Repository navigation
feat: import papers (retry after libspelling publish) - #532
Conversation
There was a problem hiding this comment.
HUMAN DECISION NEEDED — @hanthor: this PR builds and would publish the full Papers app RPM, but the maintainer direction recorded on #523 and your own #476 update say the app is Flatpak-delivered and only the native thumbnailer is wanted. Confirm whether publishing papers (app), papers-previewer and papers-nautilus alongside the Flatpak is acceptable, or whether the recipe should ship only the thumbnailer/libs.
Reviewed at head 2275fb6.
correctness
High — the non-RHEL build path needs Rust crate RPMs the factory does not have. packages/papers/papers.spec:1 keeps Fedora's %bcond bundled_rust_deps %{defined rhel}, so on Hummingbird the spec takes the system-crates path: %cargo_prep (:187) and %generate_buildrequires / %cargo_generate_buildrequires -a -t (:194-195). The hermetic lane resolves those generated crate(...) BuildRequires in a real mock root (.github/workflows/build-stage.yml:213-220), and #527 records that the factory carries essentially no Rust crate stack — crate(gio) already fails for rust-libglycin-rebind; papers' vendor set needs gio-0.22, gtk4-0.11, libadwaita-0.9. packages/glycin/glycin.spec:5-8 hit the same gap and forced the vendor path. The vendor tarball is already provisioned: packages/papers/sources:2 lists it, tools/source_pipeline.py:163-190 fetches it from the Fedora lookaside by hash, and I verified the downloaded tarball (sha512 cbe77ffd…) contains formatx-0.4.2, matching 0001-bump-formatx-dependency-from-0.2-to-0.4.patch. The PR body also says local validation has not happened yet, so nothing has exercised this path.
# Hummingbird has no Rust crate RPMs (see #527 and glycin.spec), so the Fedora
# system-crates path cannot resolve. Always use the lookaside vendor tarball.
%bcond bundled_rust_deps 1
intent-alignment
Medium — scope conflicts with the recorded Flatpak decision. #523's flatpak-overlap review states "do not build apps already shipped as Flatpaks — org.gnome.Papers … papers (app): stays OUT", and the #476 follow-up says "papers app RPM dropped permanently … Remaining question is only the native PDF thumbnailer". This PR's %files for the main package (packages/papers/papers.spec:234-235, /usr/bin/papers + org.gnome.Papers.desktop) and the previewer/nautilus subpackages would all be published by the factory's publish job. The backlog move (config/factory-build-backlog.toml:669-672) correctly lists only the four subpackages, never papers itself — the catalog never asked for the app. If the app RPM is unwanted, one route is to keep the spec but exclude the main/previewer/nautilus packages from publication; that is a policy call, not one I can make here.
No findings from: security, style, docs-currency.
Confidence: 2/5 (needs attention) — 1 high finding; 1 medium finding
— hive: agent=reviewer backend=copilot model=claude-fable-5.1 copilot=1.0.88
hanthor
left a comment
There was a problem hiding this comment.
Import mechanics check out and libspelling is published, but the recipe cannot build as written: packages/papers/papers.spec:1 sets %bcond bundled_rust_deps %{defined rhel}, taking the system-crate path on factory roots. Your own Canary run (39188213702) failed with 43 unmet crate() BuildRequires. Apply the glycin fix verbatim (%bcond bundled_rust_deps 1; the vendor tarball is already in sources). See factory #527.
Hummingbird has no Rust crate RPMs, so the Fedora system-crate path cannot resolve (43 unmet crate() BuildRequires in the papers-only lane). Force the bundled vendor tarball, as glycin already does. Assisted-by: omp via GitHub Copilot Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Retry: papers builddep needs libspelling, now published to factory. Recipe re-imported with byte-verified upstream lock. Backlog moves land next with this PR number. Local validation after the moves commit.