os-package-availability: fedora package-list drift

[github-actions]

os-package-availability: fedora package-list drift — 2026-07-18

I have enough to report. The dust/xh/doggo/carapace lookups returned 404 (no positive hit → they remain unpackaged, so the exclusion comment stays correct; no new-availability drift). Here's the audit.

---

/os-package-availability — Fedora

Question: does every name in dotfiles-Fedora/install/packages.txt still resolve upstream, under the name used, and still belong here?

Verdict: ✅ Clean. No broken names, no renames, no stale exclusions. Every active entry resolves in the Fedora repos; every documented exclusion is still genuinely unpackaged.

Index queried: Fedora Packages (packages.fedoraproject.org) + mdapi (mdapi.fedoraproject.org, rawhide = F45). Note: PORTING-MATRIX.md has no Fedora column by design (Fedora is the template), so names were verified against the upstream index + matrix footnotes, not a matrix column.

Broken (fix needed)

None.

Drifted (likely)

None. Two things that looked like drift but aren't:

• sd (packages.txt:24) — mdapi returns HTTP 400 for the 2-char name and the /pkgs/rust-sd/sd/ table under-renders (only shows F41). This is a tooling quirk, not a drop: dnf install sd resolves on F44 (confirmed via search + the live package page). Installable as written.
• gron exclusion (packages.txt:28–31) — comment claims it was orphaned and dropped (last F41/42). Confirmed accurate: the Fedora page lists gron only in F41/F42 (0.7.1-10.fc41, maintainer orphan), absent from F43+. Correctly kept out with a truthful comment.

Clean — checked and still resolves

37 entries in packages.txt. The 14 drift-prone / modern-CLI names were each confirmed live against the Fedora index:

Package	Fedora status
eza	✓ 0.23.4 (rawhide)
fd-find	✓ 10.4.2 (binary fd, per comment)
git-delta	✓ 0.19.1
tealdeer	✓ 1.7.3 (F43/F44; binary tldr)
duf	✓ 0.9.1
procs	✓ 0.14.10 (F43/F44)
sd	✓ 1.0.0 (see note above)
yq	✓ 4.53.3 — confirmed the Go/mikefarah build (python-yq is separate), matching the packages.txt:25 comment
glow	✓ 2.1.2
gum	✓ 0.17.0
git-subtree	✓ 2.55.0 (separate package, per comment)
python3-neovim	✓ 0.6.0 (real pynvim provider)
direnv	✓ 2.37.1
tree-sitter-cli	✓ 0.26.9 — clears the ≥0.26.1 floor from matrix footnote ⁵; the "drop to mise/cargo" fallback comment is still just a fallback, not needed

The remaining ~22 are long-stable Fedora base packages accepted as resolving without individual query: zsh, tmux, ripgrep, zoxide, fzf, btop, bat, neovim, git, openssh-clients, wl-clipboard, xclip, gcc, gcc-c++, make, cargo, python3-pip, curl, wget, jq, htop, flatpak, gawk. The wget comment (packages.txt:61–64) — F42+ ships wget as a virtual capability defaulting to wget2-wget — remains accurate.

New availability

None to move in. The bootstrap-only exclusions (dust, xh, doggo, carapace, sesh, gron, op) showed no positive Fedora-index hit, so they remain genuinely unpackaged — the packages.txt:28 comment stands.

Matrix footnotes (Fedora-relevant)

• Footnote ⁵ (tree-sitter-cli via dnf, verify ≥0.26.1): still accurate — F45 ships 0.26.9.

---

Report-only, per the routine — no files edited. If you want, any package-list change would land in caller/install/packages.txt (the OS repo); nothing here needs a PORTING-MATRIX.md edit since Fedora has no matrix column and footnote ⁵ is current.

Sources: Fedora Packages, mdapi, sd on Fedora Packages

Filed by the claude-routines workflow. Report-first: review and act — nothing was changed.