Repository navigation
Repair macOS validation portability and failure diagnostics - #948
Merged
seathatflowsinourveins merged 2 commits intoOct 10, 2026
Merged
Conversation
|
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. |
seathatflowsinourveins
force-pushed
the
codex/ns2604-validate-macos-20261009
branch
from
October 10, 2026 00:38
6459b97 to
379f9df
Compare
seathatflowsinourveins
marked this pull request as ready for review
October 10, 2026 02:25
|
You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard. |
seathatflowsinourveins
deleted the
codex/ns2604-validate-macos-20261009
branch
October 10, 2026 02:43
5 tasks done
seathatflowsinourveins
added a commit
that referenced
this pull request
Oct 10, 2026
… manifest on main since #948 origin/main 30aa2fa fails scripts/validate.py on its own: the manifest row for this test holds sha256 72d7e709... (20,631 bytes) while the file, last changed by a56f85e (#948), is 4601bbba... (20,940 bytes). Re-registered with host_receipts.register_file; this PR does not change the file. Co-Authored-By: Claude Opus 5.5 <[email protected]>
seathatflowsinourveins
added a commit
that referenced
this pull request
Oct 10, 2026
…slot (#923) * Claude Code native practice, 2026-10: one evidenced default per role slot Decision record for Claude Code 2.1.295 covering 33 role slots across the native client, skills, workers, MCP, headless, GitHub automation, token and cost practice: a capability matrix, a supersession table, one default per slot with ranked alternatives and overturn conditions, the GitHub automation requirements, measurements owed and a weekly refresh loop owned by the currency lane. The defaults ship inside the claude-native-practice project skill (reference/slots.md, one page per layer, slots.json), so they travel with a user-scope install. Evidence artifacts: claims with locators, supersession, verification and adjudication results, G5 candidate rows for grand-catalog, a practice-references candidate file, the trigger evaluation and the fresh-session proofs. The 2026-10-02 automation record gets a dated addendum: seven required checks since 2026-10-05. Co-Authored-By: Claude Opus 5.5 <[email protected]> * Escape the quoted $USER in verification.json The previous rewrite inserted raw double quotes into a JSON string, so validate_convergence.py could not parse the file during record discovery (CI validate-shard-0). The builder now sanitizes string values before serializing, and each written artifact is parsed back. Co-Authored-By: Claude Opus 5.5 <[email protected]> * Show organic activation of the practice skill in the fresh-session proof The co-op's read of b0e713f (P2) found that the fresh-session evidence kept only booleans and hashes. fresh-session.json now carries, for each of the 11 runs, the launch command, the exact request (none names the skill), the first Skill invocation with its arguments and the first read of reference/slots.md, copied from the private transcripts; the record's handoff text cites them. Co-Authored-By: Claude Opus 5.5 <[email protected]> * Re-register this branch's files after rebasing onto 3497f50 Co-Authored-By: Claude Opus 5.5 <[email protected]> * Publish every skill-directory read so the activation rates recompute The excerpts kept only the first call that named "claude-native-practice/reference" as one string, so shell reads that cd into the skill directory and then name reference/slots.md were missed or left as a placeholder. Every call that reads from the skill directory is now listed with its position, tool and exact (sanitized) path or command. Recomputed from those events for the shipped description: the skill was the first tool call in 5 of 6 sessions; 4 of 6 then read reference/slots.md whole (2 by Read, 2 by cat) and a fifth grepped it; the sixth searched the skill directory without invoking the skill. The per-run flags, rates, statement and the decision record now say so. Co-Authored-By: Claude Opus 5.5 <[email protected]> * Grant the practice skill's SKILL.md in the reviewed architecture inventory scripts/local_pages_policy_grants.py selects every .claude/skills/*/SKILL.md, and this branch adds .claude/skills/claude-native-practice/SKILL.md, so the committed repository grants (90) no longer matched the inventory at the merge head (CI run 38016037590, test_committed_repository_grants_match_current_git_inventory_paths). Adds the repo-root row the script proposes (`--pin 998eea1`, 91 rows, equal to the committed set) and moves the test's count to 91. The receipt grants are pinned to their reviewed SHA and do not move. Co-Authored-By: Claude Opus 5.5 <[email protected]> * Re-register tests/test_local_pages_fleet_regressions.py, stale in the manifest on main since #948 origin/main 30aa2fa fails scripts/validate.py on its own: the manifest row for this test holds sha256 72d7e709... (20,631 bytes) while the file, last changed by a56f85e (#948), is 4601bbba... (20,940 bytes). Re-registered with host_receipts.register_file; this PR does not change the file. Co-Authored-By: Claude Opus 5.5 <[email protected]> --------- Co-authored-by: seathatflowsinourveins <[email protected]> Co-authored-by: Claude Opus 5.5 <[email protected]>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Scope
The macOS adoption suite failed when canonical paths were mixed with
/varaliases, test instrumentation assumed Linux/proc, and Ubuntu/WSL integration commands ran on Darwin. This repairs portable path/descriptor/report interfaces, names the native platform requirements, and prints failure headers in both macOS test modes while preserving unittest's exit status.a7bb4a51eea9cc321dd06dd6243a5b2654a5f805.eb90128f6ae2a1e4b48c5517332b8bb594d5ad9e, two commits on the base.lane:foundation.The observability decision's isolated-render citation now points to
tests/test_new_wsl_client_config.py:641-656. Its registered evidence hash matches the corrected document bytes.SDK containment and relative references use one resolved copy while the independent reader guard retains original lexical declarations, exact roles and no-follow checks. The reviewed reader digest/caller positions are updated together. Normalization cannot hide an unapproved descendant symlink or traversal selector.
Sealed producer execution remains a native Linux guarantee. Unsupported systems expose finite unavailable metadata and preserve dated snapshot/Actions observations. Snapshot/seal failures and later transport failures have distinct reasons. The cache uses exclusive mode-0600 creation through its checked directory descriptor instead of a
/proctemporary pathname. Linux collection tests retain the real sealed-descriptor path; only systems without that native API mock the transport. The cache regression asserts explicit close-on-exec creation flags and fails when that flag is removed.Both review workflows already used
wc -c; the change normalizes BSD padding with POSIXtr. Static/event-policy and portable tests remain active. Only integrations that execute frozen Linux/Ubuntu commands or unsupported shell capabilities are guarded, with specific reasons. The current advisory macOS event policy is preserved.Scout alias migration runs on Darwin with a resolved fixture root. Apple's maintained
readlink -fimplementation is portable for this command; the former skip hid a fixture alias mismatch. An actual symlink temporary prefix reproduced failures for the previous/current owners before resolution, and all six ownership cases pass after it. Claude capture guards nameflock; the Inspector probe guard names the Linux port-range file andsetsid, whilessis stubbed.SOTA sources
_mkstemp_inner, native API gates and public os descriptor interfaces: exclusive/random/mode-0600 creation, descriptor-relative access and supported memfd availability. Installed tempfile.mkstemp has no dir_fd; os.open supplies the existing checked-descriptor adaptation.wc -candtr: portable byte-count formatting.dfeea13377cfb15936f856d9ed8df3c6575a7895, job114030975988, step21: actual Darwin full-suite failure, distinct from local simulations.-fand callsrealpath. This is primary source evidence for the portable Scout command.Evidence-class table
The artifact's412 headers are source-policy tests, including407
/procwatcher failures; six are direct readiness failures. The decision record separates these families from the dispatch's shorthand and does not claim a fresh100-job census.Local commands run
Implementation regression runs below were recorded at379f9df6. The citation-only
forward commit changes no executable code. FULL validation and the actual-host
tip/new-push history gate were refreshed at eb90128.
The final checks above ran as ordinary uncapped native commands. Alias, missing-API and BSD-output fixtures establish the indicated Linux simulations. A native Darwin after-result awaits the next eligible full-suite run.
The initial integration batch found one strict inventory omission for Fleet's
new platform boundary; the selector inventory now includes it and the strict
checks remain intact. An unmodified hang-trace control failed once during a
follow-up workflow group, then passed in an isolated17-test check and the final
146-test native group without a code change. Both outcomes are retained. That
earlier15-module batch ran946tests/4skips; those historical measurements are
distinct from the implementation400-test gate. The first attempt at that gate
used tests/ as CWD and reported three loader import errors; repository-root
imports passed and the corrected implementation gate returned rc0 with400tests/3skips.
Final FULL output SHA256:
edbe447c81a10cad7c1cf4e3eba19131dd89781e912314823f9e30ba02d3cc53.
Actual native-session tip/history output SHA256:
ba498e69af02df2e2273e3d9d7fae58bf436b71f8a2e2a730453a16b8c527949.
Implementation400-test output SHA256:
ba521499a7ea59e5a2bdab26e8c19f643a89dea16429bfb788767e274d6b087c.
Implementation165-test alias output SHA256:
b522a10377193ac5784eb9b8be22751cec56199845570f5b8c28256570efa74c.
Decision record
docs/decisions/2026-10-09-validate-macos-portability.mdrecords native before evidence, root causes, adopted primary interfaces, regression controls, exact platform boundaries and the native after-run limitation. The next eligible main/nightly/dispatch macOS run supplies that native evidence.Host evidence
N/A: no evidence/hosts receipt or platform-status change.
Checklist
This remains a draft foundation PR. Hosted checks at the stated head are tracked
in the PR Checks view; they are separate from the retained native fail-before
artifact, local tests and the outstanding native Darwin after-run.