Shared queue, not private guessing
When multiple hosts hit the same failure, the work can converge in one public issue family instead of being rediscovered from scratch.
Crowd-mode Linux repair, but opt-in and honest
Fixer turns recurring Linux failures into shared repair work. Quiet participants can submit sanitized evidence. Workers with a coding agent can investigate, ship a diff when there really is one, or publish a clean handoff when pretending to have a patch would be dishonest.
When multiple hosts hit the same failure, the work can converge in one public issue family instead of being rediscovered from scratch.
If a patch would be hand-wavy or wrong, Fixer can publish triage, handoff targets, and next steps instead of pretending the diff is ready.
Patches are pushed toward plan-first reasoning, review, and git-friendly writeups so the result is easier to understand and submit upstream.
Merged upstream
The goal is not to generate local workarounds forever. When the evidence is strong and the patch is maintainable, Fixer should help move the fix back to the project that owns the code.
PackageKit
The merged daemon patch makes the progress timer a one-shot source after flushing D-Bus updates, so idle transactions no longer wake every 100 ms.
Merged 2026-05-06 15:09 UTC
htop
The merged patch enumerates existing zram devices instead of probing missing sysfs paths, follows htop's compat/openat style, and reads sysfs files with Compat_readfileat.
Merged 2026-05-01 14:42 UTC
The system tries to stay useful without being creepy, noisy, or overconfident.
The package can live quietly on a machine before anyone decides to join the network. Local mode stays the default.
sudo curl -fsSL https://fixer.maumap.com/apt/fixer-archive-keyring.gpg -o /usr/share/keyrings/fixer-archive-keyring.gpg
echo "deb [arch=$(dpkg --print-architecture) signed-by=/usr/share/keyrings/fixer-archive-keyring.gpg] https://fixer.maumap.com/apt/ stable main" | \
sudo tee /etc/apt/sources.list.d/fixer.list >/dev/null
sudo apt update
sudo apt install fixerAPT repo: https://fixer.maumap.com/apt/
Signing key: https://fixer.maumap.com/apt/fixer-archive-keyring.gpg
Fixer may unintentionally collect private or sensitive data, including local paths, package metadata, command lines, stack traces, debugger backtraces with local variables, warning lines, and other evidence gathered to diagnose issues. Uploading findings to a server is opt-in. Raw coredumps and whole repositories are not uploaded automatically.
Public pages and public issue JSON expose only aggregate sanitized metadata. They do not expose hostnames, install IDs, raw command lines, or private evidence bundles.
The homepage is meant to answer the human question first: what is likely to hurt someone the most right now, and how widely is it happening?
These numbers help explain whether the queue is mostly single-host sightings or already-converging public issue families.
Each issue card now spells out its cluster size directly, for example single-host issue family or 4-host issue family.
This is the work that already made it through quarantine and into the public queue. These cards are sorted by likely user impact first, then by how large the issue family is, then by fixability and freshness.
Desktop graphics failure
Repeated EGL/Mesa/Qt desktop warnings affected plasmashell, org_kde_powerdevil, xdg-desktop-portal-kde, DiscoverNotifier, google-chrome-stable on KDE Wayland desktop, suggesting a shared compositor or graphics-session failure.
Likely user impact. The desktop graphics session likely felt broadly broken: multiple apps may have failed to launch, windows may have stopped painting correctly, or the session may have become unstable.
Cluster size: 2-host issue family. Last seen 2026-05-19 01:26 UTC. Details · JSON
Wake-from-sleep failure
After suspend/resume, radeon X11 desktop failed: Xorg, kwin_x11 crashed after GPU/display errors, and sddm restarted the display stack.
Likely user impact. After wake-from-sleep, the desktop likely came back blank, broken, or dropped the user back to login.
Cluster size: single-host issue family. Last seen 2026-05-19 01:35 UTC. Details · JSON
Wake-from-sleep failure
After suspend/resume, nouveau X11 desktop failed: Xorg crashed after GPU/display errors, and sddm restarted the display stack.
Likely user impact. After wake-from-sleep, the desktop likely came back blank, broken, or dropped the user back to login.
Cluster size: single-host issue family. Last seen 2026-05-19 01:35 UTC. Details · JSON
Desktop graphics failure
Repeated EGL/Mesa/Qt desktop warnings affected chrome, adb, node, systemsettings, sddm-helper-start-wayland on KDE Wayland desktop (nouveau), suggesting a shared compositor or graphics-session failure.
Likely user impact. The desktop graphics session likely felt broadly broken: multiple apps may have failed to launch, windows may have stopped painting correctly, or the session may have become unstable.
Cluster size: single-host issue family. Last seen 2026-05-19 01:22 UTC. Details · JSON
Desktop graphics failure
Repeated EGL/Mesa/Qt desktop warnings affected Xorg on desktop desktop desktop (radeon), suggesting a shared compositor or graphics-session failure.
Likely user impact. The desktop graphics session likely felt broadly broken: multiple apps may have failed to launch, windows may have stopped painting correctly, or the session may have become unstable.
Cluster size: single-host issue family. Last seen 2026-05-19 01:35 UTC. Details · JSON
Out of memory kill
element-desktop was killed by the kernel OOM killer after reaching about 200 MiB anonymous RSS in `element-desktop`.
Likely user impact. The system likely ran out of memory and killed element-desktop, so work in that app or task may have vanished.
Cluster size: 2-host issue family. Last seen 2026-04-03 22:55 UTC. Details · JSON
perl is stuck in a likely busy poll loop: 100.00% of sampled CPU passed through Perl_runops_standard, with repeated pselect6 x4.
Likely user impact. perl likely made the app or machine feel hot, loud, or sluggish by burning CPU continuously.
Cluster size: 3-host issue family. Last seen 2026-05-19 01:26 UTC. Details · JSON
Out of memory kill
python was killed by the kernel OOM killer after reaching about 11717 MiB anonymous RSS in `org.kde.yakuake`.
Likely user impact. The system likely ran out of memory and killed a Python workload, so work in that app or task may have vanished.
Cluster size: single-host issue family. Last seen 2026-03-31 12:57 UTC. Details · JSON