Repository navigation
feat: import thumbnailers and Disks wave - #523
Conversation
There was a problem hiding this comment.
Reviewed at head c057ed9. All five new recipes are byte-identical to Fedora rawhide at the commits recorded in each .hummingbird-upstream.json, and just check (validate, factory-contract, backlog --check, multimedia, runtime-contract, 1275 tests) passes locally on this head. One gap between what the PR says and what it ships:
intent-alignment
- Imported names are not moved to
[resolved]in the backlog catalog (medium). Issue #513 step 3 ("Move the name from its area to[resolved]... in the same PR") and the repository rule indocs/skills/factory-build-backlog.md:73-77("already_recipeis the auditor's observation of an area name whose move is still owed, not an alternative way to close it") both require the two-edit move. This diff does not touchconfig/factory-build-backlog.toml:papers-libs/-nautilus/-previewer/-thumbnailerremain atconfig/factory-build-backlog.toml:79-82,gnome-disk-utilityandgnome-epub-thumbnailerat:180-181,gst-thumbnailersat:468, andreports/factory-build-backlog.jsononly flips them toalready_recipe. The PR body says the[resolved]entries "lackprnumbers", but no such entries exist in the diff. The previous wave did this correctly (config/factory-build-backlog.toml:651-655, PR #514). Fix: remove the seven names from their areas, add{ name = ..., pr = 523 }under[resolved], rerunpython3 tools/factory_build_backlog.pyto regenerate the report.
correctness — open question, not a verified defect: the new recipes carry runtime Requires that are neither built in packages/ nor listed in config/hummingbird-provided-sources.json — libhandy (packages/gnome-disk-utility/gnome-disk-utility.spec:34), libpeas-loader-python3, grilo-plugins, gstreamer1-plugins-good-gtk (packages/totem/totem.spec:60,67,73). Only totem-video-thumbnailer is contract-listed (config/bluefin-packages.toml:131), so the final publish transaction is not at risk, but a consumer=built canary resolves every binary in the candidate (.github/workflows/publish-repository.yml:561-565). I could not check what the Hummingbird image provides; worth confirming when the canary is dispatched.
No findings from: security, style, docs-currency.
Confidence: 3/5 (needs attention) — 1 medium finding; 1 perspective requested changes
— hive: agent=reviewer backend=copilot model=claude-fable-5.1 copilot=1.0.88
There was a problem hiding this comment.
Reviewed at head 1b3ebb3. Five recipes imported byte-identical to the pinned Fedora rawhide commits in each .hummingbird-upstream.json (spec and sources compared against src.fedoraproject.org); just check, pytest (1275 passed) and tools/validate.py ("validated 431 source RPMs (427 rawhide, 4 upstream)") all pass locally, matching the counts written into docs/architecture.md:81,133-144,163. The papers vendor tarball (packages/papers/sources:2) resolves from the lookaside (HTTP 200) via tools/source_pipeline.py:162-190.
intent-alignment
- info:
#513asks for runtime deps built in graph order.packages/gnome-disk-utility/gnome-disk-utility.spec:22,28BuildRequireslibhandy-1andpwquality, andpackages/totem/totem.spec:30-31,54-55requirelibpeas-gtk-1.0/grilo-0.3; none of these has a recipe under.packit.yaml. Buildability therefore depends on the Hummingbird buildroot shipping them — not something I could verify here; the canary will settle it.
No findings from: correctness, security, style, docs-currency.
Confidence: 5/5 (safe)
— hive: agent=reviewer backend=copilot model=claude-fable-5.1 copilot=1.0.88
…r, gst-thumbnailers, totem, gnome-disk-utility) papers-thumbnailer is a subpackage of source papers; totem provides the contract-listed totem-video-thumbnailer; gnome-disk-utility covers factory #513 scope. Assisted-by: muse-spark via GitHub Copilot Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
…ler builds papers needs libspelling (ICU-77-free factory build); gst-thumbnailers needs crate(libglycin-rebind). Locks verified byte-identical to Fedora pins.
1b3ebb3 to
069eedf
Compare
There was a problem hiding this comment.
intent-alignment
The PR body says the seven imported names are moved to [resolved] with pr = 523, but config/factory-build-backlog.toml is not in this diff. At head 069eedf the names are still listed under their areas (config/factory-build-backlog.toml:76-79 papers-libs/nautilus/previewer/thumbnailer, :171-172 gnome-disk-utility/gnome-epub-thumbnailer, :192 libspelling, :449 gst-thumbnailers) and [resolved] at :602 has no 523 entries. The regenerated reports/factory-build-backlog.json therefore records them as already_recipe (31→39), which docs/skills/factory-build-backlog.md:73-77 defines as "an area name whose move is still owed", and :55-58 says an import PR makes the two-edit move (remove from area, add to [resolved]) in the same PR. Please add the eight { name = ..., pr = 523 } entries and regenerate the report, or drop the claim from the body.
Minor: libspelling and rust-libglycin-rebind are imported (as BuildRequires of papers at packages/papers/papers.spec:60 and gst-thumbnailers via 0003-Bump-libglycin-rebind-*.patch) but are not named in the body; worth a line so the scope is explicit.
No findings from: correctness, security, style, docs-currency.
Confidence: 3/5 (needs attention) — 1 medium finding; 1 perspective requested changes
— hive: agent=reviewer backend=copilot model=claude-fable-5.1 copilot=1.0.88
Move papers-libs, papers-nautilus, papers-previewer, papers-thumbnailer, gnome-disk-utility, gnome-epub-thumbnailer, gst-thumbnailers and libspelling from their backlog areas into [resolved] with pr = 523, and regenerate reports/factory-build-backlog.json. Hive-Fix-Attempt: 2/6 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
There was a problem hiding this comment.
Reviewed at head 52740cb. This PR adds seven new package recipes (document viewer Papers, three thumbnailers, the Disks app, and two build-only libraries) imported from Fedora with pinned source checksums; the diff matches what the description says and all repository gates pass. Two things worth knowing before merge, neither blocking.
correctness
- (low) First-run build order inside this wave is not visible to the planner.
tools/build_graph.py:100-125builds edges fromrpmspec --buildrequiresplus capabilities the published factory already provides. Two edges introduced here are neither:packages/papers/papers.spec:110needspkgconfig(libspelling-1), a generated provide of the not-yet-publishedlibspelling;packages/gst-thumbnailers/gst-thumbnailers.spec:55-56pullscrate(libglycin-rebind)only via%cargo_generate_buildrequires, whichrpmspec -qnever sees. So papers/gst-thumbnailers may be scheduled alongside or before their in-wave providers and satisfy those BuildRequires from Fedora's buildroot on the first run (works, but the first published binaries are not built against the factory's own copies). Self-corrects on the next run once primary.xml carries the provides — just don't read a first-canary ordering surprise as a recipe bug.
intent-alignment
- (info) The body says this "covers #513 scope", and the recipe +
[resolved]move match #513's implementation steps (config/factory-build-backlog.toml:657-665). But #513's acceptance items (factory run URLs, binary NEVRAs, Hummingbird-only runtime resolution) are deferred to the canary per the PR body, and the PR has no closing keyword, so merging will not close #513 — someone needs to attach the canary evidence there afterwards.
No findings from: security, style, docs-currency.
Confidence: 5/5 (safe)
— hive: agent=reviewer backend=copilot model=claude-fable-5.1 copilot=1.0.88
…ollow-ups papers needs published libspelling (ICU-77 builddep); gst-thumbnailers + rust crate need gtk-rs crate stack (#527). Keep lane-green epub, totem, gdu, libspelling.
52740cb to
b11e4d3
Compare
There was a problem hiding this comment.
Reviewed at head b11e4d3. The four recipes that are actually in the diff (gnome-epub-thumbnailer, totem, gnome-disk-utility, libspelling) validate cleanly (tools/validate.py → 448 recipes; tools/factory_build_backlog.py --check consistent; backlog/validate tests pass). The findings are about what the PR says versus what it does.
intent-alignment
- PR body is stale after the deferral commit. The body still says this wave imports
papers,gst-thumbnailersandrust-libglycin-rebind, and that eight backlog names move to[resolved]withpr = 523. Head commit b11e4d3 ("defer papers, gst-thumbnailers, rust-libglycin-rebind to follow-ups") deletedpackages/papers/,packages/gst-thumbnailers/andpackages/rust-libglycin-rebind/; none exist in the tree, andreports/factory-build-backlog.json:234-250(papers-*) and:1534(gst-thumbnailers) remain"state": "pending". The body's stated reason forlibspelling("BuildRequires of papers") no longer applies inside this PR either — it now lands as a standalone pre-stage. Please rewrite the body to describe the four-package wave and link the follow-ups (#527 is named in the commit).
docs-currency
config/factory-build-backlog.tomlwas not updated even though the body says it was (git diff 336dd44f..b11e4d3 -- config/factory-build-backlog.tomlis empty). Every previous wave closed its names under[resolved]— #466, #514, #522 atconfig/factory-build-backlog.toml:606-640, #524 at:646-648— and the catalog header (:24-26) describes that as the two-edit move.gnome-disk-utility(:171),gnome-epub-thumbnailer(:172) andlibspelling(:192) are still listed in[areas.gnome-apps]; the auditor only infersalready_recipefrompackages/. Add{ name = ..., pr = 523 }entries for the three and regenerate the report so the catalog matches the other waves.
No findings from: correctness, security, style.
Confidence: 3/5 (needs attention) — 1 medium finding; 2 perspectives requested changes
— hive: agent=reviewer backend=copilot model=claude-fable-5.1 copilot=1.0.88
…ibspelling via #523 Move the three catalogued names this wave imports from [areas.gnome-apps] to [resolved] with pr = 523 and regenerate reports/factory-build-backlog.json. Hive-Fix-Attempt: 3/6 Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
There was a problem hiding this comment.
Reviewed at head 6af6dd0. Four recipes (GNOME Disks, EPUB thumbnailer, Totem video thumbnailer, libspelling) are byte-identical imports of Fedora rawhide at the commits recorded in each .hummingbird-upstream.json; the source lock, Packit registration, backlog catalog and architecture counts all agree, and the repo's own gates pass locally (tools/validate.py → 448 recipes, factory_build_backlog.py --check, factory_contract.py, runtime_contract.py --check, 1281 pytest).
intent-alignment
- (low) Body says
gnome-disk-utility"covers #513 scope", but the body has no closing keyword, so merging leaves #513 open with nothing pointing back at this PR. #513 also asks for runtime deps "in graph order" with blockers linked:libhandyis a BuildRequires and a hardRequiresof Disks (packages/gnome-disk-utility/gnome-disk-utility.spec:22,34) and a BuildRequires of Totem (packages/totem/totem.spec:31), yet it stays pending in[areas.gnome-apps](config/factory-build-backlog.toml:188). SuggestCloses #513(or an explicit "partial" note) and a line in #513 naminglibhandyas the remaining runtime blocker.
correctness
- (info) The PR canary run 39133894024 only exercises the fixed libical/vulkan set; none of the four new recipes were built. This matches how #522/#524/#526 landed ("canary at its turn"), so not a blocker — just noting the build itself is unverified at merge.
No findings from: security, style, docs-currency.
Confidence: 5/5 (safe)
— hive: agent=reviewer backend=copilot model=claude-fable-5.1 copilot=1.0.88
|
Flatpak-overlap review (maintainer direction: do not build apps already shipped as Flatpaks —
|
Wave import:
gnome-epub-thumbnailer,totem(contract-listed video-thumbnailer),gnome-disk-utility(covers #513 scope), andlibspelling.Four packages land in this PR:
gnome-disk-utility— GNOME Disks.gnome-epub-thumbnailer— EPUB/MOBI thumbnailer.totem— provides the contract-listed video thumbnailer.libspelling— imported as a standalone recipe (it is a catalogued gnome-apps backlog name). It was originally pulled in as a BuildRequires of papers; papers is now deferred, so within this PR it stands on its own.Deferred to follow-ups (#527):
papers(thumbnailer subpackage),gst-thumbnailers, and its build dependencyrust-libglycin-rebind. These were removed from this PR in b11e4d3; their backlog entries (papers-*,gst-thumbnailers) remain pending.Local
just check+ tests green. Canary to be dispatched after earlier waves clear (one at a time).Backlog: the three catalogued names this wave closes —
gnome-disk-utility,gnome-epub-thumbnailer,libspelling— are moved from[areas.gnome-apps]to[resolved]withpr = 523inconfig/factory-build-backlog.toml, andreports/factory-build-backlog.jsonis regenerated.totemis not in the backlog catalog, so no move applies.