Repository navigation
feat: add build recipes for the VM networking and streaming-video libraries - #789
Conversation
Refs #503 Hive-Run: #503 Hive-Plan: issue #503 Implementation step 5 (backlog reconciliation) Hive-Spec: factory-build-backlog#closing-a-backlog-gap Assisted-by: Claude Opus 5.5 via Claude Code Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Signed-off-by: Danathar <Danathar@users.noreply.github.com>
|
@Danathar this PR is back in your court ( Review asks:
Reply here or push a new head to this branch (same sign-off); CI and the review swarm re-run automatically and the PR returns to the normal path. Queued in your relay fix queue (3 waiting): fixes are handed out before new work as soon as you have capacity. |
30fe811 to
37a4842
Compare
Refs #503 Hive-Run: #503 Hive-Plan: issue #503 Implementation step 5 (backlog reconciliation) Hive-Spec: factory-build-backlog#closing-a-backlog-gap Assisted-by: Claude Opus 5.5 via Claude Code Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Signed-off-by: Danathar <Danathar@users.noreply.github.com>
There was a problem hiding this comment.
Looks correct to me, and I reproduced the generator's claims rather than taking them from the comments:
- libslirp: ran the exact sequence in
tools/generated_sources.py:1047-1065by hand (git fetch --depth 1 … refs/tags/v4.9.6→rev-parse FETCH_HEAD^{commit}=392af902…→git archive --prefix=libslirp-4.9.6/→lzmaxz/CRC64/preset 6). SHA-512 is66cf1461af28…, which is both the new lock entry inconfig/upstream-sources.jsonand Fedora's ownsourcesdigest at the imported dist-git commit. The "byte for byte" comment atgenerated_sources.py:986-989holds. - rtmpdump: the
git archivetar of138fdb25…iscmp-identical to Fedora's lookasidertmpdump-2.6.tar.gzonce gunzipped, andgzip -n -9over it gives423d7c94…, the digest the PR pins inpackages/rtmpdump/sources. So the re-pin away from Fedora's gzip digest (7ebe4fe5…) is justified exactly asgenerated_sources.py:989-991describes — same tar, different deflate stream. - Both specs are byte-identical to Fedora dist-git at the commits in
.hummingbird-upstream.json. _git_archive_generatefails closed on a moved tag before writing anything (:1059-1061), andtests/test_generated_sources_generate.py:329-402covers the unpinned-version, moved-tag, gz-reproducibility and metadata paths against a local repo, so the suite does not depend on network.
One thing worth stating for the record, not a defect: both locks now depend on git archive's tar writer and the xz/gzip encoders staying stable across tool versions (I reproduced with git 2.47.3 / gzip 1.13). If a future runner image changes either, the pipeline will fail closed with a digest mismatch rather than ship different bytes, which is the right failure — just expect that to be the symptom.
Confidence: 5/5 (safe)
— hive: agent=reviewer backend=copilot model=claude-fable-5.1 copilot=1.0.88
|
@Danathar this PR is back in your court ( Review asks:
Reply here or push a new head to this branch (same sign-off); CI and the review swarm re-run automatically and the PR returns to the normal path. Your relay could not take this as a |
|
@Danathar this PR is back in your court ( Review asks:
Reply here or push a new head to this branch (same sign-off); CI and the review swarm re-run automatically and the PR returns to the normal path. Queued in your relay fix queue (4 waiting): fixes are handed out before new work as soon as you have capacity. |
|
@Danathar this PR is back in your court ( Review asks:
Reply here or push a new head to this branch (same sign-off); CI and the review swarm re-run automatically and the PR returns to the normal path. Queued in your relay fix queue (5 waiting): fixes are handed out before new work as soon as you have capacity. |
1 similar comment
|
@Danathar this PR is back in your court ( Review asks:
Reply here or push a new head to this branch (same sign-off); CI and the review swarm re-run automatically and the PR returns to the normal path. Queued in your relay fix queue (5 waiting): fixes are handed out before new work as soon as you have capacity. |
|
@Danathar the hive has adopted this PR ( |
hanthor
left a comment
There was a problem hiding this comment.
CI is green. It only needs a rebase onto main and regenerated counts.
|
@Danathar this PR has a new head without confirmed hive push evidence, so the hive has handed this PR back to you ( |
There was a problem hiding this comment.
Read tools/generated_sources.py against the existing generators and the two new recipes; this looks correct to me.
What I verified:
_git_archive_generate(tools/generated_sources.py, new block after_bluefin_artwork_generate) fetches onlyrefs/tags/<tag>, peelsFETCH_HEAD^{commit}, and raises before writing anything if the tag no longer resolves to the pinned commit. The unpinned-version check runs before anysubprocesscall. Both paths are covered intests/test_generated_sources_generate.py::GitArchiveGenerateTests.gzip -n -9strips the mtime, so thertmpdumparchive is reproducible; the test asserts bytes 4–8 are zero and that two runs are identical.config/upstream-sources.jsongenerateentries carry the same digests aspackages/libslirp/sourcesandpackages/rtmpdump/sources, andrtmpdump.spec:1pins the same%global commit 138fdb25…thatGIT_ARCHIVE_SOURCESdoes.libslirp.spec:11hasBuildRequires: git-core, which%autosetup -S git_am(line 31) needs.
One clarification for whoever merges, since the green checks can read as more than they are: the pass1/pass6 jobs on this PR built the Canary's fixed smoke set (libical, vulkan-headers, vulkan-loader) — not libslirp or rtmpdump. The body's "Hosted build: not run yet" is still accurate after CI; the [resolved] entries record the recipe import only, as the body says.
Confidence: 5/5 (safe)
— hive: agent=reviewer backend=copilot model=claude-fable-5.1 copilot=1.0.88
|
@Danathar this PR is back in your court ( Review asks:
Reply here or push a new head to this branch (same sign-off); CI and the review swarm re-run automatically and the PR returns to the normal path. Queued in your relay fix queue (2 waiting): fixes are handed out before new work as soon as you have capacity. |
|
stage 095-pre-merge-soak → 072-returned-to-contributor: conflicting |
|
@Danathar the hive has adopted this PR ( |
…raries Import Fedora Rawhide libslirp 4.9.6 (user-mode networking for virtual machines) and rtmpdump 2.6 (builds librtmp) for long-tail batch #503. Neither project publishes a release file. Add a git-archive generator to tools/generated_sources.py: fetch the upstream tag, require that it still resolves to the pinned commit, and `git archive` it under <name>-<version>/. - libslirp: xz preset 6 over the git archive tar reproduces Fedora's pinned tar.xz byte for byte, so `sources` is unchanged. - rtmpdump: the tar is byte-identical to Fedora's, but the packager's zlib-ng gzip stream is not reproducible, so the archive is `gzip -n -9` and `sources` is repinned to the generated digest. Move libslirp and librtmp to [resolved], regenerate .packit.yaml, the architecture counts and the backlog report, and record the method in the factory-build-backlog skill. Refs #503 Hive-Run: #503 Hive-Plan: issue #503 Implementation steps 2-5 (long-tail batch) Hive-Spec: factory-build-backlog#closing-a-backlog-gap Assisted-by: Claude Opus 5.5 via Claude Code Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Signed-off-by: Danathar <Danathar@users.noreply.github.com>
Refs #503 Hive-Run: #503 Hive-Plan: issue #503 Implementation step 5 (backlog reconciliation) Hive-Spec: factory-build-backlog#closing-a-backlog-gap Assisted-by: Claude Opus 5.5 via Claude Code Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Signed-off-by: Danathar <Danathar@users.noreply.github.com>
Signed-off-by: hanthor <hanthor@users.noreply.github.com>
62d1891 to
81b8a97
Compare
|
@Danathar this PR has a new head without confirmed hive push evidence, so the hive has handed this PR back to you ( |
Summary: Utah is the Bluefin desktop build. It gets its software from a small "factory" in this repository, which rebuilds the packages that Red Hat's minimal base system doesn't provide. This change adds build recipes for two more missing pieces: a library that gives virtual machines network access without special permissions, and a library that streaming-video tools use to receive live streams. Neither project publishes a ready-made download, so the factory now rebuilds their source code from the projects' own version history in a way anyone can repeat and check. This is one more step in a batch that earlier added a disk/network library, a serial file-transfer tool (#742) and two other libraries (#756). The batch issue stays open.
What I need from you: Review it. This branch is in the main repository, so its build check can run.
Refs #503 (partial: 2 of the 5 names still outstanding in this batch; see "Remaining" below)
What changed
libslirplibslirp4.9.6-1 (Fedora Rawhidef92164cc, also builds-devel)[resolved]generateentry: tagv4.9.6ofhttps://gitlab.freedesktop.org/slirp/libslirp.git, pinned to commit392af902…, run throughgit archiveand xz preset 6. The output SHA-51266cf1461…c00be49equals Fedora's dist-git pin, sosourcesis unchanged.librtmprtmpdump2.6-6 (Fedora Rawhide4d382e19, also buildsrtmpdump,librtmp-devel)[resolved]generateentry: tagv2.6ofhttps://git.ffmpeg.org/rtmpdump.git, pinned to commit138fdb25…(the spec's own%global commit). The decompressed tar is byte-identical to Fedora's lookaside tarball. Fedora's gzip stream is zlib-ng output, which neither GNU gzip nor Python's zlib can reproduce, so the archive isgzip -n -9andsourcesis repinned to423d7c94…ff7b9913. Generated twice: same bytes.How it works:
tools/generated_sources.pynow hasGIT_ARCHIVE_SOURCES. For each entry it fetches onlyrefs/tags/<tag>into an empty repository, refuses to continue if the tag no longer resolves to the pinned commit, and writesgit archive --format=tar --prefix=<name>-<version>/. It fetches only the tag becausegit.ffmpeg.orgadvertises refs that point at the all-zero object, and a plaingit clonefails on them (upload-pack: not our ref 0000…).Build dependencies, checked against Hummingbird's current repodata:
nettle-devel4.0 (whichrtmpdump-nettle4.patchneeds),gnutls-devel3.8.13,libgcrypt-devel1.12.4,glib2-devel2.90.1,meson1.12.1,git-core2.56.0. No cross-batch blockers.Also included:
.packit.yaml,docs/architecture.mdandreports/factory-build-backlog.jsonregenerated;libslirpandlibrtmpmoved to[resolved]; new tests intests/test_generated_sources_generate.py(a local tagged repository: xz/gz output, reproducibility, moved-tag refusal, an unpinned version refused before git runs); and the method written up indocs/skills/factory-build-backlog.md.Not yet proven
projectbluefin/utah-packages, so the PR's canary should run, unlike the fork PRs feat: add libsigc++30 Rawhide recipe #545, feat: add build recipes for the liburing and lrzsz system tools #742 and feat: add build recipes for the RabbitMQ client and Windows drawing-file libraries #756. Run URLs and built NEVRAs still need recording.libslirp.so.0andlibrtmp.so.1, and runrtmpdump --help. Utah consumer issue: [Epic] Bluefin package parity: image integration and runtime verification utah#263.[resolved]entry records the recipe import. It does not prove publication or installability.Remaining in #503
lilv-libs: needs the lv2/serd/sord/sratom chain (zixis already a recipe).libreport-filesystem: large source package; not started.lockdev: no upstream serves its MD5-pinned 2011 snapshot. Moving it to[wontfix]needs a linked maintainer decision.Validation
python3 tools/generated_sources.py libslirp|rtmpdump packages/<name> <out>: digests abovepython3 tools/factory_build_backlog.py --check: consistent (227 entries, 213 pending)just check: passjust test: 1395 passed, 2 skippedpre-commitisn't installed here; CI runs it.🤖 Generated with Claude Code
— hive: backend=claude model=claude-opus-5-5
🐝 Hive Agent:
contributor| SHA:fcc4cfce