* feat(ship): pause /ship on a red base instead of failing (gstack-base-red, conditional receipts) (GSTA-157 PR 1) Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com> * fix: base-red shell runner launches bash through runBashSync Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com> * test: base-red coverage additions (unreproduced head, unresolved head, readPause, wait unconfigured) Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com> * docs: base-red pause in unattended endings; BASE_RED_UNCONFIGURED covers an invalid base policy Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com> * test: base-red coverage pass 2 (write --base-red, summary line, classify usage before fetch, unresolved-head detail) Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com> * feat(ship-queue): merge trains for any repo — train.gates (workflows/command), every prefix gated, generalized train land (direct/update, merge method), base-red trains, train init/doctor, watch --land (GSTA-157 PR 2) Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com> * fix: pre-landing review fixes (train gates: dispatch bound, lease, rebase, post-merge checks, output tail, caps) Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com> * fix: train land refuses the next slot until the previous slot's post-merge checks are green Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com> * docs: ship-policy, troubleshooting and captain for train gates review fixes Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com> * chore: changelog fragment for merge trains for any repo Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com> * test: keep ship-queue-train-generic out of the Windows subset (gh shim resolves gh.exe there) Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com> --------- Co-authored-by: capy-ai[bot] <230910855+capy-ai[bot]@users.noreply.github.com>
2.5 KiB
Merge train for a peer repository
For a crew whose repository has its own CI and no gstack eval workflow (gbrain, an app repository). Copy it into the crew's NOTES.md or CREW.md and fill the angle brackets. The rules are docs/ship-policy.md#train-gates; this file is the order to do them in.
Turn it on (once)
- In a checkout of
<REPO>on<BASE>, with gstack installed:gstack-ship-queue train init --base <BASE>, adding--command "<test command>" --spend-cap <usd>when a VM-run test command should gate too (it runs through scripts/ubicloud/ubi-runner.sh and needsUBI_OWNER). - If it refuses with
QUEUE_TRAIN_INIT_REFUSED, PRs here bump the version: adoptchangelog_fragmentsfirst (docs/ship-policy.md#changelog-fragments). - Open one PR that adds the printed
trainblock andmin_gstack_versionto.gstack/ship-policy.json, and the printednonceinput, run-name and concurrency lines to each gate workflow. Fix everyWARN:line in the same PR (a push trigger that matchestrain-prefix/refs getsbranches-ignore: ['train-prefix/**']). - After it merges:
gstack-ship-queue train doctor --base <BASE>until it printsDOCTOR: no findings.
Parameters for CREW.md
| Parameter | Value |
|---|---|
GATES |
<workflow>/<job> list from the block, plus the command gate if any |
REQUIRED_CHECKS |
train.land.required_checks (or changelog_fragments.required_checks) |
MERGE_METHOD |
train.land.merge_method (merge when history is merge-only) |
LAND_MODE |
what init printed: direct when protection is not strict, else update |
COMMAND_SPEND_CAP |
the command gate's spend_cap_usd, per train |
Lanes
READY with gstack-ship-queue ready --pr <n> --no-screen after /ship at the pinned head.
There is no screen and no slot call. A lane whose /ship paused on a red base (its receipt
carries base_red) READYs with --base-red --fixer <fixer PR> once the fixer lane READYs
with --fixes-base-red <case,...>; it never merges the base or waits by hand.
Controller loop
gstack-ship-queue train build --prs <n>@<sha>,... --base <BASE>
gstack-ship-queue train watch --train <nonce> --land --base <BASE>
watch --land waits for every prefix run, reads the verdict and lands each landable slot
in order, then prints NEXT. A culprit (QUEUE_TRAIN_RED) goes back to its lane and the rest
ride the next train; red on main (QUEUE_TRAIN_RED_ON_MAIN) is fixed forward;
QUEUE_TRAIN_UNATTRIBUTED means repair or a new order, never the same train again.