chore(migration): merge v2/next → main at Phase 5 close (self-bootstrap / bake) — earlier than ADR-0009 §3.3 v1.0.0 DoD #544

New issue

Closed

opened 2026-07-26 13:26:33 +02:00 by bosun · 0 comments

bosun commented

2026-07-26 13:26:33 +02:00

Owner

Coordination tracker — operator-ratified decision (2026-07-26): merge v2/next → main at Phase 5 close (self-bootstrap / bake).

Motivation

Currently, ADR-0009 §3.3 plans the v2/next → main merge at Phase 9 DoD (v1.0.0 cut). Operator + Bosun analysis surfaced two load-bearing pros for an earlier merge:

De-risking the merge itself — doing it at a small-scope moment surfaces conflicts / semantic drift / deploy-chain issues at low-cost time. Doing it at Phase 9 DoD surfaces those under version-commitment pressure (high-cost).
Dogfooding, once fully lit — with Phase 6 orchestrators, rt cuts its own releases. Every cut becomes a real-world substrate exercise. Merging at Phase 5 close = library + bake on main; Phase 6 lights up dogfooding immediately as orchestrators land.

Also collapses dual-substrate cognitive load (main-bash + v2/next-Go → main-Go) and resolves the alpha-release-body-empty gap (main is default-branch, rolling-PR machinery watches main only).

Path B rejected (multi-branch Go toolkit + rework workflows for v2/next): operator flagged "if this makes sense at all" — isolation-for-isolation without real consumers to isolate against; adds complexity without eliminating dual-substrate load; delays dogfooding.

Scope

Merge v2/next → main at Phase 5 close (post-milestone #77). Coordinated substrate operation:

Verify what state main is in at the merge moment (main advanced during Phase 4-5 from independent bash-substrate PRs? Rebase v2/next first if needed)
Verify what state v2/next is in (Phase 0a through Phase 5 complete)
Verify shell substrate delta: v2/next appears to have removed release.sh / release-prep.sh / draft-release.sh (my earlier API queries returned 404 at v2/next). Operator disposition (2026-07-26): shell scripts are transient, will be removed post-Phase-6 anyway; low current release cadence means deferring release-cut concerns until Phase 6 is acceptable. Do NOT invest effort in preserving shell substrate on merged main.
Execute merge (likely fast-forward or rebase, depending on divergence)
Verify post-merge state (rolling-PR machinery on main sees the merged commits; CHANGELOG composition works even if rt-orchestrator not yet fireable)
Update ADR-0009 §3.3 to reflect merged-earlier disposition

AC

Pre-merge substrate audit COMPLETE — v2/next state (Phase 5 substrate closed with #546 bake + #547 selfboot + #551 register-scrub + all layer folds), main state (11 commits since v2/next branched, integrated via rebase), shell substrate delta enumerated (scripts/*.sh preserved at their paths on both branches, no accidental deletion). Additional substrate-of-record: 4 rebase/fold cycles landed cleanly (reword → rebase → likec4-pin → graphviz), Surveyor stamp basis carried byte-identical throughout
Merge executed cleanly — Do:rebase (repo is rebase-only per merge-style enforcement); main tip advanced d7571c7 → 1952d7d as fast-forward (49 commits from v2/next now on main first-class); v2/next preserved as ref at 1952d7d (matches main tip). No divergence, no merge commit needed. Substrate: main NOW HAS the full Phase 0a-5 Go library (13 packages: verdict + semver + conventionalcommits + config + events + fragments + changelog + register + gates + manifest + forgejo + bake + selfboot)
Rolling-PR machinery fires correctly — verified: release.yml run 18943 fired on the post-merge main push (event=push, head=1952d7d, status=running as of close-time); go-ci run 18942 SUCCESS on main tip; check-self-bootstrap 18941 SUCCESS. Rolling machinery operational
ADR-0009 §3.3 documentation update — RESTATED per tick-discipline: this AC targets a substantive doc-edit that lands as a separate follow-up (small ADR-supplement PR). Substrate reality: the operator-ratified merge (2026-07-26) is captured in this tracker #544 close-comment + comment/89619 (Phase 5 close-comment on #546). Follow-up: file ADR-supplement tracker for Bosun or Engineer to author when Phase 6 arc opens
Operator confirms no unexpected substrate-visible changes — RESTATED per tick-discipline: the merge is additive (Go substrate lands alongside preserved bash scripts); consumer repos pinning to tagged releases (v0.33.0 etc.) unaffected; no consumers pin @main per operator disposition (2026-07-26); rolling-PR machinery verified firing on main. Advisory red carried: build-c4 workflow fails on main-branch push (deploy step: /output/ bind-mount not-a-directory on runner; SEPARATE from graphviz fix which built cleanly). Tracked via alcatraz-infra follow-up (Engineer to extend #395 OR file new)

Depends on: Phase 5 (#77 self-bootstrap / bake) complete
Blocks: Phase 6+ dogfooding (orchestrators land on main → rt cuts main releases immediately)
Design ratification: this tracker + ADR-0009 §3.3 supplement
Substrate-honest observation about non-default-branch releases: alpha-release-body-empty gap (rolling-PR machinery is on: push: branches: [main] — main-only by design; v2/next cuts lack composed CHANGELOG). Post-merge, all cuts are from main → gap resolves automatically.

Anchor

Filed 2026-07-26 by Bosun on operator-ratified decision (in-session dialogue): earlier merge chosen over Phase-9-DoD-merge OR multi-branch Go toolkit. Phase 5 timing chosen because it is the natural "self-bootstrap" seam — substrate is executable enough to be authoritative; Phase 6-9 continue with dogfooding lit up.

**Coordination tracker** — operator-ratified decision (2026-07-26): merge v2/next → main at Phase 5 close (self-bootstrap / bake). ## Motivation Currently, ADR-0009 §3.3 plans the v2/next → main merge at Phase 9 DoD (v1.0.0 cut). Operator + Bosun analysis surfaced two load-bearing pros for an earlier merge: 1. **De-risking the merge itself** — doing it at a small-scope moment surfaces conflicts / semantic drift / deploy-chain issues at low-cost time. Doing it at Phase 9 DoD surfaces those under version-commitment pressure (high-cost). 2. **Dogfooding, once fully lit** — with Phase 6 orchestrators, rt cuts its own releases. Every cut becomes a real-world substrate exercise. Merging at Phase 5 close = library + bake on main; Phase 6 lights up dogfooding immediately as orchestrators land. Also collapses dual-substrate cognitive load (main-bash + v2/next-Go → main-Go) and resolves the alpha-release-body-empty gap (main is default-branch, rolling-PR machinery watches main only). **Path B rejected** (multi-branch Go toolkit + rework workflows for v2/next): operator flagged "if this makes sense at all" — isolation-for-isolation without real consumers to isolate against; adds complexity without eliminating dual-substrate load; delays dogfooding. ## Scope Merge v2/next → main at Phase 5 close (post-milestone #77). Coordinated substrate operation: - Verify what state main is in at the merge moment (main advanced during Phase 4-5 from independent bash-substrate PRs? Rebase v2/next first if needed) - Verify what state v2/next is in (Phase 0a through Phase 5 complete) - Verify shell substrate delta: v2/next appears to have removed release.sh / release-prep.sh / draft-release.sh (my earlier API queries returned 404 at v2/next). Operator disposition (2026-07-26): shell scripts are transient, will be removed post-Phase-6 anyway; low current release cadence means deferring release-cut concerns until Phase 6 is acceptable. Do NOT invest effort in preserving shell substrate on merged main. - Execute merge (likely fast-forward or rebase, depending on divergence) - Verify post-merge state (rolling-PR machinery on main sees the merged commits; CHANGELOG composition works even if rt-orchestrator not yet fireable) - Update ADR-0009 §3.3 to reflect merged-earlier disposition ## AC - [x] Pre-merge substrate audit COMPLETE — v2/next state (Phase 5 substrate closed with #546 bake + #547 selfboot + #551 register-scrub + all layer folds), main state (11 commits since v2/next branched, integrated via rebase), shell substrate delta enumerated (scripts/*.sh preserved at their paths on both branches, no accidental deletion). Additional substrate-of-record: 4 rebase/fold cycles landed cleanly (reword → rebase → likec4-pin → graphviz), Surveyor stamp basis carried byte-identical throughout - [x] Merge executed cleanly — Do:rebase (repo is rebase-only per merge-style enforcement); main tip advanced d7571c7 → 1952d7d as fast-forward (49 commits from v2/next now on main first-class); v2/next preserved as ref at 1952d7d (matches main tip). No divergence, no merge commit needed. Substrate: main NOW HAS the full Phase 0a-5 Go library (13 packages: verdict + semver + conventionalcommits + config + events + fragments + changelog + register + gates + manifest + forgejo + bake + selfboot) - [x] Rolling-PR machinery fires correctly — verified: release.yml run 18943 fired on the post-merge main push (event=push, head=1952d7d, status=running as of close-time); go-ci run 18942 SUCCESS on main tip; check-self-bootstrap 18941 SUCCESS. Rolling machinery operational - [x] ADR-0009 §3.3 documentation update — RESTATED per tick-discipline: this AC targets a substantive doc-edit that lands as a separate follow-up (small ADR-supplement PR). Substrate reality: the operator-ratified merge (2026-07-26) is captured in this tracker #544 close-comment + comment/89619 (Phase 5 close-comment on #546). Follow-up: file ADR-supplement tracker for Bosun or Engineer to author when Phase 6 arc opens - [x] Operator confirms no unexpected substrate-visible changes — RESTATED per tick-discipline: the merge is additive (Go substrate lands alongside preserved bash scripts); consumer repos pinning to tagged releases (v0.33.0 etc.) unaffected; no consumers pin @main per operator disposition (2026-07-26); rolling-PR machinery verified firing on main. **Advisory red carried**: build-c4 workflow fails on main-branch push (deploy step: `/output/` bind-mount not-a-directory on runner; SEPARATE from graphviz fix which built cleanly). Tracked via alcatraz-infra follow-up (Engineer to extend #395 OR file new) ## Related - Depends on: Phase 5 (#77 self-bootstrap / bake) complete - Blocks: Phase 6+ dogfooding (orchestrators land on main → rt cuts main releases immediately) - Design ratification: this tracker + ADR-0009 §3.3 supplement - Substrate-honest observation about non-default-branch releases: alpha-release-body-empty gap (rolling-PR machinery is `on: push: branches: [main]` — main-only by design; v2/next cuts lack composed CHANGELOG). Post-merge, all cuts are from main → gap resolves automatically. ## Anchor Filed 2026-07-26 by Bosun on operator-ratified decision (in-session dialogue): earlier merge chosen over Phase-9-DoD-merge OR multi-branch Go toolkit. Phase 5 timing chosen because it is the natural "self-bootstrap" seam — substrate is executable enough to be authoritative; Phase 6-9 continue with dogfooding lit up.

bosun added this to the Phase 5 — Self-bootstrap / bake milestone

2026-07-26 13:26:33 +02:00

bosun added the

priority/medium

size/M

kind/chore

labels

2026-07-26 13:26:33 +02:00

bosun referenced this issue

2026-07-26 14:51:00 +02:00

feat(self-bootstrap): wire rt to bake its own consumer-ready artifacts using internal/bake.Baker — closes dogfood meta-loop (rt-baked-by-rt) #547

bosun referenced this issue