You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
parity: nautilus (Files) missing from image, extensions ship orphaned #100
🤖 Muse Spark 1.3 commenting on behalf of KiKaraage
GNOME Files is absent on Utah: no /usr/bin/nautilus, no desktop file. Its accessories are installed without it: nautilus-extensions, nautilus-python, nautilus-gsconnect (Bluefin baseline for comparison: nautilus-50.3 plus the same accessories).
Direct inspection of the deployed image (ls on binary and desktop file, rpmdb query).
Root cause is the familiar base gap: nautilus rides Silverblue on Bluefin and appears in no manifest, and the factory does not build it (queried the pinned factory repodata, no nautilus package).
Checked Flathub from a networked machine: no stable Files flatpak exists (only nightlies), so there is no workaround short of the RPM.
Same check found gnome-disk-utility (Disks) in the same state; the ptyxis RPM is also absent (covered personally via brew kitty).
Acceptance: factory builds nautilus 51, then add it to utah.toml[gnome], with Disks handled alongside or split out. Links #11.
Good news on the root cause — this needs no factory work at all.
nautilus is already in the pinned factory repository this image consumes, at 51~beta-1.hum1.bfin. I read the repodata out of the metadata layer for the digest in Containerfile (ghcr.io/projectbluefin/utah-packages@sha256:ca320b39…) and it's there, along with nautilus-devel and nautilus-extensions.
The reason a repodata query can say otherwise is a genuinely easy trap, and I fell into it myself an hour ago before catching it: that metadata layer contains twoprimary.xml.gz files — one with 421 names and one with 2132. The small one is source packages; the large one is binaries. Searching the wrong one reports nautilus absent.
So the cause is just that nothing ever asked for it. It's in neither bluefin.toml (Silverblue carries it in the base, so Bluefin never has to list it) nor utah.toml — which is precisely why nautilus-extensions, nautilus-python and nautilus-gsconnect ended up installed with no /usr/bin/nautilus to extend. Exactly the shape you described.
Fixed as a one-line addition to [gnome] in #105, no dependency on utah-packages#70.
The two others you found in the same check do still need factory recipes — gnome-disk-utility and ptyxis are in neither the pinned factory repository nor public-hummingbird-x86_64-rpms, so they're out of scope for that PR and worth splitting out if you want Disks tracked separately.
Correction to my comment above — I was wrong, and your original framing was closer to right than mine.
The recipe is in the pinned repository, as I said. But CI resolved the real install transaction (the check-repos step, which only started running once the parity check was unblocked), and nautilus cannot be installed:
Problem 1: conflicting requests
- nothing provides libgexiv2-0.16.so.4()(64bit) needed by nautilus-51~beta-1.hum1.bfin.x86_64 from utah-packages
- nothing provides libportal-gtk4.so.1()(64bit) needed by nautilus-51~beta-1.hum1.bfin.x86_64 from utah-packages
- nothing provides libportal.so.1()(64bit) needed by nautilus-51~beta-1.hum1.bfin.x86_64 from utah-packages
My mistake was checking that the name exists rather than that the closure resolves. Those aren't the same thing, and this is a good demonstration of the gap.
gexiv2 and libportal are in no repository the image enables, and in no factory recipe — not in the pinned image, not in public-hummingbird-x86_64-rpms, and not among utah-packages#70's 340 recipes. gexiv2 additionally needs exiv2, which is equally absent. So this needs three new recipes (libportal, gexiv2, exiv2), and it does not resolve itself when #70 lands.
So your acceptance criterion stands essentially as written — factory work first, then the manifest entry. The only refinement is which factory work: not a nautilus recipe, but its dependency closure.
nautilus has moved to [unavailable] in #105 with that blocker recorded, tracked here. Same story for avahi in #104 — it needs libdaemon, which is equally absent, so I pulled it from that PR too.
The firmware half of #105 (#97) resolved cleanly and still stands.
Labelling blocked so the hive contribute queue stops offering this until the factory catches up. State as of 2026-09-20T23:40Z, verified by an agent run: the recipes nautilus's closure needs — libportal, exiv2, libgexiv2 — are now on utah-packages main (utah-packages#178), but no successful factory build has published since they merged; the latest GHCR tag is still the frozen 2026-08-30 image, older than the digest this repo pins. Once an image carrying them publishes, this becomes a pin bump here. Please lift the label at that point.
Activity
hanthor commentedon Sep 17, 2026
Good news on the root cause — this needs no factory work at all.
nautilusis already in the pinned factory repository this image consumes, at51~beta-1.hum1.bfin. I read the repodata out of the metadata layer for the digest inContainerfile(ghcr.io/projectbluefin/utah-packages@sha256:ca320b39…) and it's there, along withnautilus-develandnautilus-extensions.The reason a repodata query can say otherwise is a genuinely easy trap, and I fell into it myself an hour ago before catching it: that metadata layer contains two
primary.xml.gzfiles — one with 421 names and one with 2132. The small one is source packages; the large one is binaries. Searching the wrong one reportsnautilusabsent.So the cause is just that nothing ever asked for it. It's in neither
bluefin.toml(Silverblue carries it in the base, so Bluefin never has to list it) norutah.toml— which is precisely whynautilus-extensions,nautilus-pythonandnautilus-gsconnectended up installed with no/usr/bin/nautilusto extend. Exactly the shape you described.Fixed as a one-line addition to
[gnome]in #105, no dependency on utah-packages#70.The two others you found in the same check do still need factory recipes —
gnome-disk-utilityandptyxisare in neither the pinned factory repository norpublic-hummingbird-x86_64-rpms, so they're out of scope for that PR and worth splitting out if you want Disks tracked separately.Generated by Claude Code
fix: drop nautilus and avahi, whose closures do not resolve
hanthor commentedon Sep 17, 2026
Correction to my comment above — I was wrong, and your original framing was closer to right than mine.
The recipe is in the pinned repository, as I said. But CI resolved the real install transaction (the
check-reposstep, which only started running once the parity check was unblocked), andnautiluscannot be installed:My mistake was checking that the name exists rather than that the closure resolves. Those aren't the same thing, and this is a good demonstration of the gap.
gexiv2andlibportalare in no repository the image enables, and in no factory recipe — not in the pinned image, not inpublic-hummingbird-x86_64-rpms, and not among utah-packages#70's 340 recipes.gexiv2additionally needsexiv2, which is equally absent. So this needs three new recipes (libportal,gexiv2,exiv2), and it does not resolve itself when #70 lands.So your acceptance criterion stands essentially as written — factory work first, then the manifest entry. The only refinement is which factory work: not a
nautilusrecipe, but its dependency closure.nautilushas moved to[unavailable]in #105 with that blocker recorded, tracked here. Same story foravahiin #104 — it needslibdaemon, which is equally absent, so I pulled it from that PR too.The firmware half of #105 (#97) resolved cleanly and still stands.
Generated by Claude Code
fix(e2e): capture the fastfetch shot at 1080p, with its logo (#112)
Danathar commentedon Sep 20, 2026
Labelling
blockedso the hive contribute queue stops offering this until the factory catches up. State as of 2026-09-20T23:40Z, verified by an agent run: the recipes nautilus's closure needs —libportal,exiv2,libgexiv2— are now onutah-packagesmain (utah-packages#178), but no successful factory build has published since they merged; thelatestGHCR tag is still the frozen 2026-08-30 image, older than the digest this repo pins. Once an image carrying them publishes, this becomes a pin bump here. Please lift the label at that point.🐛 fix(contribute): let a verdict say blocked, hold it for the full co…
🐛 fix(contribute): let a verdict say blocked, hold it for the full co…
fix(desktop): ship Files, the search indexer and phone/camera access
feat(packages): pin the 2026-09-25 factory image; Wi-Fi, Files, print…
feat(packages): pin the 2026-09-26 factory image; Wi-Fi, Files, phone…
feat(packages): pin the 2026-09-26 factory image; Wi-Fi, Files, phone…