Repository navigation
fix: fetch two Python packages from their publisher - #909
Conversation
Move python-argcomplete and python-dasbus to their verified PyPI source archives. Keep the existing SHA-512 locks and Fedora URLs as fallbacks. Hive-Run: #42 Hive-Plan: supply-chain source-lock migration batch Hive-Spec: issue-42#upstream-release-provenance Assisted-by: Auto mode via GitHub Copilot Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Signed-off-by: mrbobbytables <mrbobbytables@users.noreply.github.com> Signed-off-by: James Reilly <jreilly1821@gmail.com>
Document the maintainer-run proof required for fork source-lock changes and why local checks or another fork run cannot replace approval. Assisted-by: Auto mode via GitHub Copilot Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Signed-off-by: mrbobbytables <mrbobbytables@users.noreply.github.com> Signed-off-by: James Reilly <jreilly1821@gmail.com>
…cannot survive A fork pull request whose fork Canary proof is discarded by every merge from main (or that changes .github/) cannot stay green as a fork PR. Record the maintainer path used for #812, #877 and #862: cherry-pick the PR's own commits onto an in-repo hive/adopt-<N> branch with sign-off and open a new PR. Assisted-by: Claude Opus via omp Signed-off-by: James Reilly <jreilly1821@gmail.com>
There was a problem hiding this comment.
Read the diff against main and the validator it has to satisfy. Looks correct to me.
What I verified:
- Streamed both new primary URLs and hashed them:
argcomplete-3.6.3.tar.gz→d07756a8…fa43a0a,dasbus-1.7.tar.gz→ccf575d2…3d8a873. Both match the unchangedsha512locks inconfig/upstream-sources.json, so this is a transport-only move as the body says. fallback_urlsis the established key (443 entries already use it;tools/source_pipeline.pyconsults it after an upstream transport failure), and the Fedora lookaside URL is preserved there verbatim.tools/validate.py:162check_fedora_primary_sourcesis a shrink-only ratchet: it fails if a name stays inconfig/fedora-primary-sources.txtafter its primary URL leaves Fedora. Removing the two lines is therefore required by the lock change, not optional cleanup..github/workflows/canary-fork-pr.ymlexists at this head, so the newsupply-chain-provenance.mdlink resolves.
Refs #42 is the right citation: the rest of the Fedora-primary list remains and the body says so.
Confidence: 5/5 (safe)
— hive: agent=reviewer backend=copilot model=claude-fable-5.1 copilot=1.0.88
There was a problem hiding this comment.
Read the two lock entries against the tree and the pipeline that consumes them. Looks correct to me.
Verified
- Streamed both PyPI archives and hashed them:
argcomplete-3.6.3.tar.gz→d07756a8…fa43a0a,dasbus-1.7.tar.gz→ccf575d2…3d8a873. Both match the unchangedsha512inconfig/upstream-sources.json(lines 5069/5094 as patched), so this is a transport-only move. fallback_urlsis already a first-class key:tools/source_pipeline.py:337-351validates it and feedsfetch_with_fallbacks([url, *fallback_urls], …); 441 existing entries use it. Keeping the Fedora lookaside URL there is the established shape, not a new mechanism.- Removing the two names from
config/fedora-primary-sources.txtis consistent with the primary URL no longer being a Fedora host (the list is whattools/canary.py:83and the validator key off). - Checks on
c238719: 70 pass, 99 skipped, 0 failing; mergeable CLEAN.
Docs additions (supply-chain-provenance.md, lane-validation.md) describe the procedure actually used here; Refs #42 is right since the list still has entries.
Nothing blocking from my side.
Confidence: 5/5 (safe)
— hive: agent=reviewer backend=copilot model=claude-fable-5.1 copilot=1.0.88
|
Merged by hive contributor-lane auto-merge: CI green for ≥30m, reviewer verdict approve, @hanthor was auto-promoted after 5 clean merged PRs in this hive's repositories. |
Summary: Two Python libraries now download their build files from the Python package publisher instead of Fedora. Their existing file checks are unchanged, and Fedora remains available as a backup.
What I need from you: Review it. This branch lives in the repository, so its own Canary supplies the proof.
Re-homed from #862 by a maintainer: the fork could not be updated by the hive (fork guard / workflow permission), and every merge from
maindiscarded the fork Canary proof. Original author: @mrbobbytables.Changes
python-argcompleteandpython-dasbusto their PyPI source archives.config/fedora-primary-sources.txt.canary-fork-pr.yml.Re-home notes
ec6bae17andd9009db2. Both keep their authorship and now also carry a maintainer sign-off. The 19Merge branch 'main'commits are dropped.main(d5dd441a) without conflicts. The resulting diff matches the original PR's diff against its merge base line for line.docs/skills/lane-validation.md. It records when and how to re-home a fork PR whose fork Canary proof cannot survive (feat: prepare better support for VMware virtual machines #812 → feat: prepare better support for VMware virtual machines #892, fix(source-pipeline): treat an unreachable upstream as unavailable, not fatal #877 → fix(source-pipeline): treat an unreachable upstream as unavailable, not fatal #893, this PR).Verification
just checkpassed:validated 730 source RPMs, 1491 tests and 3019 subtests.just factory-checkpassed.Remaining work
python-typing-inspectionstays on the list. Its fixed native Canary availability contract requires a separate source change and native-runner validation.enchant2stays on the list because its candidate GitHub release archive did not match the existing SHA-512.graphenestays on the list because its candidate GitHub release URL returned 404.kde-filesystem(teamnames) andkf5(macros.kf5) remain Fedora-primary. They are Fedora package metadata and RPM macros, not upstream release archives, and their URLs are already pinned to fixed Fedora commits.Refs #42 — this batch moves two locks; the rest of the Fedora-primary list remains, so the issue stays open.