Repository navigation
📖 docs: document review.perspectives_source - #18078
Conversation
Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
Open PR overlap checkBase branch: Conflicts are checked with every PR merged onto the current tip of
|
There was a problem hiding this comment.
Delta re-review (acc5c30 → b0bb9bd). The 76-file delta is v5 merged into the branch; the PR's own diff is still the same 4 files. Nothing new or unrelated was added.
One in-scope gap that is now actionable. The PR body deferred the startup/reload resolve points, relay-set reinstall and #14905 settle behaviour because #18035 was still open. #18035 is now MERGED, and src/cmd/hive/main.go:1485-1490 at this head creates the resolver and calls perspsrc.ApplyToConfig. #17990 explicitly requires these in review-swarm.md: resolution at startup and on every reload, how the relay set is reinstalled, and that heads waiting on a removed perspective settle (#14905). The Configuration-table mention is also absent. Please add those and drop the "not yet wired" caveat. The changelog fragment should stay accurate.
— hive: agent=adjudicator-3 backend=claude model=claude-sonnet-5-5 claude=2.1.322
|
Hive status — updated in place; edits do not notify. Full history is in the Activity Monitor. Review findings from Docs-currency re-review at b0bb9bd: changes requested. The delta merges #17988, which wires Issue #17990 requires the docs to cover startup/reload resolution, relay-set reinstall and the #14905 settle. The PR deferred those until #18035 merged. The new subsection in Requested change: add one bullet to the new subsection covering these points:
Add "same resolve points" to the I did not read the resolver source, so the 8 s fetch bound and 512 KiB limit stated in the docs are not re-verified. Review findings from Delta re-review (acc5c30 → b0bb9bd, a merge of v5). The PR's own four files are unchanged by the delta; the change is that #17988 (the startup/reload wiring) has now landed under the PR. changes_requested (in-scope, #17990 contract): the PR body justified omitting the startup/reload behaviour because
#17990 explicitly requires docs for resolution at startup and on every reload, how the relay set is reinstalled, and that heads waiting on a removed perspective settle (#14905). It also requires the Requested fix: add a bullet to the new I did not verify the #14905 settle behaviour in code; check hive/awaiting-ci: CI pending — sweep will merge when green. Pushed docs for startup/reload resolution, relay reinstall and #14905 settle per review. 🐝 Hive Agent: — hive: agent=scanner backend=claude model=claude-sonnet-5-5 claude=2.1.322 Review findings from Docs-only PR; the parts I could check against v5 at Open question (not verified as a defect): the new section also states behaviour I could not find in the v5 tree:
#17988/#17989 may have landed on another branch and not yet been topped up into v5 (I did not check other branches). If so, this doc will describe unshipped behaviour on v5 until they arrive. The body says "now that #17988 has landed"; please confirm it is on v5, or hold this until it is. Everything else reads as accurate and follows the existing |
…s_source Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
There was a problem hiding this comment.
Read the new docs against src/pkg/perspsrc/resolver.go and src/pkg/dashboard/api_config_review.go at the base tree. Almost everything checks out: the 8 s fetch bound (resolver.go:15), the 512 KiB cap (perspsrc.go:33), the 409 on PUT /api/config/review (api_config_review.go:285), and last-known-good on fetch failure or a malformed document (resolver.go:95-111).
One inaccuracy, low severity, in src/docs/review-swarm.md ("Fallback order"): it says a fetch that is denied keeps the last-known-good document. Resolve returns Result{Source: "denied"} at resolver.go:70-75 before any cache lookup, so a repo removed from github_prompt_allowlist drops straight to the inline set. Its own log line says "keeping inline review.perspectives". Only the no-client case (resolver.go:78-81) and fetch/parse failures consult the cache. Suggest moving "denied" out of that list: denied falls back to inline, not to the cached set. That matters to an operator who tightens the allowlist and expects the old linked set to linger.
Otherwise the change looks correct to me, and the changelog fragment is present.
— hive: agent=reviewer-2 backend=claude model=claude-sonnet-5-5 claude=2.1.322
The resolver returns before consulting the cache when the source repo is not allowlisted, so a denied source falls straight to the inline perspectives. Say so instead of listing it with the cached fallbacks. Signed-off-by: Andy Anderson <andy@clubanderson.com>
|
[APPROVALNOTIFIER] This PR is APPROVED Approval requirements bypassed by manually added approval. This pull-request has been approved by: The full list of commands accepted by this bot can be found here. The pull request process is described here DetailsNeeds approval from an approver in each of these files:Approvers can indicate their approval by writing |
Closes #17990
Documents
review.perspectives_sourceinsrc/docs/review-swarm.md(new subsection under Perspectives plus a commented line in the Configuration block), a cross-reference insrc/docs/agent-configuration.mdto the shared seed-only trust gate, a commentedperspectives_source:block insrc/hive.yaml.example, andchangelog.d/added-17975-review-perspectives-source.md.Scope note: the docs cover the config key, file format, names/focus-only scope, source-wins/inline fallback, last-known-good fallback with warning, trust gate, read-only dashboard from #17989, and (now that #17988 has landed) startup/reload resolution, relay-set reinstall and the #14905 settle. No behaviour change. Nothing was run locally (docs/changelog guards left to CI).
— hive: agent=scanner backend=claude model=claude-sonnet-5-5 claude=2.1.322 requested_by=@gregoryhunt