fix(worktree): survive the renderer-era bmad-dev-auto (BMAD-METHOD #2587/#2588)#155
Open
pbean wants to merge 1 commit into
Open
fix(worktree): survive the renderer-era bmad-dev-auto (BMAD-METHOD #2587/#2588)#155pbean wants to merge 1 commit into
pbean wants to merge 1 commit into
Conversation
…587/#2588)
Upstream bmad-dev-auto's SKILL.md is now a two-line shim that runs
`uv run render.py` and follows its stdout; the renderer anchors its
project root on the first _bmad/ dir walking UP from the session cwd,
requires the post-#2285 _bmad/config.toml, and rewrites
_bmad/render/bmad-dev-auto/ with absolute paths baked in. The completion
contract bmad-loop parses is unchanged — the breakage is environmental:
- Worktrees nest under <project>/.bmad-loop/runs/ and never carried a
_bmad/, so render.py silently adopted the MAIN checkout's: rendered
workflows baked main-checkout absolute artifact paths (deferred-work
appends, no-spec fallback artifacts, and step-01 context escaped the
worktree — fallbacks landing where _artifact_dirs never looks) and
every parallel session raced on one shared, delete-rewritten render
dir. provision_worktree now merge-seeds the main repo's _bmad/ config
surface per-file (copy-when-absent: a committed _bmad/ keeps every
tracked file; the gitignored *.user.toml layers — whose absence
hard-HALTs render.py since #2588 — are filled in), never seeds the
regenerated render/, and shields it: one anchored /_bmad exclude when
the checkout had none, per-file rels when merging into an existing
tree.
- /_bmad/render/ is always excluded (seeded or not) so session-rendered
output can't be swept into story commits by the unit's `git add -A`;
`bmad-loop init` also gitignores _bmad/render/ for in-place runs.
- The shim HALTs on any failure BEFORE the workflow's HALT protocol is
loaded, so a broken environment (no uv, no central config, missing
render.py) yields one result-less Stop per story with no spec status
ever written. A renderer preflight (validate + run/sweep/resume start)
now fails loudly instead — gated on the installed SKILL.md actually
being the render shim, so pre-render installs see zero new checks and
byte-stable validate output.
- `validate --render-probe` (opt-in) executes the real renderer against
the project to surface render-time HALTs (#2588: missing {{.var}}
keys, unparseable override layers) before a run burns a story.
- validate warns (never fails) when the four-layer central TOML and the
legacy _bmad/bmm/config.yaml disagree on implementation_artifacts
(the renderer follows the TOML, the orchestrator reads the yaml —
TOML-first migration tracked in #154), and once when a project has
already committed _bmad/render/** (excludes can't shield tracked
files; `git rm -r --cached` remediation printed).
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Pro Run ID: 📒 Files selected for processing (11)
WalkthroughRenderer-era preflight, render probing, TOML artifact parity warnings, and worktree ChangesRenderer validation and artifact parity
Worktree configuration seeding and render shielding
Estimated code review effort: 4 (Complex) | ~45 minutes Sequence Diagram(s)sequenceDiagram
participant Validator as bmad-loop validate
participant Support as missing_renderer_support
participant UV as uv
participant Renderer as render.py
Validator->>Support: check render.py, uv, and _bmad/config.toml
Validator->>UV: execute uv run render.py
UV->>Renderer: start renderer probe
Renderer-->>Validator: dispatch output or HALT details
Possibly related issues
Suggested reviewers: Poem
🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
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. Comment |
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.
What upstream changed
BMAD-METHOD 64157d39 (#2587) made
bmad-dev-auto's SKILL.md a two-line shim —uv run {skill-root}/render.py, follow its stdout; on ANY failure "report what it printed and HALT". The old SKILL.md body moved toworkflow.md, rendered into{project-root}/_bmad/render/bmad-dev-auto/with config values and absolute paths baked in at render time.render.pyfinds{project-root}by walking UP from the session cwd to the first_bmad/directory and requires the post-#2285 central_bmad/config.toml. Follow-up 1cd4a7f5 (#2588, upstream HEAD) hardened it: a{{.var}}referenced by the templates but missing from the merged config, or a present-but-unparseable customization layer, now hard-HALTs.The completion contract bmad-loop parses is unchanged (verified against the diff):
## Auto Run Result,Status:line, frontmatter statuses,baseline_revision, thebmad-dev-auto-result-fallback prefix, folder+id dispatch, and the"folder+id dispatch"stories-probe literal all survive; every preflight marker file still exists.devcontract.pyis untouched. The breakage is environmental, in two critical shapes:<project>/.bmad-loop/runs/<id>/worktrees/…and never carried a_bmad/(defaultworktree_seed = []), so render.py silently adopted the main checkout's_bmad/: rendered workflows baked main-checkout absolute artifact paths (deferred-work appends, no-spec fallback artifacts, and step-01 context all escaped the worktree — fallback artifacts landing where_artifact_dirsnever looks), and every parallel session raced on one shared render dir that render.py delete-rewrites per invocation.workflow.md's HALT protocol is ever loaded, so a broken environment (nouv, no_bmad/config.toml, missingrender.py, or a #2588 config HALT) writes no spec status at all — every story becomes a result-less Stop (the cascade fix(adapters): bound dev/review stall nudges + diagnose result-less Stops #151's breadcrumbs diagnose, one story at a time).What this PR does
_bmadmerge-seed inprovision_worktree: per-FILE copy-when-absent of the main repo's_bmad/into the worktree (a committed_bmad/keeps every tracked file; the gitignored*.user.tomllayers — whose absence hard-HALTs render.py since #2588 — are filled in), never seeding the regeneratedrender/. Shield rels follow the existing "shield exactly what we wrote" doctrine: one anchored/_bmadwhen the checkout had none, per-file rels when merging./_bmad/render/always excluded (git-common-dirinfo/exclude), so session-rendered output (machine-absolute paths baked in) can't be swept into story commits by the unit'sgit add -A;bmad-loop initalso gitignores_bmad/render/for in-place (isolation = "none") runs.install.missing_renderer_support) invalidateand at run/sweep/resume start:render.pynext to the shim,uvon PATH,_bmad/config.tomlpresent. Gated on the installed SKILL.md actually containingrender.py— pre-render installs get zero new checks and byte-stable validate output.validate --render-probe(opt-in): executes the realuv run render.pyagainst the project; pass = theread and follow …dispatch line. Surfaces #2588 render-time HALTs before a run. Side effect documented in--help(rewrites the regenerated_bmad/render/bmad-dev-auto/cache).bmm/config.yamldesync onimplementation_artifacts(bmadconfig.toml_artifact_parity, mirroring render.py's exact merge precedence — deep-merge layers, then[modules.bmm]beats[core]; TOML-first migration tracked in bmadconfig: read the four-layer central TOML as primary, legacy bmm/config.yaml as fallback #154), and a one-time "you have_bmad/render/**committed —git rm -r --cachedit" hint (excludes can't shield tracked files).Verification
trunk checkclean. New coverage: seed/merge/shield/symlink-escape, renderer-support matrix (incl. uv-missing reported once across trees), probe pass/HALT via a fake-uv launcher (no test executes real uv — Windows CI safe), an end-to-end engine test proving a committed-_bmadproject's worktree gets the gitignored user layer seeded and the session's render residue never reaches the story commit, and TOML-parity precedence tests.validatedetects the shim and passes the preflight;validate --render-proberan the real renderer via uv and produced its dispatch line; the rendered step-04 confirms the absolute-path baking the seed protects against.Follow-ups
_bmad/render/; render.py's walk-up could stop at a git-worktree/repo boundary as defense-in-depth. Filed separately.Summary by CodeRabbit
New Features
render.py,uv, and centralized configuration.validate --render-probeto detect render-time failures and configuration issues.Bug Fixes
Documentation