release: promote develop to main (8debf013) — the runtime images stop shipping their build tree - #626
Merged
Merged
Conversation
sync: carry the maturity-evidence refresh from main back into develop
…at moved (#621) auditClientIngestion read `rulesets/schema`, a path that has not existed since rulesets/ moved under src/ (ADR-0048). exists() turned the moved directory into a silent zero, so the audit reported ZERO client schemas over a directory holding 50. Measured before and after on the same tree: Schemas 0 -> 50, the dimension's verdict PARCIAL -> SÓLIDO, and the global score 8/9 -> 9/9 dimensiones SÓLIDO. The score was wrong in the pessimistic direction. Salvaged from a commit that lived only in a stale local main and was about to be discarded. The narrative rewrite that travelled with it is deliberately not included — that is a claim about the product, not a defect fix.
…oubled segment (#622) auditSdlc read `reference/core/sdlc/sdlc` — a doubled segment. That path has never existed; the SDLC surface is at reference/core/sdlc. exists() turned it into a silent zero, so the dimension was scored over no files at all. Measured before and after: Archivos SDLC 0 -> 174, datos estructurados 5 -> 7, playbooks 0 -> 22. Unlike #621 this moves no verdict: the dimension was already SÓLIDO on Fases 5/5 and Gates 5/5, computed elsewhere. That is why it survived — the zero never hurt the score, so an evidence line reading 'Archivos SDLC: 0' sat next to a green verdict. A wrong number that costs nothing is the one nobody reports. Left out on purpose, same function: '0 markdown fases' is still 0, because the filter matches phase-0[1-5] while the files are phase-1, phase-1.1, phase-2, phase-3, phase-4. Whether phase-1.1 counts as a phase is a question about the model, not about a regex.
… the Core (#623) Advances GT-708 — the Tracker half is beyondnetcode/evolith_tracker#153. Two things travelled under three letters and neither was ever built: Phase 1.1 (an optional subphase with its own readiness gate and seven artifact templates) and KDD = Knowledge-Driven Development (an optional section INSIDE the PRD, per-tenant, from the 2026-07-04 owner session). Owner decision: neither survives, in either repo. Measured before deleting anything: five gates for phases 1..5; none of the seven KDD artifacts among the 33 registered; zero TypeScript files; 31 CLI commands with zero mentions (--phase discovery maps to phase 1 ENTIRE); zero in MCP; no Tracker screen or entity; no KDD section in prd.schema.json. The prose was load-bearing anyway: phase-1-business-signoff made 'Phase 1.1 adoption level has been declared' a PRECONDITION for opening Gate 1, with 'a FAIL result blocks this gate'. A gate nothing implements was blocking a gate everything implements. 16 files deleted, the Gate 1 precondition and its three KDD clauses removed, the Subphase 01.1 tables and index rows removed, D-004/L-009 rewritten to what survives them. ADR-0127 carries the retirement and AMENDS ADR-0103 rather than editing it. CHANGELOG.md and ADR-0103 keep their KDD text on purpose. Three pinned corpus counts move by one because the ADR itself owes a generated conformance ruleset — the corpus grows by one rule nothing can run, to record the removal of a concept nothing could run either.
…624) The row was left IN-PROGRESS when the Core half merged, because one of its criteria was the Tracker's REQ-DIS-12/REQ-DIS-13. That criterion is met by evolith_tracker#153 (97e1bc8e), so the row closes. Both repositories had the same shape of teeth in different words: the Core made 'Phase 1.1 adoption level has been declared' a precondition for opening Gate 1, and the Tracker gave REQ-DIS-13 the acceptance criterion 'a FAIL result blocks opening the Business Sign-Off gate'. One non-existent gate, described twice, blocking on paper the only gate every initiative must pass — while no ruleset, schema, CLI command, MCP tool, screen or entity implemented it anywhere. Board: 677 / 706 done, 0 pending.
…ted it on the way out (#625) Advances GT-692 (DEFERRED -> IN-PROGRESS). The Tracker's Deploy (kind + Helm + smoke) died importing evolith-core-api with 'no space left on device' on /repo/node_modules/get-intrinsic/CHANGELOG.md — the same shape this row registered in July, still biting in another repository. Two causes, and the second was not in the original evidence: 1. The runner received the build tree, compiler included. Fixed with 'npm prune --omit=dev' at the end of every builder stage. 2. Found by reading docker history rather than the Dockerfile: 'RUN … chown -R' is a 586 MB layer on core-api alone, a byte-for-byte duplicate of everything copied above it, because a recursive chown rewrites every file into a new layer. Ownership now travels on COPY --chown. Measured before and after, same tree and same day: core-api 1.96 GB -> 862 MB, agent-runtime-api 2.49 GB -> 1.13 GB, mcp-server 1.89 GB -> 825 MB, cli 2.00 GB -> 890 MB. Total 8.34 GB -> 3.71 GB. Every image was BOOTED, not merely built — and that is what earned the second finding: mcp-server first died with 'Cannot find module keyv', a runtime dependency of @nestjs/cache-manager it never declared, surviving only because eslint hoisted a copy. The prune exposed the defect; it did not create it. 70-validate-runtime-image-shape.mjs keeps both causes fixed and was observed red against the previous Dockerfile, naming line 87. Merged with the non-required 'Governance guards' job stuck ~26 min on an npm install step — the same flakiness seen twice today. All 8 required checks are green.
|
You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard. |
📊 Bilingual Coverage ImpactPR Changes
Repository Coverage
✅ Good: All EN changes have ES counterparts. Generated by GitHub Actions |
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.
Promotes
developtomain. Six commits; two of them matter.8.34 GB → 3.71 GBtotal. This is the fix for the Tracker'sDeploy (kind + Helm + smoke), which has been dying onno space left on deviceimportingevolith-core-api.ADR-0127records it.Schemas: 0over 50 files) and the SDLC scan at a doubled segment (Archivos SDLC: 0over 174).Why this promotion is the one that unblocks another repository
The image sizes are the point. Measured before and after, same tree and same day:
core-apiagent-runtime-apimcp-servercliHalf of the saving was not the dependency prune.
RUN … chown -Rwas a 586 MB layer oncore-apialone — a recursive chown rewrites every file into a new layer — and that cause was not inGT-692's original evidence. Ownership now travels onCOPY --chown.Every image was booted, not merely built, and that is what exposed
mcp-server'sCannot find module 'keyv': a runtime dependency of@nestjs/cache-managerit never declared, surviving only becauseeslinthoisted a copy. The prune found the defect; it did not create it.After this merges
A
v*tag is what publishes the images to GHCR, so the Tracker keeps importing the 1.91 GB one until a tag is cut. That is the next step, and it closesGT-692's fourth criterion — re-running the consumer's job against the new image.Board
677 / 706 done · 0 pending · 3 in progress · 26 deferred.
🤖 Generated with Claude Code