Repository navigation
Conversation
[scanner] CI root cause for the red srpm jobs: pinned packit image digest was pruned from quay (exit 125, manifest unknown). One-line fix at .github/workflows/packit-srpm-pilot.yml:90 — details and verified replacement digest: #74. Agent cannot push workflow-file changes (App lacks workflows permission).🐝 Hive Agent: — hive: agent=scanner backend=copilot model=kimi-k3 |
[scanner] CI root cause identified:
|
[scanner] CI red is systemic, two clusters: (1) all
|
|
[scanner] rebuild failures share the same root cause as #56/#57/#58: ruby3.3/3.4/4.0 default-gems file conflicts in the buildroot transaction. Blocked on utah-packages#83; rebase after it merges. scanner agent (ACMM L5 — hold-gated mode)🐝 Hive Agent: — hive: agent=scanner backend=copilot model=kimi-k3 copilot=1.0.78 |
hanthor
left a comment
There was a problem hiding this comment.
Clean suites, but both files this PR adds are also added by other open PRs. Full matrix in #59; the collisions involving this branch, from three-dot diffs against main (54cc8d1):
61 71 ['tests/test_validate.py']
63 71 ['tests/test_hummingbird_gaps.py']
tests/test_validate.py is also added by #61, and tests/test_hummingbird_gaps.py by #63. Four PRs (#59, #61, #63, #71) between them write four test files, each one twice.
They form a cycle, so two of them cover everything with zero overlap: #59+#71 or #61+#63. Either pair is complete and needs no rebasing; the other two can be closed as duplicates.
This PR is in good shape. Merged into main locally (baseline 72 passed):
PR#71 fc=0 val=0 pytest=0 | validated 193 source RPMs | 103 passed in 1.47s
PR#71 quoting=0 runtime_contract=0 unittest=0 | validated runtime contract: 72 packages
+31 tests, every other gate unchanged. The body's 31 passed for the two files in isolation reproduces as part of the +31 total.
Comparing the overlapping halves, since the pairs are otherwise equivalent:
test_validate.py— #61's version drivesvalidate.pyas a subprocess over a synthetic factory tree, which is exactly how.github/workflows/validate.ymlinvokes it (python3 tools/validate.pyis the whole step). This PR's 15 cases cover the same ground at a similar depth. Close to a wash; #61's harness is marginally more faithful to CI.test_hummingbird_gaps.py— #63's version carries 14 cases plus a guard that no[excluded]package leaks into the contract. That one is worth keeping whichever pair wins:reports/hummingbird-gap.jsonis consumed byprojectbluefin/utah, so a leak there misreports coverage to a downstream repo, not just locally. This PR's 16 cases cover the ISO-8601measured_atand output-directory creation, which #63 also has.
So on content the pairs are close, with #63 holding the single most valuable individual case. Either choice is defensible; the important thing is making it once rather than merging three of the four and discovering the collisions at merge time.
One note on the test_validate.py interaction with #73: #73 refactors tools/validate.py into check_provenance() / main(root) with a __main__ guard. I merged #73 and #61 together and the suite stays green (97 passed), because these tests go through the CLI rather than the internals — the same should hold for this PR's version. Landing #73 first would let a later round of tests call the functions directly instead of shelling out.
Recommendation: pick a pair, close the other two with a pointer to which PR carries their coverage, and have the quality agent re-check disjointness against open PRs at review time rather than only at filing time.
Generated by Claude Code
|
[scanner] CI root cause: both failure classes here are systemic, not from this PR.
Once #83 and #87 merge and this branch is rebased, CI should validate this PRs actual changes. Note: the hive App token currently lacks the scanner agent (ACMM L5 — hold-gated mode)🐝 Hive Agent: — hive: agent=scanner backend=copilot model=kimi-k3 copilot=1.0.78 |
[scanner] The mass red here is environmental, not this PR: (1) every
|
hanthor
left a comment
There was a problem hiding this comment.
Approving. The lgtm here resolves the one thing my earlier review was waiting on — which of the two disjoint pairs to take. The branch is unchanged since I verified it (still 2c45477), so that verification stands: merged into main locally, 103 passed against a 72 passed baseline, +31 tests, every other gate unchanged.
This approval picks the #59 + #71 pair, so #61 and #63 are now duplicates. The four PRs form an exact 4-cycle over four files, each file claimed by exactly two PRs:
| PR | dist_git |
quoting |
validate |
gaps |
|---|---|---|---|---|
| #59 | ● | ● | ||
| #61 | ● | ● | ||
| #71 | ● | ● | ||
| #63 | ● | ● |
Reading down the columns: #59 + #71 covers all four files with zero overlap, and so does #61 + #63. There is no third option, and taking any three of the four means a file gets added twice and the second one fails to merge.
So the remaining work is mechanical, and worth doing before either of these lands rather than after:
- #59 needs the same call — it is the other half of this pair and is still
hold-labelled. - Close #61 and #63 as duplicates, each with a pointer to where its coverage now lives: #61's
test_validate.py→ this PR, #61'stest_check_workflow_quoting.py→ #59, #63'stest_hummingbird_gaps.py→ this PR, #63'stest_dist_git.py→ #59.
One case is genuinely lost in that trade, and it is the single most valuable individual case across all four PRs: #63's guard that no [excluded] package leaks into the contract. reports/hummingbird-gap.json is consumed by projectbluefin/utah, so a leak there misreports coverage to a downstream repo rather than just failing locally. This PR's test_hummingbird_gaps.py covers the ISO-8601 measured_at and output-directory creation that #63 also has, but not that one. Worth lifting it across in a follow-up — it is about fifteen lines and does not need to block this.
Separately, and not blocking: #73 refactors tools/validate.py into check_provenance() / main(root) behind a __main__ guard. These tests drive the CLI as a subprocess rather than the internals, so they survive it — I confirmed that by merging #73 with #61's near-identical version (97 passed). Landing #73 first would let a later round call the functions directly instead of shelling out.
Last one for the quality agent rather than this PR: disjointness is being checked at filing time and not re-checked at review time, which is exactly how four PRs came to write four files twice each. Re-checking against open PRs when the review is requested would have caught this on the second filing.
Generated by Claude Code
…e import, and contract (#71, #72, #56, #57) (#92) * test(quality): unit coverage for validate.py and recalculate_hummingbird_gaps.py * fix(ci): verify cosign binary sha256 checksum in rebuild-rpms Verify the downloaded cosign v2.4.1 binary against its release SHA-256 checksum prior to installation and keyless signing. Closes #46 Closes #66 Closes #68 * fix(tools): validate package and branch inputs in import_rawhide (#56) * fix(contract): exclude anaconda-live from runtime transaction (#57) --------- Co-authored-by: castrojo <castrojo@users.noreply.github.com>
Summary
Adds unit coverage for
tools/validate.pyandtools/recalculate_hummingbird_gaps.py.Closes #60
What this adds
tests/test_validate.py: 15 cases testingtools/validate.pyacross synthetic factory trees (rawhide imports with required commit/tree, upstream imports with no snapshot, invalid schema, missing locks, missing Packit config, and validated count output).tests/test_hummingbird_gaps.py: 16 cases testingtools/recalculate_hummingbird_gaps.py(whitespace/comment stripping inlines(), duplicate collapsing, contract partitioning across image, repo-only, and missing, excluded section handling, ISO-8601 timestamps, output directory creation, and real manifest validation).Verification