Repository navigation
Catalog freshness: see prerelease releases of a pinned prerelease stream (Nautilus 2.0.0rcN; trading PR-6) - #707
Conversation
…eam (Nautilus 2.0.0rcN) github_freshness.py and catalog-freshness.yml read a bounded, paginated release list beside releases/latest and, for a pin in a prerelease stream, report the newest published, non-draft release of the pinned major stream, prereleases included; reaching the page cap reports unknown, never a silent miss. A stable 1.x backport created after an rc no longer hides it. Fixtures cover rc6 after rc5, a later stable backport, drafts and the cap. GPT-6.1 Sol build freshness-1 through the packaged SDK worker; the coordinator commits. Co-Authored-By: Claude Opus 5.5 <[email protected]>
…fore the final hot-file commit Co-Authored-By: Claude Opus 5.5 <[email protected]>
… (hot-file protocol: every hot-file edit in the last commit) Co-Authored-By: Claude Opus 5.5 <[email protected]>
…it unknown/no-release states (Opus p1 + p2s) GPT-6.1 Sol repair round freshness-r1 (answers the Claude Opus batch-7 read) through the packaged SDK worker; the coordinator commits. Co-Authored-By: Claude Opus 5.5 <[email protected]>
…fore the final hot-file commit Co-Authored-By: Claude Opus 5.5 <[email protected]>
… (hot-file protocol: every hot-file edit in the last commit) Co-Authored-By: Claude Opus 5.5 <[email protected]>
seathatflowsinourveins
added
the
lane:shared
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: d3f068a059
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
…egistry plus the owned rows) Co-Authored-By: Claude Opus 5.5 <[email protected]>
…fore the final hot-file commit Co-Authored-By: Claude Opus 5.5 <[email protected]>
… (hot-file protocol: every hot-file edit in the last commit) Co-Authored-By: Claude Opus 5.5 <[email protected]>
|
Dependency limit exceeded — report not shown. This pull request scan exceeded the 10,000-dependency limit applied to this scan, so the results are incomplete and may be inaccurate. To avoid reporting false positives, Socket has not posted a report. Upgrade your plan to raise the dependency limit and get complete reports, or view the partial scan in the dashboard. Socket is always free for open source. If this is a non-commercial open source project, contact us to request a free Team account. |
…anking, dormancy, fetch scope, aliases, anti-pattern log) - github_freshness.py ranks release candidates by parsed version first, with publication time only as a tie-breaker, so a republished rc5 can no longer hide rc7 from an rc6 pin. - Runtime entries that declare tags no longer require a release-list fetch. A slug needs the list only when one of its entries actually uses the release-stream path, so a failed unneeded fetch cannot suppress valid drift. - build_manifest.py counts the selected release-list publication as dormancy activity, so a row cannot show a fresh RC and dormant: true together. - A refetch removes every retained record with the same normalized slug, aliases included, before storing the new one. Unrelated repositories are preserved. - docs/harness-defaults.md's anti-pattern log gains rows for the proven mistakes: import-time execution, uncertainty reporting and runtime-tag comparison. Each row has its prevention and verification path. Each fix has a regression test that was red before it. The decision record follows the behaviour. Built by GPT-6.1 Sol through the packaged SDK worker (round freshness-t1); committed by the coordinator. Co-Authored-By: Claude Opus 5.5 <[email protected]>
…fore the final hot-file commit Co-Authored-By: Claude Opus 5.5 <[email protected]>
… (hot-file protocol: every hot-file edit in the last commit) Co-Authored-By: Claude Opus 5.5 <[email protected]>
…ues, row checks, documented dormancy, fail-closed tag misses) - The alias cleanup guards dictionary access and normalizes aliases with github_slug. Unrelated null, string and list records survive refetch and resume, which fixes the t1 regression. Mixed-case and .git release URLs resolve to the fresh record. - The anti-pattern log regression now covers all seven prior repair rows and the three new rows. It checks uniqueness, non-empty fields, and every cited method and its owning class by AST, with twelve negative controls. The decision record's coverage claim is corrected. - Dormancy keeps its per-row channel semantics, documented: stable-pinned and tag-declared rows ignore RC activity fetched for another row. The rendered explanations say what activity each row considered. - For a tag-declared prerelease pin, a missing or unmatched tag list now publishes not_compared, with pin_behind_upstream null and an explicit tag_pattern_unfetched or tag_pattern_unmatched reason, instead of comparing against another package's releases/latest. The behaviour is fail-closed and covered by regressions. Built by GPT-6.1 Sol through the packaged SDK worker (round freshness-t2); committed by the coordinator. Co-Authored-By: Claude Opus 5.5 <[email protected]>
…fore the final hot-file commit Co-Authored-By: Claude Opus 5.5 <[email protected]>
… (hot-file protocol: every hot-file edit in the last commit) Co-Authored-By: Claude Opus 5.5 <[email protected]>
…ing; anti-pattern controls reach the AST checks; stable-pin fallback covered - tests/test_catalog_freshness_trading.py:587 asserts the reworded dormant summary sentence. It is shared through the module, so wording and test cannot drift apart again. The trading report module joins the record's targeted acceptance. - The anti-pattern verification gains source-side negative controls with the log unchanged: a renamed cited method, and a cited method moved to another class. Each control asserts the exact failing subtest and message, so deleting the class or method ownership check now fails a control. - A tag-declared stable pin with a tag miss keeps its documented fallback comparison against releases/latest, and a regression now covers it. Built by GPT-6.1 Sol through the packaged SDK worker (round freshness-t3); committed by the coordinator. The coordinator re-ran the trading, runtime, propose and sota_convergence suites (exit 0) and validate (exit 0). Co-Authored-By: Claude Opus 5.5 <[email protected]>
…fore the final hot-file commit Co-Authored-By: Claude Opus 5.5 <[email protected]>
… (hot-file protocol: every hot-file edit in the last commit) Co-Authored-By: Claude Opus 5.5 <[email protected]>
|
Trading-lane acknowledgement (lane:shared, docs/lanes.md) at The trading side of this change is acknowledged:
The basis is the cross-family review chain. The Claude Opus delta reads t1, t2 and t3 cover every round through this head. t3 returned ACCEPT with no P1 or P2, and confirmed that the t2 trading report assertion (tests/test_catalog_freshness_trading.py) follows the new wording. The coordinator's runs of the trading, runtime, propose and sota_convergence suites and of scripts/validate.py exited 0 at this head. All five review threads are answered and resolved. The one CI failure at this head is tests.test_adaptive_paper_metrics FileSdRegistrationTests, which hit its 30-second exporter registration deadline. That module is outside this change, and the failed job has been re-run. 🤖 Generated with Claude Code |
…egistry plus the owned rows) Co-Authored-By: Claude Opus 5.5 <[email protected]>
…fore the final hot-file commit Co-Authored-By: Claude Opus 5.5 <[email protected]>
… (hot-file protocol: every hot-file edit in the last commit) Co-Authored-By: Claude Opus 5.5 <[email protected]>
|
Claude session native-agent-stack-5f: landing at head Observed main Required checks at this head: 7 pass . Unresolved review threads: 0. |
|
Claude session native-agent-stack-5f: post-merge observation. Landed as |
… its own readiness message, not by a set difference The FileSdRegistrationTests helper start() treated a reused address as missing registration. An unfinished trial keeps its target registered after a clean stop. When a later exporter in the same test binds the same address, update_file_sd replaces that address's group, so set(listed()) - before stays empty and the helper waited out its 30-second deadline. This is the recurring CI failure "exporter never registered its target" (tests/test_adaptive_paper_metrics.py:467), seen on #707 and #725 today. Root cause, confirmed by measurement: - CPython's HTTPServer sets allow_reuse_address, and TCPServer.server_bind sets SO_REUSEADDR. - 1,000 sequential --port 0 exporters reused a retained address 378 times (37.8%). Every one registered, in 55.6 ms median and 79.3 ms maximum. The fix: - start() reads this child's own readiness message (its listening address and file_sd path) from non-blocking stderr, and confirms that the announced address is listed. - It keeps the 30-second bound and fails at once if the child exits. Regressions: - Forced sequential port reuse: failed after 30.2 s with the old helper; passes now. - A never-registering exporter (--no-file-sd on a reused address) is rejected within its bound. - 20 consecutive runs of FileSdRegistrationTests: 380 tests, all passing. Built by GPT-6.1 Sol through the packaged SDK worker (round filesd-flake-r1); committed by the coordinator. The builder's runs each exited 0: tests.test_adaptive_paper_metrics (42 tests), the 20-run loop, validate, evidence_manifest --check and git diff --check. Co-Authored-By: Claude Opus 5.5 <[email protected]>
…md gained #707's rows) and cite only what publication-validation.json records (gate read 730c b, c) Co-Authored-By: Claude Opus 5.5 <[email protected]>
Scope
A published NautilusTrader 2.x prerelease can be absent from
releases/latest, leaving av2.0.0rc5pin without a drift report when rc6 is published. This change adds bounded release-list reads for prerelease pins and applies the same selection and comparison policy to the workflow summary, manifest, and trading report.3b8f9c8a1b938358d65bf422f61592dc3b89407dlane:shared; the other lane's acknowledgement is required before merge..github/workflows/catalog-freshness.yml,tools/sota-convergence/github_freshness.py,tools/sota-convergence/build_manifest.py,tests/test_catalog_freshness_pins.py,tests/test_catalog_freshness_trading.py,docs/decisions/2026-10-04-nautilus-prerelease-freshness.md, andmanifests/evidence.json.Prerelease pins select the newest published, non-draft release in the pinned major, including prereleases and final releases. A later stable 1.x backport does not hide a 2.x RC. Stable pins retain their existing
releases/latestbehavior. Reads are capped at three pages of 100 records; a full third page reportsunknown beyond cap, even when a candidate appears within the window.SOTA sources
per_page/pagecontrols, and publication fields; official documentation read on 2026-10-04.Evidence-class table
Local commands run
All resumed checks used
rtk nice -n 19withTMPDIR="$PWD/.tmp-build/.pytest_cache/freshness".Historical full-suite command
python3 -m unittest -v --durations 50exited 1 with ENOSPC before completion and workspace-TMPDIR-sensitive fixture failures. It was not rerun under the latest instruction, so full-suite acceptance remains unqualified. Earlier interrupted or failed attempts are retained in the decision record.Decision record
docs/decisions/2026-10-04-nautilus-prerelease-freshness.mdrecords the endpoint choice, pagination bound, alternatives, regression evidence, completeness review, and conditions for revisiting the cap or supported version grammar.Host evidence
Not applicable: no files under
evidence/hosts/changed. These checks establish local integration and artifact consistency. Live upstream rc6 observation and hosted CI remain unperformed.Checklist
permissions: {}and job-specific grants.No commit or push was performed; the coordinator owns integration.
Review chain
d3f068a0593a(0 p1, 0 p2).🤖 Generated with Claude Code