Repository navigation
fix(services): install avahi and enable avahi-daemon - #150
mrbobbytables wants to merge 2 commits into
Conversation
hanthor
left a comment
There was a problem hiding this comment.
This does not build, and the failure is the point of #104 rather than an incidental problem.
The contract job on 20d5161 failed at step 6, Check package availability in Utah's repositories — just check-repos — and every image build was skipped as a result. That step resolves the whole install transaction against the only two runtime repositories Utah enables, so what it is saying is that avahi is not obtainable: Hummingbird ships the client libraries that gvfs, geoclue2, gtk3/4, samba and pipewire build against, but not the daemon package. That is exactly the situation #104 describes, and it is why the comment this PR deletes named avahi as the same shape as ppp and nautilus — a plugin shipped without the daemon behind it.
So the three lines that enable the unit are right, and the one line that installs it cannot work yet. avahi has to exist before Utah can install it:
- Add an
avahirecipe to projectbluefin/utah-packages (there is none today — every hit for "avahi" there is aBuildRequires, satisfied from the buildroot, not from a runtime repository), let it publish, and repin the factory digest. - Then this PR's change to
packages/utah.toml,scripts/configure-services.shand the preset lands as-is andcheck-reposgoes green.
Two smaller notes for when it comes back:
- Please add
avahi-daemon.serviceto[services] enabledincontracts/bluefin-desktop.toml. The in-image verifier asserts that list, and without the entry the enable is only asserted by agrepover the source files — which cannot tell you whether the unit is actually enabled in the built image. (systemd-resolved.serviceandtailscaled.serviceare missing from that list too; worth adding all three in the same commit.) - Keep the avahi sentence in the
[unavailable]comment block until the package genuinely resolves. Deleting the explanation while the gap is still open is how a documented gap turns into a silent one.
Happy to review again as soon as the factory carries it.
Generated by Claude Code
Thank you for the review, @hanthor. Understood completely — holding this PR pending the addition and publication of the
|
castrojo
left a comment
There was a problem hiding this comment.
Tested avahi transaction resolution on branch against enabled repos: fails with 'nothing provides libdaemon.so.0()(64bit) needed by avahi-0.9~rc4-0.1.1.hum1.x86_64 from public-hummingbird-x86_64-rpms'. avahi cannot be installed until libdaemon is built and published to utah-packages (#104). PR is blocked on upstream package build.
hanthor
left a comment
There was a problem hiding this comment.
Requesting changes (re-affirming): contract check fails — avahi package unresolvable upstream. Please resolve the package source issue.
hanthor
left a comment
There was a problem hiding this comment.
This is blocked outside this repo: contract fails on check-repos with nothing provides libdaemon.so.0()(64bit) needed by avahi-0.9~rc4-0.1.1.hum1, and it has been red since 09-18. The PR also deletes the utah.toml comment that documents avahi as an orphaned plugin. Suggest closing it for now and reopening (or folding the three enable lines into #265's batch) once libdaemon ships in utah-packages.
Danathar
left a comment
There was a problem hiding this comment.
This has been idle since 09-18 and now also conflicts with main. The conflict is one comment line in packages/utah.toml. #265 cut "the same shape as avahi (#104) and nautilus (#100)" down to avahi, and this branch cut it down to nautilus. Once this PR installs avahi, neither example fits, so I'd drop that clause.
The libdaemon blocker from hanthor's review may be gone. utah-packages has built libdaemon since 2ef8ed1 (09-20), and its config/runtime-contract.toml now lists avahi as installable. Utah main has pinned the 2026-09-26 factory image since #265. I haven't run check-repos against it, so treat that as a guess. A rebase and a fresh contract run would settle it. If avahi resolves, this is a small PR again. If it still fails on libdaemon, closing it as hanthor suggested makes sense.
Geoclue cannot connect to avahi service and mDNS resolution is degraded because only avahi client libraries ship without the daemon package. Hummingbird provides avahi in its repository. Add avahi to [services] in packages/utah.toml, enable avahi-daemon.service in scripts/configure-services.sh and system_files/shared/usr/lib/systemd/system-preset/85-utah-desktop.preset, and add regression tests. Fixes projectbluefin#104 Assisted-by: Gemini 3.8 Flash via GitHub Copilot Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Signed-off-by: mrbobbytables <mrbobbytables@users.noreply.github.com>
Add avahi-daemon.service, systemd-resolved.service and tailscaled.service to [services] enabled in contracts/bluefin-desktop.toml so the in-image verifier asserts they are enabled, rather than relying on a grep over the source files. avahi now resolves: utah-packages builds libdaemon and main pins the 2026-09-26 factory image, so check-repos installs avahi from public-hummingbird-x86_64-rpms with libdaemon from utah-packages. The Utah additions count moves from 84 to 85; regenerate site data and the documented counts. Assisted-by: Claude via omp Signed-off-by: castrojo <castrojo@users.noreply.github.com>
20d5161 to
710760c
Compare
|
Rebased onto current
Two side effects of the rebase:
Local checks run: |
There was a problem hiding this comment.
Read against the tree at merge-base first, then the body. The avahi change looks correct to me; one thing in the diff is not in the body.
Correct: avahi goes into the [services] group of packages/utah.toml, the daemon is enabled in both scripts/configure-services.sh:76 and the 85-utah-desktop.preset, and contracts/bluefin-desktop.toml gains avahi-daemon.service, which scripts/verify-desktop-contract.py:195-197 will check with systemctl is-enabled in the built image. #104 states the Hummingbird repository carries avahi, so this is a manifest gap and the fix is the right shape. Counts in README/package-contract/packages.json move 84→85 / 141→142 consistently, which scripts/check-doc-counts.py enforces.
Not in the body: contracts/bluefin-desktop.toml also adds systemd-resolved.service and tailscaled.service to [services].enabled. Both are already enabled by scripts/configure-services.sh:73,75, so the contract check will pass and nothing breaks — but the body enumerates every change and omits these two, and they are unrelated to #104. Suggest either a one-line mention in the body (the contract now declares what the image already enables) or dropping them to a separate commit so the #104 fix stays reviewable on its own.
I have not approved or changed anything.
Confidence: 3/5 (needs attention) — 1 perspective requested changes
— hive: agent=reviewer backend=copilot model=claude-fable-5.1 copilot=1.0.88
|
Re-checked at 710760c. Both of my points are addressed. The ppp clause is gone from utah.toml, and the rebase settled the libdaemon question. In run 36282922871 the contract job's check-repos transaction pulls avahi 0.9~rc4-0.1.1.hum1 from Hummingbird and libdaemon 0.14-34.hum1.bfin from utah-packages. build_main installs avahi and enables avahi-daemon.service, and the in-image verifier reports 15 enabled services. I merged the branch onto main 77a2a94 locally: tests/run_suite.py passes (407, 1 skipped), and check-doc-counts and generate-site-data --check are current. One thing to handle before merge: #297 (alsa-utils) makes the same count change on the same lines, 84 to 85 additions and 141 to 142 installed. Because the hunks are identical, git merges the two without a conflict, and the result fails |
|
Re-checked hanthor's change requests against the current head (710760c), so they can be re-reviewed or dismissed. Fixed:
tests/run_suite.py passes on the head (400). One thing left before merge: #297 merged on 09-28 with the same count edits (84 to 85, 141 to 142). Merging current main into this head goes through with no conflict, but |
|
Closing as superseded by #411, which installs Avahi and also enables switcheroo-control. This branch conflicts with Generated by Claude Code |
Geoclue cannot reach Avahi because only the client libraries ship without the daemon package, leaving mDNS resolution and WiFi-based location degraded.
avahito[services]inpackages/utah.tomlavahi-daemon.serviceinscripts/configure-services.shavahi-daemon.serviceinsystem_files/shared/usr/lib/systemd/system-preset/85-utah-desktop.presetdocs/skills/desktop-contract.mdtests/test_package_resolution.pyFixes #104
— hive: backend=copilot model=gemini-3.8-flash
🐝 Hive Agent:
contributor| SHA:20d5161