Skip to content

fix: restore main CI — platform-aware installer test + the F28 doctrine obeys its own rule (#1109 follow-up) - #1130

Merged
aarontrowbridge merged 1 commit into
mainfrom
1110-main-ci-restore
Sep 14, 2026
Merged

aarontrowbridge merged 1 commit into
mainfrom
1110-main-ci-restore

Conversation

@aarontrowbridge

Copy link
Copy Markdown
Member

Restores main's fast lane — #1109's merge run failed on two darwin-blind spots my local gates (darwin) could not see:

  1. The installer fleet-branch test asserted darwin's terminal outcome ("guard not installed", exit 1) — but on non-darwin the installer deliberately skips the host check ("the fleet is a darwin fleet") and completes --check green. The test now asserts the BRANCH (the parsed role + port flowed through; never the standalone skip) and accepts the platform's correct terminal outcome — darwin: exit 1 + guard-not-installed; non-darwin: exit 0 + host-check-skipped. Read straight from install.sh's own uname branch.
  2. amico-vault's F28 proof case embedded the literal rotted vault names — the doctrine violated its own "never in prose or examples" rule, and its own mechanical lint ([vpath], the skills campaign's skills-integrity lint: catch retired namespaces, dead skill paths, and frontmatter drift in CI #1045/skills-integrity lint: [vpath] rotted vault-path rule (F28 repo half) #1100) correctly caught it. Reworded without the literals, with the self-obedience stated — the lint cannot tell a proof case from a violation, and that is the point.

Gate evidence: both suites green locally (28/28); typecheck clean. The third CI red in #1109's run (skills_integrity's second test) is the same F28 finding — one fix covers both. The two amico-run profile_verb failures in the same window are the known deploy-lag family (machine-specific, CI-green — not this PR's concern).

Process note, on the record: #1109 was merged on a partial CI view (the long-pole lanes green in the tail; the run conclusion checked after the merge — backwards). The run-conclusion check now happens BEFORE merge, every time. The skills_integrity failures are the skills campaign's surface — fixed here minimally to unblock main; the owning agent may rework the doctrine further.

…in skips the host check by design) + the F28 doctrine obeys its own rule (#1109 follow-up)

CI's ubuntu runner exposed two darwin-blind spots merged in #1109's window:
(1) the installer fleet-branch test asserted darwin's terminal outcome
(guard not installed, exit 1) — on non-darwin the installer deliberately
skips the host check ('the fleet is a darwin fleet') and completes green;
the test now asserts the BRANCH (parsed role, never the standalone skip)
and accepts the platform's correct terminal outcome. (2) amico-vault's
F28 proof case embedded the literal rotted vault names — the doctrine
violated its own 'never in prose or examples' rule and its own mechanical
lint caught it; reworded, with the self-obedience stated.
@aarontrowbridge aarontrowbridge added the hitl Needs human review before merge label Sep 14, 2026
@coderabbitai

coderabbitai Bot commented Sep 14, 2026

Copy link
Copy Markdown

Important

  • 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Advanced

Run ID: a42d9430-5f50-4778-b91b-5b63a9860840


Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

Comment @coderabbitai help to get the list of available commands.

@aarontrowbridge
aarontrowbridge merged commit c6144b3 into main Sep 14, 2026
11 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

hitl Needs human review before merge

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant