rectangles from a server or session are clipped to the picture before anything is uploaded; the
view'sview-statsline is checked to be digits before it is logged. - The 16-bit picture buffer is sized from the same checked desktop size as the 32-bit one.
- The clipboard watch (
wl-paste --watch) only reports that the clipboard changed, never reads its
contents into OMARemote, and ends with the window that started it. - The session link is now safe to write from two threads; a write that fails shuts the socket down and
leaves closing it to the main thread.
OMARemote is in early development: it works day to day, but expect rough edges. Please report what you run into: https://github.com/RFdeGroot/OMARemote/issues
Install on Omarchy, on Intel/AMD (x86_64) or Apple Silicon (aarch64); pacman pulls in every dependency:
curl -LO https://github.com/RFdeGroot/OMARemote/releases/download/v0.9.17/omaremote-0.9.17-1-$(uname -m).pkg.tar.zst
sudo pacman -U ./omaremote-0.9.17-1-$(uname -m).pkg.tar.zstThe package is signed with OMARemote's release key; OMARemote's Update button checks that before installing. To check it by hand, with the key from packaging/arch/omaremote-release.gpg in the repository: curl -L -o omaremote.sig https://github.com/RFdeGroot/OMARemote/releases/download/v0.9.17/omaremote-0.9.17-1-$(uname -m).pkg.tar.zst.sig, then gpgv --keyring ./omaremote-release.gpg omaremote.sig omaremote-0.9.17-1-$(uname -m).pkg.tar.zst. Keep the signature under that other name: pacman checks a .sig it finds beside the package against its own keyring, which does not have OMARemote's key, and refuses the install.
Updating from an earlier release works the same way: pacman upgrades in place. Restart OMARemote afterwards.
Or build it from source with the attached PKGBUILD: makepkg -si. See the README for more.
OMARemote 0.9.16
What's new
- Theme: the windows no longer read Omarchy's theme files on the UI thread, so a theme file
that is slow or never ends cannot freeze them. A window shows once its theme is in (at the latest
after a second), so it never flashes the default colours first. A theme switch still applies
live. - Own windows open without a flash. Hyprland first opens a new window tiled; OMARemote now has
everything ready before the window shows and floats, sizes and centres it in one step, with only
the background visible until then (measured: 52–85 ms instead of up to 670 ms). - Super+C / Super+V in an SSH session in its own window now copy and paste, as they do in a
tab. The window never got Hyprland's terminal tag before.
Security review
The Reviewer reviewed every change and ran the omarchy-plugin-security skill on it; nothing open.
This closes the last item of its second sweep (S5): theme files are read in the background, and
text longer than 64 K characters is not parsed. Tested with a theme file that is a FIFO, a link to
/dev/zero and 300 MB: the window shows with the default colours and quits cleanly. A size bound on
the read itself is not possible here, because Omarchy's Commons.Color reads the same files whole
in the same process; that bound belongs in Omarchy (an issue is drafted). Placement commands sent to Hyprland are built
from Hyprland's own window address and whole numbers only.
OMARemote is in early development: it works day to day, but expect rough edges. Please report what you run into: https://github.com/RFdeGroot/OMARemote/issues
Install on Omarchy, on Intel/AMD (x86_64) or Apple Silicon (aarch64); pacman pulls in every dependency:
curl -LO https://github.com/RFdeGroot/OMARemote/releases/download/v0.9.16/omaremote-0.9.16-1-$(uname -m).pkg.tar.zst
sudo pacman -U ./omaremote-0.9.16-1-$(uname -m).pkg.tar.zstThe package is signed with OMARemote's release key; OMARemote's Update button checks that before installing. To check it by hand, with the key from packaging/arch/omaremote-release.gpg in the repository: curl -L -o omaremote.sig https://github.com/RFdeGroot/OMARemote/releases/download/v0.9.16/omaremote-0.9.16-1-$(uname -m).pkg.tar.zst.sig, then gpgv --keyring ./omaremote-release.gpg omaremote.sig omaremote-0.9.16-1-$(uname -m).pkg.tar.zst. Keep the signature under that other name: pacman checks a .sig it finds beside the package against its own keyring, which does not have OMARemote's key, and refuses the install.
Updating from an earlier release works the same way: pacman upgrades in place. Restart OMARemote afterwards.
Or build it from source with the attached PKGBUILD: makepkg -si. See the README for more.
OMARemote 0.9.15
What's new
A security release. Nothing changes in how OMARemote looks or works.
- connections.json, last-connected.json and session logs are read through
omaremote-session read.
It reads only ordinary files, with size, nesting and count limits, so the window never reads
them whole itself. The bar plugin usesread favourites(plugin 0.4.17). - connections.json holds at most 2000 connections, 200 groups, 200 credential sets and 2000 in the
trash. Sync, import and the window add no more, and say why. - Sync merges in linear time: a round against a full library takes seconds, not minutes. A sync
file holding more than a library would is not read. Only deletions of what this system knows
are kept, and a deletion wins over a stale copy whichever file is read first. - Opening a connection while OMARemote is closed (the bar plugin's click) waits for the library
to load. An unreadable library is retried.
Security review
The Reviewer reviewed this work in four rounds, then ran a second full omarchy-plugin-security
sweep of the whole tree (backend, QML, native code, packaging, CI). Every finding was fixed with a
test: JSON parsing that could escape on deep nesting, a FIFO that could hang the last-connected
stamp or a session record rewrite, FileViews that read whole files on every change, and least
privilege in the release workflow. One hardening item (bounded reads of the theme files) is
deferred.
OMARemote is in early development: it works day to day, but expect rough edges. Please report what you run into: https://github.com/RFdeGroot/OMARemote/issues
Install on Omarchy, on Intel/AMD (x86_64) or Apple Silicon (aarch64); pacman pulls in every dependency:
curl -LO https://github.com/RFdeGroot/OMARemote/releases/download/v0.9.15/omaremote-0.9.15-1-$(uname -m).pkg.tar.zst
sudo pacman -U ./omaremote-0.9.15-1-$(uname -m).pkg.tar.zstThe package is signed with OMARemote's release key; OMARemote's Update button checks that before installing. To check it by hand, with the key from packaging/arch/omaremote-release.gpg in the repository: curl -L -o omaremote.sig https://github.com/RFdeGroot/OMARemote/releases/download/v0.9.15/omaremote-0.9.15-1-$(uname -m).pkg.tar.zst.sig, then gpgv --keyring ./omaremote-release.gpg omaremote.sig omaremote-0.9.15-1-$(uname -m).pkg.tar.zst. Keep the signature under that other name: pacman checks a .sig it finds beside the package against its own keyring, which does not have OMARemote's key, and refuses the install.
Updating from an earlier release works the same way: pacman upgrades in place. Restart OMARemote afterwards.
Or build it from source with the attached PKGBUILD: makepkg -si. See the README for more.
OMARemote 0.9.14
What's new in 0.9.14
A security release. Nothing changes in how OMARemote looks or works, unless a file it reads is far larger or stranger than any real one.
- Sync and import tell you what they skipped. When the sync folder or an import holds more than OMARemote takes in, the report says so in one line ("3 files skipped: …") and everything else still comes through.
- If this system's own sync file grows past 4 MiB, the sync report says so: other systems no longer read a file that large.
Security review
Reviewed everything since 0.9.13 against docs/security-review.md. The Reviewer session reviewed every change, ran the omarchy-plugin-security skill over it, and then swept the whole code (not only the changes) for the same classes, so they all ship in this one release.
- Marketplace finding #12587 (HANCORE-linux): sync had no limit on what it read in total. Each file in the sync folder was capped at 16 MiB, but a round read any number of them and kept them all, so whoever can write the sync folder could fill the memory of OMARemote's backend, and the window read the answer whole. A round now takes in at most 4 MiB per system, 16 MiB in all and 64 files (newest first). It reads nothing from a folder holding more than 1024 OMARemote files. The answer to the window is refused above 16 MiB, and the sync command runs with a 1 GiB memory ceiling.
- Found on the way, fixed:
- A file nested very deeply stopped sync for good.
[[[[…crashed the backend's JSON reader with an error nothing caught. Caught, a value 2000 levels deep inside a connection was still merged, and the window's own JSON reader (at most 1024 levels) then refused every round. Every JSON file OMARemote reads is now refused deeper than 64 levels (a real one is about 5). - Import from other clients. A connection the import cannot take (a 400 KB user name, say) still carried its fields to the window: 3 such
.rdpfiles made a 2.3 MB answer. It now shows only a short name. An import reads at most 64 MiB, 500 files and 2000 connections, names at most 100 problems, and its answer is capped like sync's. - Importing an OMARemote export file is held to the same limits as sync.
- The session list.
omaremote-session list(read whole by the window and the bar plugin) read every record in the sessions folder. It now shows the newest 256, each at most 16 KiB. A record could also name its own id:../victimmade the clean-up of old sessions delete a file outside the folder, and a number as id crashed the list. A record now counts only under its own file name, and an id is never a path. - The log view opens only the session's own log, and only when it is an ordinary file.
- Your own configs.
hyprland.lua,~/.ssh/configand OMARemote's managed files are read with a size limit, never through a FIFO (which hung the command), and text that is not UTF-8 is left alone rather than written back changed. A rewritten config gets its permissions on the open file itself. - connections.json. A new one is made readable only by you (0600); the
.brokencopy of a file that does not read is never written through a link. It is still written in place: an atomic write in Quickshell 0.3.2 makes the window reload the old text right after saving the new. - Hyprland's answers (window and monitor lists) are read up to 8 MiB with a deadline; the folders walked for window rules and SSH keys have a file limit.
- The release build's container is pinned by digest, as its actions already were by commit.
- A file nested very deeply stopped sync for good.
- Tests that fail without each fix: the sync limits (count, size, total, nesting, the answer, the memory ceiling), the import limits and refused entries, the session ids and log paths, configs that are a FIFO or not UTF-8, and connections.json's mode and
.brokencopy against the real Store in Quickshell.
OMARemote is in early development: it works day to day, but expect rough edges. Please report what you run into: https://github.com/RFdeGroot/OMARemote/issues
Install on Omarchy, on Intel/AMD (x86_64) or Apple Silicon (aarch64); pacman pulls in every dependency:
curl -LO https://github.com/RFdeGroot/OMARemote/releases/download/v0.9.14/omaremote-0.9.14-1-$(uname -m).pkg.tar.zst
sudo pacman -U ./omaremote-0.9.14-1-$(uname -m).pkg.tar.zstThe package is signed with OMARemote's release key; OMARemote's Update button checks that before installing. To check it by hand, with the key from packaging/arch/omaremote-release.gpg in the repository: curl -L -o omaremote.sig https://github.com/RFdeGroot/OMARemote/releases/download/v0.9.14/omaremote-0.9.14-1-$(uname -m).pkg.tar.zst.sig, then gpgv --keyring ./omaremote-release.gpg omaremote.sig omaremote-0.9.14-1-$(uname -m).pkg.tar.zst. Keep the signature under that other name: pacman checks a .sig it finds beside the package against its own keyring, which does not have OMARemote's key, and refuses the install.
Updating from an earlier release works the same way: pacman upgrades in place. Restart OMARemote afterwards.
Or build it from source with the attached PKGBUILD: makepkg -si. See the README for more.
OMARemote 0.9.13
What's new in 0.9.13
- SSH sessions reconnect by themselves. When a working connection drops (the computer restarts, you move from cable to wifi), the session starts ssh again in the same terminal, the screen and scrollback kept. A banner at the top says what happens ("Connection lost · trying again in 4 s", "reconnecting to pc.lan…"), with Try now (
⏎) and Cancel. It tries with growing pauses for up to 10 minutes. It is a new shell: programs that ran on the server are gone unless they ran in tmux or screen. A refused sign-in, a host key that is not accepted or a refusal before signing in is never tried again. Reconnect is on by default and can be switched off per SSH connection. - Much less CPU with a session on screen. Since the first release, a hidden "Connecting…" animation kept running in every open session (RDP, VNC, Gliff and SSH alike), and the window redrew at the screen's refresh rate the whole time: some 40% of a CPU core on a 240 Hz monitor. It now only runs while a session is actually connecting.
- A session that ended is shown on a card. The "Connection failed" / "Disconnected" panel sits on its own card over the session's last picture, dimmed, instead of across the remote text.
- ssh failures in plain words. "client_loop: send disconnect: Broken pipe" now reads "The connection was lost: the computer may have restarted, or the network changed"; refused connections, unknown names, unreachable networks, refused sign-ins and host keys get a sentence too. ssh's own words stay in the log.
Security review
Reviewed everything since 0.9.12 against docs/security-review.md. The Reviewer session reviewed every change and ran the omarchy-plugin-security skill (built from the marketplace's own review comments) over the whole release until it found nothing more.
- Marketplace finding #12422 (HANCORE-linux): a Gliff server could make a file paste write outside its folder. A new clipboard offer arriving while files were being pasted replaced the offer under the running paste: its next file came from the new list, with no folder, so it was created relative to where OMARemote was started (usually your home folder). Each paste now keeps its own copy of the offer and of its folder (as an open descriptor), creates every file and folder beneath that descriptor without following links (
openat2), and a new offer ends every paste still running. - Found with the skill, all fixed:
- Stopping a session could signal another program.
omaremote-session stopsignalled the process number in the session's record; after a crash that number can belong to another program. Every record now keeps its processes' start times, and a signal only goes to that same process (pinned with a pidfd first). - A server could grow the session log without limit. The log is in memory (the runtime folder) and the log view reads it whole. It is capped at 8 MiB, lines at 8 KiB, and the rest of a long line can no longer pass for a line of its own.
- Files of the user's, read by name after a check. Hyprland config,
known_hosts, ssh keys, terminal configs and OMARemote's own state files are now read through one open: only an ordinary file, never a FIFO that would hang, with a size limit. - Own files written through a planted link. The Kerberos cache and config are replaced atomically; the backups of
hyprland.luaand~/.ssh/configare created new, and when something else sits at a backup's name the config is left alone rather than edited without a backup.
- Stopping a session could signal another program.
- Reconnecting cannot be steered into dialling forever. Each try is judged by its own output; 20 drops in a row with nothing typed in between end it, so a server that keeps saying "Broken pipe" cannot keep the client calling.
- Tests that fail without each fix: the clipboard swap against the real clipboard code, process identity before signals, the log cap, FIFOs and links at file names, the backups, and SSH reconnect against the real
omaremote-sshwith a stand-in ssh. - Left as they are (hardening, same-user only): the window reads its own backend's output and its own config files whole.
OMARemote is in early development: it works day to day, but expect rough edges. Please report what you run into: https://github.com/RFdeGroot/OMARemote/issues
Install on Omarchy, on Intel/AMD (x86_64) or Apple Silicon (aarch64); pacman pulls in every dependency:
curl -LO https://github.com/RFdeGroot/OMARemote/releases/download/v0.9.13/omaremote-0.9.13-1-$(uname -m).pkg.tar.zst
sudo pacman -U ./omaremote-0.9.13-1-$(uname -m).pkg.tar.zstThe package is signed with OMARemote's release key; OMARemote's Update button checks that before installing. To check it by hand, with the key from packaging/arch/omaremote-release.gpg in the repository: curl -L -o omaremote.sig https://github.com/RFdeGroot/OMARemote/releases/download/v0.9.13/omaremote-0.9.13-1-$(uname -m).pkg.tar.zst.sig, then gpgv --keyring ./omaremote-release.gpg omaremote.sig omaremote-0.9.13-1-$(uname -m).pkg.tar.zst. Keep the signature under that other name: pacman checks a .sig it finds beside the package against its own keyring, which does not have OMARemote's key, and refuses the install.
Updating from an earlier release works the same way: pacman upgrades in place. Restart OMARemote afterwards.
Or build it from source with the attached PKGBUILD: makepkg -si. See the README for more.
OMARemote 0.9.12
What's new in 0.9.12
- Fullscreen for RDP, for real. Resolution → Fullscreen now opens the connection in a window of its own that fills the screen, straight away. Choose Whole screen (the Omarchy bar hidden, only the remote desktop showing) or In a window (maximised, the bar stays). In Whole screen, rest the pointer on the top edge: a bar slides in with Exit fullscreen, To a tab and Disconnect, each with its key. Fullscreen opens in its own window, so a tab or a hotkey window is not offered with it; a hotkey window can still go fullscreen with
ctrl+alt+⏎while it is open. ctrl+alt+⏎in a tab works like Omarchy's SUPER+F. The manager goes fullscreen with the Omarchy bar hidden and keeps its own tabs, pinned list and status bar.- Windows come back where they belong. An own window leaving fullscreen comes back centred (Hyprland put it partly off the screen). A Hyprland window rule of your own for OMARemote's windows (
omaremote-session) is now respected: OMARemote no longer floats, sizes or centres them over it. - Gliff asks before trusting a new computer. The first connection to a computer used to fail with ssh's "Host key verification failed.", which read like a wrong key. The tab now asks "Trust pc.lan?" with its key's fingerprint; Trust lets ssh add it to
~/.ssh/known_hosts, as answering yes in a terminal does. A changed key is refused, as ssh refuses it, with a clear warning and how to clear the old key if the machine was reinstalled. A key of a new type for a known computer gets a warning in the question. - Gliff no longer connects to its own computer. Streaming a computer into itself only echoes itself. A connection that reaches this computer (by its address, its name, or its ssh host key) is greyed in the list, marked "this computer", and refuses to connect, also when it arrived through sync from another computer. A clone with the same ssh host keys can be marked It's another computer (this computer only, never synced, for that host and port).
- "All monitors" is gone. FreeRDP's multi-monitor mode measures every monitor wrongly on Hyprland and cannot connect. It comes back once it works; a connection that had it opens as before otherwise.
Security review
Reviewed everything since 0.9.11 against docs/security-review.md; the Reviewer session reviewed each change as it was made.
- Marketplace finding #12352 (HANCORE-linux): a server could make the session allocate without bound. Gliff's H.264 decoder took its size from the server's stream with no limit before decoding. Now a stream larger than the 3840×2160 the client asks for ends the session, and a picture larger than its stream is refused when FFmpeg first learns its size, before it allocates anything for it (
get_format, plus FFmpeg's own pixel limit). The same pass found the same class in RDP: FreeRDP's graphics pipeline let a server create surfaces up to 65535×65535 (17 GB each), as many as it liked, and copy any of them into a cache with no total. Surfaces are now capped at 8192 a side and 2 GiB together (RemoteFX tiles counted), the cache at 128 MiB (the RDP specification's limit is 100 MB). Every session process also has a 4 GiB memory ceiling for what a library might still allocate, and reaching it ends the session cleanly instead of crashing. - The new host-key question cannot be posed by a server. ssh hands a server's own login questions to the same helper; only a question that starts with ssh's own wording is shown as "Trust …?", so a server cannot bring up that dialog. Only "yes" after Trust is ever answered, and ssh itself still decides and writes known_hosts.
- Command lines stay anonymous. Finding "this computer" runs in-process (addresses, known_hosts with hashed entries, key fingerprints computed here); the only new command line with a host is a Gliff session's
ssh -Gjust before its ownssh, documented with it. - Found in this review: the Hyprland rule scan and the known_hosts reader opened files by path without checking they were regular files; a FIFO by such a name would hang the placement. Both read regular files only now, with a test that fails without the fix.
- Tests that fail without each fix: oversized streams and pictures against a fake Gliff server, the RDP surface and cache limits, the memory ceiling, the host-key question (a fake ssh using the real askpass path, and the real OpenSSH 10.6 checked by hand), and this computer's detection.
OMARemote is in early development: it works day to day, but expect rough edges. Please report what you run into: https://github.com/RFdeGroot/OMARemote/issues
Install on Omarchy, on Intel/AMD (x86_64) or Apple Silicon (aarch64); pacman pulls in every dependency:
curl -LO https://github.com/RFdeGroot/OMARemote/releases/download/v0.9.12/omaremote-0.9.12-1-$(uname -m).pkg.tar.zst
sudo pacman -U ./omaremote-0.9.12-1-$(uname -m).pkg.tar.zstThe package is signed with OMARemote's release key; OMARemote's Update button checks that before installing. To check it by hand, with the key from packaging/arch/omaremote-release.gpg in the repository: curl -L -o omaremote.sig https://github.com/RFdeGroot/OMARemote/releases/download/v0.9.12/omaremote-0.9.12-1-$(uname -m).pkg.tar.zst.sig, then gpgv --keyring ./omaremote-release.gpg omaremote.sig omaremote-0.9.12-1-$(uname -m).pkg.tar.zst. Keep the signature under that other name: pacman checks a .sig it finds beside the package against its own keyring, which does not have OMARemote's key, and refuses the install.
Updating from an earlier release works the same way: pacman upgrades in place. Restart OMARemote afterwards.
Or build it from source with the attached PKGBUILD: makepkg -si. See the README for more.
OMARemote 0.9.11
What's new in 0.9.11
- The update check gives up on time, whatever the server does. 0.9.10 gave the releases list (which OMARemote reads to see whether there is an update for itself or the bar plugin) a one-minute limit, but that limit only started once the server had sent its headers. A server sending those headers very slowly could keep the check waiting far longer. The minute now covers the whole request, headers and redirects included.
- A garbled answer no longer breaks the check. If the server's answer was not valid HTTP, the update and plugin checks stopped with an error instead of reporting "no update found". They now treat it like any other failed download.
Thanks
The Reviewer session found both while reviewing 0.9.10's download limits.
Security review
Reviewed everything since 0.9.10 against docs/security-review.md. The only change is this fix.
- The deadline covers the whole transfer.
fetch_jsoninomaremote-sessionnow runs the request (connecting, status line, headers, every redirect and the body) in a background thread and gives up when the deadline passes. The byte cap and the deadline still hold inside that thread, so a body that keeps coming stops there too. A thread stuck in the headers is left behind; that is safe because the update and plugin checks are short command-line runs that exit right after. - Errors stay errors.
http.client.HTTPException(a bad status line, a header line that is too long) becomes anOSError, which both checks already handle. - New tests use a real local server: one that sends its headers a byte at a time (11 s past a 0.5 s deadline before the fix) and one that sends a garbled status line (a traceback before). Both fail without the fix.
- The lesson in
docs/security-review.mdnow says that a deadline covers headers and redirects, not only the body.
OMARemote is in early development: it works day to day, but expect rough edges. Please report what you run into: https://github.com/RFdeGroot/OMARemote/issues
Install on Omarchy, on Intel/AMD (x86_64) or Apple Silicon (aarch64); pacman pulls in every dependency:
curl -LO https://github.com/RFdeGroot/OMARemote/releases/download/v0.9.11/omaremote-0.9.11-1-$(uname -m).pkg.tar.zst
sudo pacman -U ./omaremote-0.9.11-1-$(uname -m).pkg.tar.zstThe package is signed with OMARemote's release key; OMARemote's Update button checks that before installing. To check it by hand, with the key from packaging/arch/omaremote-release.gpg in the repository: curl -L -o omaremote.sig https://github.com/RFdeGroot/OMARemote/releases/download/v0.9.11/omaremote-0.9.11-1-$(uname -m).pkg.tar.zst.sig, then gpgv --keyring ./omaremote-release.gpg omaremote.sig omaremote-0.9.11-1-$(uname -m).pkg.tar.zst. Keep the signature under that other name: pacman checks a .sig it finds beside the package against its own keyring, which does not have OMARemote's key, and refuses the install.
Updating from an earlier release works the same way: pacman upgrades in place. Restart OMARemote afterwards.
Or build it from source with the attached PKGBUILD: makepkg -si. See the README for more.
OMARemote 0.9.10
What's new in 0.9.10
- Updates download within limits. The Update button downloads the new package and its signature first, then checks the signature. Until now, those downloads had no size or time limit: a server pretending to be GitHub could keep sending data until your disk was full, before the check ever ran. Each download now has a size cap (64 MiB for the package, which is about 0.5 MB today, and 64 KiB for the signature) and a time limit, and it only goes over HTTPS, redirects included. The same goes for the releases list that OMARemote reads to see whether there is an update for itself or for the bar plugin (4 MiB, one minute).
Thanks
The marketplace reviewer (HANCORE-linux) found this in the bar plugin's install button, in #12133. OMARemote's own updater had the same gap.
Security review
Reviewed everything since 0.9.9 against docs/security-review.md. The only change is this fix.
- Downloaded before it is checked, now bounded.
omaremote-update:curl --max-filesize(which also stops a stream that announces no length, at the cap),--max-time,--connect-timeout,--protoand--proto-redir '=https'on the package and the signature.omaremote-session: the update check and the plugin check read the releases list through one function that reads at most 4 MiB within 60 s in all (a socket timeout alone counts per read, so a slow trickle never tripped it) and refuses a redirect away from HTTPS. - The whole class checked. Every other place that reads data from outside already had a byte cap and a deadline: sync and import files, the clipboard, session messages, Gliff. The release workflow's Arch Linux ARM image, also downloaded before its signature is checked, now has a cap and a deadline too.
- New tests: every download carries its cap and deadline, a failed download installs nothing, and a list past the cap, a trickle past the deadline and a redirect to plain HTTP are all refused. They fail without the fix.
- The plugin uses the same limits; they are now part of
docs/plugin-interface.md.
OMARemote is in early development: it works day to day, but expect rough edges. Please report what you run into: https://github.com/RFdeGroot/OMARemote/issues
Install on Omarchy, on Intel/AMD (x86_64) or Apple Silicon (aarch64); pacman pulls in every dependency:
curl -LO https://github.com/RFdeGroot/OMARemote/releases/download/v0.9.10/omaremote-0.9.10-1-$(uname -m).pkg.tar.zst
sudo pacman -U ./omaremote-0.9.10-1-$(uname -m).pkg.tar.zstThe package is signed with OMARemote's release key; OMARemote's Update button checks that before installing. To check it by hand, with the key from packaging/arch/omaremote-release.gpg in the repository: curl -L -o omaremote.sig https://github.com/RFdeGroot/OMARemote/releases/download/v0.9.10/omaremote-0.9.10-1-$(uname -m).pkg.tar.zst.sig, then gpgv --keyring ./omaremote-release.gpg omaremote.sig omaremote-0.9.10-1-$(uname -m).pkg.tar.zst. Keep the signature under that other name: pacman checks a .sig it finds beside the package against its own keyring, which does not have OMARemote's key, and refuses the install.
Updating from an earlier release works the same way: pacman upgrades in place. Restart OMARemote afterwards.
Or build it from source with the attached PKGBUILD: makepkg -si. See the README for more.
OMARemote 0.9.9
What's new in 0.9.9
- VNC desktops are solid again. Some VNC servers, macOS Screen Sharing among them, showed up see-through: your own wallpaper shone through the remote windows. A VNC pixel has four bytes, and the fourth is padding that a server may fill with anything. OMARemote showed the picture as if that byte were its opacity. It now sets that byte to fully opaque in every part of the screen the server updates, and the colours stay exactly as sent.
Thanks
This fix is the work of xiemin (@peterxiemin), in #3. They found the cause on a 5120x2880 macOS session and wrote a test with a mock VNC server that checks every byte of the shared framebuffer. Thank you, xiemin!
Security review
Reviewed everything since 0.9.8 against docs/security-review.md. The only change is the VNC fix: the server's pixels are written into OMARemote's own framebuffer, only inside the rectangle the server just updated.
- No write outside the framebuffer. libvncclient refuses a rectangle that does not fit the desktop and ends the session ("Rect too large") before it reaches OMARemote. The desktop itself is at most 8192 pixels a side (since 0.9.7).
- No new cost a server could not already cause. The pass over the rectangle is cheaper than decoding it: about 6 ms for a whole 5K screen.
- The new test talks to a mock server on 127.0.0.1 only.
Nothing found.
OMARemote is in early development: it works day to day, but expect rough edges. Please report what you run into: https://github.com/RFdeGroot/OMARemote/issues
Install on Omarchy, on Intel/AMD (x86_64) or Apple Silicon (aarch64); pacman pulls in every dependency:
curl -LO https://github.com/RFdeGroot/OMARemote/releases/download/v0.9.9/omaremote-0.9.9-1-$(uname -m).pkg.tar.zst
sudo pacman -U ./omaremote-0.9.9-1-$(uname -m).pkg.tar.zstThe package is signed with OMARemote's release key; OMARemote's Update button checks that before installing. To check it by hand, with the key from packaging/arch/omaremote-release.gpg in the repository: curl -L -o omaremote.sig https://github.com/RFdeGroot/OMARemote/releases/download/v0.9.9/omaremote-0.9.9-1-$(uname -m).pkg.tar.zst.sig, then gpgv --keyring ./omaremote-release.gpg omaremote.sig omaremote-0.9.9-1-$(uname -m).pkg.tar.zst. Keep the signature under that other name: pacman checks a .sig it finds beside the package against its own keyring, which does not have OMARemote's key, and refuses the install.
Updating from an earlier release works the same way: pacman upgrades in place. Restart OMARemote afterwards.
Or build it from source with the attached PKGBUILD: makepkg -si. See the README for more.
OMARemote 0.9.8
What's new in 0.9.8
A Windows desktop that locked itself comes back signed in. Windows locks a session after a while without input. Then you had to type your password inside the session again, which is especially clumsy in a hotkey window meant for quick access. A reconnect with your saved credentials brings you back into the same session past the lock screen, with your programs and work as they were. OMARemote now does that for you:
ctrl+alt+rreconnects an RDP session in place, in a tab, an own window or a hotkey window. The old session stays on screen until the new one is up, then the new one takes its place. One that takes longer than 5 seconds is shown anyway, in case it is asking for a password, with the old one running behind it until it connects. A reconnect that fails goes back to the old session.- Unlock on return (connection or group editor, RDP, off by default): when you come back to a session after 5, 10, 15, 30 or 60 minutes away (another tab, another window, a hidden hotkey window), it reconnects. RDP does not tell whether Windows locked, so pick the remote's own lock time or a bit more. A reconnect that was not needed costs a few seconds.
- The password is never typed into the session. A reconnect is a new sign-in, where FreeRDP hands it to Windows over NLA, exactly as when you first connect. Without a saved password you are asked as usual. The remote's own lock stays: while you are away it locks as its owner wants.
Security review
Reviewed against docs/security-review.md, and by the Reviewer session in three rounds (all findings fixed before this release: closing a tab or moving it to a window during a reconnect could leave a session running unseen; a replaced session's end was reported as a failure).
- Unlock on return is this system's own. It signs in by itself, so it never syncs or exports, on connections and groups, and a sync or import file can neither set nor clear it (tested). A value other than a whole number from 1 to 1440 counts as off.
- No new credential path. A reconnect is an ordinary launch of the connection as saved here: a synced address change still waits for your confirmation, and the password goes only to FreeRDP's stdin and Windows' sign-in.
- It only reconnects when you are back (the session's tab or window has the focus again), never while you are away.
- A "keep the remote awake" option was considered and dropped: it would defeat the remote's own lock policy.
- Left as is: a reconnect started in an own window that is closed before the new session shows up leaves that session running in the background. The manager adopts it into a tab when it starts and stops it when it closes, and the bar lists it.
OMARemote is in early development: it works day to day, but expect rough edges. Please report what you run into: https://github.com/RFdeGroot/OMARemote/issues
Install on Omarchy, on Intel/AMD (x86_64) or Apple Silicon (aarch64); pacman pulls in every dependency:
curl -LO https://github.com/RFdeGroot/OMARemote/releases/download/v0.9.8/omaremote-0.9.8-1-$(uname -m).pkg.tar.zst
sudo pacman -U ./omaremote-0.9.8-1-$(uname -m).pkg.tar.zstThe package is signed with OMARemote's release key; OMARemote's Update button checks that before installing. To check it by hand, with the key from packaging/arch/omaremote-release.gpg in the repository: curl -L -o omaremote.sig https://github.com/RFdeGroot/OMARemote/releases/download/v0.9.8/omaremote-0.9.8-1-$(uname -m).pkg.tar.zst.sig, then gpgv --keyring ./omaremote-release.gpg omaremote.sig omaremote-0.9.8-1-$(uname -m).pkg.tar.zst. Keep the signature under that other name: pacman checks a .sig it finds beside the package against its own keyring, which does not have OMARemote's key, and refuses the install.
Updating from an earlier release works the same way: pacman upgrades in place. Restart OMARemote afterwards.
Or build it from source with the attached PKGBUILD: makepkg -si. See the README for more.