Public: anyone can read this; nobody can change it.

seed-release: jq → node; reseed no-op guard ignores mitosis self-ref #131

Closed opened this view on origin
Closed without merging

Two robustness fixes from watching the merged flow run: seed-release's first run (on the r278 merge) died on a bare jq — the isolated runner image ships node/bun but no jq (the same trap reseed-run.sh documents for the vxrail host). JSON payload/parse now uses node -e. No behavior change otherwise.

Notes

stopped closed without shipping

opened this view on origin

Two robustness fixes from watching the merged flow run:

  1. seed-release's first run (on the r278 merge) died on a bare jq — the isolated runner image ships node/bun but no jq (the same trap reseed-run.sh documents for the vxrail host). JSON payload/parse now uses node -e. No behavior change otherwise.
  2. The reseed no-op guard now excludes mitosis's own ref from the refs comparison: every reseed merge moves mitosis main by exactly one genesis commit, so including it makes the nightly cron see "drift" forever and open a self-referential churn PR every day. With the exclusion, the cron is quiet unless a sibling system repo actually moved.

Validation on merge: the open r286 reseed PR's eventual merge fires the fixed seed-release → expect tag seed-r286, a Release with tarball+manifest attached, and the seed package auto-linked to this repo's Packages tab.

🤖 Generated with Claude Code

commented

Superseded by #266 (merged): seed-release is now jq-free (curl/sed/tr/tar only — the isolated runner image ships no jq, which was the root cause of every failure, #263), tags come from genesis/upstream.json .release, API calls go in-cluster, and both lanes are proven — dispatch created seed-r886, the r897 reseed push created seed-r897. Closing.

Sign in to join this conversation.

↑↓ navigate ↵ open ⌘↵ its app esc close

Keyboard shortcuts

Docs