reconcile: land nightshift's 26 diverged commits + bump dips pointer - #53
Open
miles-on-nightshift wants to merge 27 commits into
Open
reconcile: land nightshift's 26 diverged commits + bump dips pointer#53miles-on-nightshift wants to merge 27 commits into
miles-on-nightshift wants to merge 27 commits into
Conversation
Fixes #47. The template was posting a generic "is this still active?" comment on all stale PRs including agent-authored ones, which is self-addressed noise. Changes: - Add reviews, headRefName, baseRefName, mergeable to gh pr list fields - Step 6 now branches on author/state/mergeable instead of posting one template - Agent-authored PRs: never nag self; file task or escalate with specifics - External-author PRs: specific nudge only when ball is genuinely with them, deduped - Null reviewDecision / UNKNOWN mergeable: skip and re-check next run Co-Authored-By: Claude Sonnet 4.6 <[email protected]>
Root cause of 2026-08-12 winston.jsonl chain fork (resolved in merge 221efd0): the log_ownership_guard's actors() trusted DATACORE_ACTOR unconditionally. On the mac, DATACORE_ACTOR=winston was set during a test, making the guard treat mac as authorized to write winston.jsonl. Fix: when the infrastructure registry maps this host to specific ledger_actors, that mapping is authoritative and DATACORE_ACTOR is ignored for ownership checks. New machines without a registry entry fall back to the prior behavior (DATACORE_ACTOR then hostname). Also fix two pre-existing test failures caused by GIT_AUTHOR_EMAIL being set in the environment, which overrode git -c user.email in test repos and broke the author-filter in changed(). Strip GIT_AUTHOR_* / GIT_COMMITTER_* from subprocess environments in: - TestLogOwnershipGuardAuthorship (test_hooks.py) - b3_ownership_guard (test_ledger_attacks2.py) Co-Authored-By: Claude Sonnet 4.6 <[email protected]>
Co-Authored-By: Claude Sonnet 4.6 <[email protected]>
Step 7 was writing extracted tasks directly to next_actions.org, bypassing GTD's single capture point. Fixed to capture to inbox.org so tasks go through triage before becoming committed work. /continue --save already targets inbox.org correctly. Co-Authored-By: Claude Sonnet 4.6 <[email protected]>
Nightshift agents sometimes write report files to the root Data repo's reports/ directory, which the pre-commit allowlist rejects (only .datacore/ and .github/ are permitted at root level). Reports belong in space repos. Ignore /reports/ to prevent pre-commit failures until agents are updated to write reports to the correct space. Co-Authored-By: Claude Sonnet 4.6 <[email protected]>
Committed by git_fleet_sync. These changes were made by agents on this host and never committed, so they existed on exactly one disk and were invisible to every other agent. Skipped as junk: 1 file(s).
Committed by git_fleet_sync. These changes were made by agents on this host and never committed, so they existed on exactly one disk and were invisible to every other agent. Skipped as junk: 1 file(s).
CLAUDE_CODE_SESSION_ID is not set in hook process env — session_id is only available in the stdin JSON payload. All four hook scripts now call set_session_id() after reading stdin, which caches the value for the process. session_state.py gains set_session_id() + _session_id_override fallback in _get_session_id(). Without this, _state_file() always returned None and the sessions/ directory stayed permanently empty. Co-Authored-By: Claude Sonnet 4.6 <[email protected]>
…3 hook industry_landscape_file default was pointing to a nonexistent path (0-personal/notes/2-knowledge/industry-landscape.yaml). The actual file is at 1-datafund/1-tracks/research/Industry landscape.md. - module.yaml: correct default path and description (YAML → md) - research_orchestrator.py: wire LANDSCAPE_FILE to _setting_path() and remove the "deliberately NOT wired" note comment - CLAUDE.base.md, README.md, research-status.md: update docs - hooks/pre-commit: fix python → python3 (python not on PATH) Co-Authored-By: Claude Sonnet 4.6 <[email protected]>
git_fleet_sync was exiting 0 even when pull conflicts occurred, causing systemd to see success and never escalate. 87 failures over 14 days went unnoticed as a result. - Return 1 when any pull conflict or push/commit failure is detected - Also surface FAILED statuses in a dedicated section - Add "N repo(s) needing a human" to the summary line (mirrors the interactive ledger_transport.py output that already gets this right) systemd's OnFailure= can now fire for real failures. Addresses items 1 and part of 3 from issue #48. Co-Authored-By: Claude Sonnet 4.6 <[email protected]>
#48 item 2) Adds two pieces: - fleet_sync_alert.sh: sends a Telegram message with the last 30 journal lines when fleet-sync exits non-zero. Reads TELEGRAM_BOT_TOKEN / TELEGRAM_CHAT_ID from the environment; degrades to a journalctl entry if credentials are absent. - install_fleet_sync_timer.sh: generates datacore-fleet-sync-alert.service alongside the main service. OnFailure= wires them together so systemd fires the alert on any non-zero exit. EnvironmentFile=-/home/${USER}/config/nightshift.env loads credentials (the `-` makes it optional). Together with c594873 (exit non-zero on conflict), the two-week silent- failure window is closed: a conflict now both exits 1 AND pages the human. Deployment: re-run install_fleet_sync_timer.sh on nightshift to install the new unit files, then systemctl daemon-reload. Co-Authored-By: Claude Sonnet 4.6 <[email protected]>
wave-trader.service was crash-looping (2927 restarts, every 30s) on `from datacore.ledger import attests` -> No module named 'datacore'. The unit's Environment=PYTHONPATH only had the gateio dir; the datacore package lives at .datacore/lib. Appended .datacore/lib so the DIP-0047 ledger import resolves. Applied live in /etc/systemd/system/ and now tracked here as a <HOME> template alongside the sibling units so a re-provision doesn't reintroduce the crash loop. Co-Authored-By: Claude Opus 4.8 <[email protected]>
The nightshift box's main was 26 ahead / 63 behind origin — real local work (nightshift batch-end output, sync'd trapped agent files, ledger and fleet-sync fixes, the wave-trader unit) that never converged with origin's canonical line. This merge integrates both directions and bumps the dips submodule pointer to the reconciled head (8522065, includes datacore-dips#5). Conflict resolution — origin adopted as canonical for all 7 conflicted files, since origin (plur9) had already superseded this box's infra edits with more-evolved versions: - git_fleet_sync.py -> origin (nuanced #48 fix: fails on conflicts, not held repos) - install_datacore_path.py-> origin (fixes fallthrough to system dir / PermissionError) - ledger/fold.py -> origin (origin ledger.py identical + has full DIP-0047 attests) - ledger_attest.py -> origin (clearer $HOME doc) - wave-trader.service -> origin (scrubbed path example) - commands/wrap-up.md -> origin (canonical) - research_orchestrator.py-> origin (line-consistency) FLAGGED for review (local had content origin lacks — dropped here, noted for follow-up, not lost; recoverable from this branch's first parent): - wrap-up.md: local auto-added GTD tasks to inbox.org (:wrap_up_extracted:); origin uses propose-then-confirm. Behavioral choice — yours to make. - research_orchestrator.py: local had a modern claude_agent_sdk async impl (budget cap, rate-limit handling); origin still uses subprocess CLI. Worth porting if you want the SDK path. - ledger/fold.py: local set item.artifact_commit from payload; origin does not. Confirm origin handles artifact_commit elsewhere before relying on it. Co-Authored-By: Claude Opus 4.8 <[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
1 participant
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.
What
The
nightshiftbox'smainhad diverged: 26 ahead / 63 behind origin. This PR reconciles both directions via a merge and bumps thedipssubmodule pointer to the reconciled head (8522065, includes datacore-dips#5's 3 landed drafts).The 26 local commits are real trapped work: nightshift batch-end output, "sync: land agent work trapped on this machine" files, ledger/fleet-sync fixes, and the wave-trader unit. All non-conflicting local content is preserved; origin's 63 commits are integrated.
Conflict resolution — origin adopted as canonical (7 files)
Origin (plur9) had already superseded this box's infra edits with more-evolved versions, so origin won every conflict:
git_fleet_sync.pyinstall_datacore_path.pyledger/fold.pyledger.pyis identical + already carries full DIP-0047attestsledger_attest.py$HOMEdocwave-trader.service/home/<account>)commands/wrap-up.mdresearch_orchestrator.pyI defaulted these to origin but they encode real choices — flip any in review:
inbox.org(:wrap_up_extracted:tag); origin uses propose-then-confirm. Behavioral/UX choice.claude_agent_sdkasync impl (budget cap, rate-limit handling); origin still shells out to theclaudeCLI via subprocess. Worth porting if you want the SDK path.item.artifact_commitfrom payload; origin does not. Confirm origin handlesartifact_commitelsewhere before relying on it.All three are recoverable from this branch's first parent if you want the local version.
Result
After merge, the nightshift box fast-forwards to origin and the 26/63 divergence is gone. dips is already reconciled (fast-forwarded locally, at 8522065).
🤖 Generated with Claude Code