Recorded 14 October 2026
Independent OSS validation
These notes describe local checks prepared with Codex. They distinguish an existing contributor's correction, a controlled export reproduction and public-package delivery.
Contribution status and next work
Status checked 14 October 2026 at 19:26 UTC
The selected catalogue lists 17 merged PRs excluding one reverted change, plus 13 open proposals. All 31 listed PR states and merge dates matched direct GitHub data. The reverted documentation change #58 remains linked to #72.
On narrow screens, scroll the tables sideways to read every column.
| Observed status | PRs | Proposals |
|---|---|---|
| Failed workflow | 4 | teslamotors/react-native-camera-kit #812 · teslamotors/react-native-camera-kit #813 · astropy/reproject #630 · UNSW-CEEM/nempy #43 |
| Awaiting approval | 4 | teslamotors/react-native-camera-kit #817 · teslamotors/react-native-camera-kit #816 · LSSTDESC/CCL #1310 · UNSW-CEEM/nempy #44 |
| Successful workflow | 2 | pydicom/pydicom #2378 · pydicom/pydicom #2379 |
| No upstream run observed | 3 | unitreerobotics/xr_teleoperate #310 · unitreerobotics/unitree_sdk2 #34 · lucidrains/open_clip #2 |
These are dated workflow observations. Successful runs describe the recorded commits; approval requirements and missing runs still need follow-up. The submitted reproject source still fails test_coadd_solar_map. Earlier baseline controls and geometry records identify nine rejected mappings whose effect propagates through background matching. Recorded Linux and Windows validation retains that failure and separately verifies a reference-only candidate. The reference change remains outside the PR pending the maintainer's decision. No new reproject test ran during this follow-up. Failed-job lists for the CameraKit and Nempy runs were empty, so their failure causes remain unconfirmed.
Eight contribution paths
The paths below cover authored PRs, independent validation and possible follow-ups. Local results remain tied to the source revisions recorded in these notes.
| Path | Current position | Next useful step |
|---|---|---|
| MuJoCo 3689 · 3690 | pranay172's two PRs remain open. CLA checks succeed; build and lint workflows await approval. | Keep the existing author credited. Offer coordinated validation or a distinct follow-up. |
| BlueOS 4472 · 4460 | #4472 is open and unassigned; #4460 remains a draft. The earlier local export reproduction is preserved. | Agree selected-file inclusion, limits and authoritative closed-file state before a production patch. |
| BSK-RL / Basilisk 354 | #354 remains open. GitHub reports mergeable=false; local conflicts were not inspected. | Resolve the existing dependency and agree ownership before a resource checker. |
| Wire Tool / EDM 2 | Issue #2 remains open with no comments. The referenced conversion archive is still uninspected. | Inspect the exact assembly and archive before reporting a missing asset. |
| OpenDrift 1877 · 1862 | #1877 remains open without comments; overlapping #1862 is closed and unmerged. | Clarify the physical and stochastic convention before proposing another correction. |
| Datarepo datarepo · data-repository 0.1.0 | Six accepted fixes remain merged. Installed public 0.1.0 still drops the null predicates corrected in #57. | Track delivery of the accepted correction in a published artifact; keep the other five contracts unverified. |
| pydicom 2378 · 2379 · 2380 | The two LUT PRs remain open with successful submitted-commit workflows. #2380 is merged; public 3.0.2 predates it. | Keep the LUT scope stable and verify a published artifact containing the encoding correction. |
| CCL 1310 | #1310 remains open; its workflow awaits approval. Scientific review is still outstanding. | Prepare one compact numerical comparison for maintainer review. |
Status snapshot with submitted commits, workflow links and limits
MuJoCo: validating an existing correction
pranay172's #3689 corrects mesh distance signs at shared edges and vertices. Independent checks compared candidate 89733a51f207c7fedafc855d22ffd4c38caa793a with parent 13ef0c8acf8863cb4246ff06f06fe55e005e0388.
The parent produced incorrect signs on a closed prism and the reporter's mesh. The candidate removed those mismatches in double and single precision. Ordinary geometry and concave controls passed on both. Loaded native-library identities were checked on macOS arm64.
| Suite | Double precision | Single precision |
|---|---|---|
| User objects | 103 passed | 98 passed, five skipped |
| User mesh | 84 passed | 84 passed |
Early setup checks caught a working-directory error and an overly strict frame guard. The guard correction follows the source normalization contract; geometry and sign tolerances stayed unchanged. Full upstream tests, Ubuntu, sanitizers, a controlled performance benchmark and a released candidate artifact remain unverified. This validates the existing contribution; the correction belongs to its author.
BlueOS: reproducing a recording omission
The unchanged Commander export handler and filebrowser client ran at source 224b80ca07156be8238ece4a096dca33dc1db775, against native filebrowser v2.30.0. Synthetic path aliases modeled the source-declared log and recorder directories. The official MCAP writer finalized two authored numeric recordings.
| Input or condition | Result |
|---|---|
| Recordings in the recorder directory | Ordinary log ZIP omitted both files |
| Same closed MCAP placed under the log root | ZIP included identical bytes |
| Authored diagnostic collector raises an error | HTTP 200 with a usable archive |
| Native archive backend stopped | HTTP 502 before streaming |
Host diagnostics were replaced with authored content, logging discovery was disabled and aliases pointed only to the synthetic system. No recorder or vehicle ran. This demonstrates a controlled source-path mismatch; deployment mounts, active-file handling, cancellation and large-storage behaviour remain untested. Production inclusion needs an agreed scope and a reliable closed-file state source.
Inspect the authored artifacts: ordinary archive, path-control archive and closed MCAP fixture. Their hashes and archive inventories are in the results file.
Upstream context: #4472. The extraction draft #4460 remains a separate contribution.
Merged work and public-package delivery
Neuralink's library is published as data-repository. The 0.1.0 wheel's official hash matched the download. Its metadata still declares Python >=3.8, while current source declares >=3.10. It predates the six accepted merges. On 14 October at 20:43 UTC, the official wheel was installed in isolation outside the source checkout. Its loaded module bytes match the downloaded wheel. Seven query-building checks of the accepted null-predicate correction produced two passing controls and five failures: null predicates disappear, AND/OR groups change, and unsupported operators are silently ignored. These checks exercised generated SQL; no ClickHouse backend ran. The other five accepted contracts remain untested. Artifact identity, expected queries and actual results.
pydicom's one-bit encoding correction is merged. The observed public 3.0.2 upload predates it; a containing public artifact has not been verified. The two lookup-table proposals remain separate open work.