0.1.6-beta

Prerelease for Windows 10/11 x64 and Linux (Ubuntu 24.04+, Debian 13+, amd64).

• own-backup-install.exe — Windows setup wizard in English and Russian: installs the program for all users and starts its backup service; needs administrator rights. It installs over an earlier version, keeping settings and jobs.
• own-backup_0.1.6.beta-1_amd64.deb — Linux package: sudo apt install ./own-backup_0.1.6.beta-1_amd64.deb installs the program, starts its service and adds you to the own-backup group (sign in again to open the window).
• OwnBackup-Linux-0.1.6.beta-1_amd64.iso — Linux boot medium (Debian 13 live): boots from a CD or a USB stick, UEFI with Secure Boot or BIOS, to back up or restore the computer's disks, Windows ones included.

Added

• Linux version. A .deb package for Ubuntu 24.04+ and Debian 13+ (amd64) with the same service, window and command line as on Windows. It backs up volumes and whole disks with ext4, XFS, btrfs, NTFS, FAT32 and exFAT, including LVM (every logical volume of a physical volume in one snapshot) and open LUKS containers (read through a device-mapper snapshot, the container header kept in the copy). Full, incremental and differential chains in VHDX, qcow2 and encrypted OBK; exclusions that really remove the files on ext4, XFS, btrfs and FAT, and on NTFS through ntfs-3g. The service key store uses systemd-creds (with the TPM 2.0 when there is one). A volume or a whole disk is restored through the service, with the choice to give the restored file systems, LUKS containers and volume groups new identities (UUIDs, labels, initramfs rebuilt) when the originals are still on the machine. The Boot page edits UEFI boot entries; the window has a tray icon and starts with the session. The service runs under systemd, asks polkit before destructive actions and logs to journald and a file.
• Backups are interchangeable between Windows and Linux, VHDX chains included: a copy made on one system opens, validates and restores on the other.
• Linux boot medium. A Debian 13 live ISO with the program on a labwc desktop with our own taskbar; it boots with UEFI (Secure Boot through Debian's signed shim) and BIOS, from a CD or a USB stick, and restores Windows copies as well as Linux ones. The window builds it from the installed package or downloads the ready one of the same version from the release, and writes it onto a USB stick.
• The boot medium card lets you choose a Winre.wim by hand (for example one taken from a Windows installation image). The service reads it first and refuses an installation image, a split or non-x64 image with the reason; a build given such a file is refused the same way.

Fixed

• An unfinished VHDX or qcow2 image of a run whose service was killed is now deleted when the service starts again (or at the next run, if the backup disk was not connected). The service records each such file in its protected folder before writing it, so only files it has proved its own are deleted. Leftovers of earlier versions and of runs outside the service stay where they are.
• The FAT boot image inside the medium's ISO keeps long names (EFI\Microsoft, boot.stl) instead of cutting them to 8.3.
• The empty part of the compression slider follows the theme.
• A job or history save that fails (for example on a full disk) no longer leaves a half-written temporary file holding the space.
• The window and the service exchange messages up to 64 MB: a job list past 4 MB (very long exclusion lists) no longer breaks every request.
• A source volume keeps its chain when it is named differently: by a new drive letter, its volume GUID path or a mount point. It used to start a new full copy, because the chain was matched by the name in the file name.
• A mount point (C:\Mount\Data) works as a source: it is read as the volume mounted there, its images are named by that volume, and a destination elsewhere on C: is no longer refused as "the same volume".
• A ReFS volume inside BitLocker is copied with its used-block map and its exclusions applied, locked or unlocked. It used to fail: under BitLocker ReFS ends before the device does, the cluster size could not be derived, and the sector-by-sector fallback hit the unreadable tail.
• A job remembers the folders it wrote into before its destination was changed, and deleting the job deletes its copies there too — even after the activity journal (200 runs) has forgotten that folder.
• A chain never grows past 60 images: the next run starts a new full copy whatever the scheme. A job of "only increments" without consolidation (saved before the modes existed, or made outside the window) used to grow past what the VHDX reader follows (64), after which every new copy was unreadable and retention stopped.
• XFS: every excluded file is now truly removed (1403 of 1403 on the test corpus, was 1369). A directory data block that would be left empty is taken out of the directory and returned to free space, as the kernel does, and is zeroed so the deleted names do not stay behind.
• The service restarts by itself after a crash (in 10 s, then 1 and 5 min). It used to stay stopped until the next reboot, and scheduled copies with it.
• A whole-disk restore keeps the partitions that are not in the backup, on Windows and on Linux. It used to rewrite the partition table from the backup alone.
• Restoring a disk keeps the boot code in LBA 0 (MBR) when the partition table is rewritten; on Linux the sectors after it (GRUB's core.img) stay too.
• NTFS exclusions cover named streams (Zone.Identifier and others) and content stored inside the MFT record. A file with many names or streams, whose record continues in extension records, no longer drops out of the exclusions entirely.
• qcow2: after consolidation the old bytes of replaced clusters are wiped, so excluded content does not survive in dead clusters.
• A backup interrupted because the service was stopped or killed is recorded as interrupted, and the schedule runs it again.
• The service no longer does one thing at a time for the whole machine: a restore is still exclusive, but copies, checks and mounts of different backup folders and the boot medium build run side by side; a scheduled check gives way to an action of a person on the same folder.
• A scheduled whole-disk backup with an encrypted partition needs the saved password only for that partition (ESP and /boot next to LUKS used to make every run skip), and a skipped run is shown under Activity.
• The end of a backup no longer waits forever behind a hung device: each VSS step has a time limit, and the window says "releasing the volume snapshot" or "opening the finished backup" instead of "flushing the cache".
• The result says how many excluded files an ext4, XFS or btrfs volume kept under their names with their contents zeroed.
• Validating an image under Tools keeps its progress bar and result when you leave the page and come back.
• On Windows the service writes the settings and the service data folder is open only to SYSTEM and administrators.
• Error messages no longer start with an English prefix.
• Dependencies: tokio past an I/O-safety flaw (RUSTSEC-2026-0386), Vite past advisories of its development server, and zbus 6 instead of the unmaintained zvariant crates; the Linux taskbar calls libgtk-layer-shell directly instead of its archived Rust bindings.

Windows: settings and jobs are kept in C:\ProgramData\OwnBackup, the service log in the Logs folder next to the program. The window needs the Microsoft Edge WebView2 Runtime: Windows 11 has it, and the installer says so if a Windows 10 does not. Linux: settings and jobs in /var/lib/own-backup, the log in journald (journalctl -u own-backup).

License: PolyForm Noncommercial 1.0.0; third-party components keep their own licenses. The checksums of the files are in SHA256SUMS.txt.