Section 05: Orphan Cleanup

Status: Not Started Goal: Remove three orphan entries whose classification and non-use are confirmed by Pass 1 research. Each deletion is independent (no cross-dependencies) and low-risk (none of these files are referenced by running code, tests, or CI). The tools/ori-lsp/ directory — the other major orphan — is explicitly NOT handled here because its deletion requires atomic updates to 20+ referencing files across Cargo.toml, release workflows, LSP design docs, version sync scripts, and Tier-3 doc rewrites. That work lives in §06.

Success Criteria:

samples/ removed from filesystem (if tracked) or confirmed as gitignored-and-absent
compiler/ori_lsp/.gitkeep removed from git tracking
compiler/ori_lsp/ directory may or may not exist on the filesystem after the deletion — both states are acceptable. The canonical LSP reservation is preserved by plans/roadmap/section-22-tooling.md:150-156 (a plan-level commitment), NOT by a filesystem artifact. When a real LSP crate is implemented later, mkdir -p compiler/ori_lsp/ + cargo new --lib . will recreate the directory with actual content.
rebuild-playground.sh removed from git tracking and filesystem
rg 'rebuild-playground' ori_lang/ ../ori-lang-website/ returns zero matches
./test-all.sh green post-cleanup
Single atomic commit: chore(plan): delete orphan directories and obsolete rebuild-playground.sh

Context: The Pass 1 research classified these three entries as orphans with the following evidence:

samples/: The Pass 1 census shows samples/ as an empty directory with 0 tracked files. .gitignore already contains /samples/ (line 24), so even if someone created files in it, they wouldn’t be tracked. The directory’s existence at all is leftover from an earlier planning state where “samples” was intended but never populated.
compiler/ori_lsp/.gitkeep: The Pass 1 research and §06 LSP disposition consensus both confirmed that compiler/ori_lsp/ is a phantom directory — its only tracked file is .gitkeep. The roadmap at plans/roadmap/section-22-tooling.md:150-156 declares compiler/ori_lsp/ as the canonical future home for the LSP, but the directory itself is empty. We keep the directory reservation (the “canonical future home” claim remains valid), but we remove the .gitkeep because it’s not needed — the directory will be re-created (or stay in git’s tree-cache) when the real LSP implementation lands.
rebuild-playground.sh: Pass 1 Agent 1 verified zero references across the entire ori_lang repo AND the ori-lang-website repo. The script’s body references ./website/playground-wasm/, which no longer exists (the website extracted to a sibling repo and has its own scripts/rebuild-wasm.sh). Classification: fully obsolete dead code.

Reference implementations:

rustc (post-unification cleanup when librustc_* was split into rustc_*): removed stale .gitkeep files and empty placeholder directories in a small follow-up commit separate from the main refactor
Swift (after cmake → swiftpm transition): removed stale shell scripts as chore: commits independent of the main migration

Depends on: §01 (the baseline must prove these three entries exist before we delete them, so mid-execution drift is detectable).

05.1 Delete samples/ Directory

File(s): samples/ directory

Verify current state:
```
cd /home/eric/projects/ori_lang
ls -la samples/ 2>&1
git ls-files samples/ 2>&1
```
- ls: directory exists, probably empty
- git ls-files: empty (gitignored by /samples/ rule at .gitignore:24)
Decision: since samples/ is already gitignored and empty, the simplest action is rmdir on the filesystem:
```
rmdir samples 2>&1
# If rmdir fails (directory not empty), list contents first
```
- If directory has any files: investigate. The Pass 1 research said “empty” — any content is drift. Capture the content, decide whether to migrate (unlikely) or delete, then proceed.
- If rmdir succeeds: the directory is gone.

Verify samples/ is gone:

ls samples 2>&1 || echo "absent (expected)"

NOTE: the /samples/ rule in .gitignore:24 can stay. Removing it is not required — it’s a historical protection that remains correct (if someone creates samples/ in the future, it stays gitignored unless the rule is removed). Leave the rule as-is to avoid gratuitous churn.
Subsection close-out (05.1) — MANDATORY before starting 05.2:
- samples/ directory absent
- .gitignore rule /samples/ left untouched (not needed to remove, not harmful)
- Update this subsection’s status in section frontmatter to complete
- Run /improve-tooling retrospectively on THIS subsection — reflect on the trivially-simple operation. rmdir on an empty directory is about as basic as it gets — no tooling gap possible. Document: “Retrospective 05.1: trivial operation; no tooling gap.”
- Run /sync-claude on THIS subsection — check whether code changes invalidated any CLAUDE.md, .claude/rules/*.md, or canon.md claims. If no API/command/phase changes, document briefly. Fix any drift NOW.

05.2 Delete compiler/ori_lsp/.gitkeep Phantom

File(s): compiler/ori_lsp/.gitkeep

This deletes the phantom .gitkeep file but PRESERVES the directory reservation (the directory stays empty but exists in the filesystem as a reminder that compiler/ori_lsp/ is the canonical future LSP home).

Actually — reconsider. A directory in git is only “tracked” if it has at least one tracked file. Once we git rm the .gitkeep, the directory becomes untracked-and-empty, and depending on workflow expectations, it might get rmdir’d automatically by git clean or similar. The “directory reservation” is really a plan-level claim, not a filesystem-level claim — the roadmap at plans/roadmap/section-22-tooling.md:150-156 already documents the intent, so the filesystem directory doesn’t need to exist to preserve the reservation.

Decision: delete the .gitkeep AND let the directory be untracked. Anyone who wants to populate the canonical LSP home in the future will mkdir -p compiler/ori_lsp/ && cargo new --lib . or equivalent. The roadmap claim is the load-bearing reservation, not the empty dir.

Verify current state:

cd /home/eric/projects/ori_lang
git ls-files compiler/ori_lsp/
# Expected: compiler/ori_lsp/.gitkeep (just the one file)

ls -la compiler/ori_lsp/
# Expected: only .gitkeep

If anything else is tracked under compiler/ori_lsp/: STOP. That’s unexpected drift — §06 LSP disposition may need to change. Investigate before deletion.
Delete the .gitkeep:
```
git rm compiler/ori_lsp/.gitkeep
```

Optionally remove the now-empty directory (git does not track empty directories anyway):

rmdir compiler/ori_lsp/ 2>&1 || true
# This may fail if the directory contains untracked files; that's fine

Verify:

git ls-files compiler/ori_lsp/
# Expected: empty

Subsection close-out (05.2) — MANDATORY before starting 05.3:
- .gitkeep removed from tracking
- Directory may or may not exist on disk (either is fine)
- §06 LSP disposition section updated if it referenced the .gitkeep deletion (cross-section sync — §06 may have been written assuming we keep the phantom directory; verify §06’s compiler/ori_lsp/ treatment is consistent with this subsection’s outcome)
- Update this subsection’s status in section frontmatter to complete
- Run /improve-tooling retrospectively on THIS subsection — reflect on whether the “delete phantom .gitkeep” pattern is something future reorgs will encounter. Phantom directories are common enough that a scripts/dev/find-phantom-dirs.sh helper (find directories whose only tracked content is .gitkeep or similar placeholders) might have value. Is this a one-off or a recurring pattern? Probably recurring across any compiler with reserved-but-unpopulated future homes. If worth building: commit via build(scripts): add find-phantom-dirs.sh — surfaced by project-reorganization/section-05.2 retrospective. If not: “Retrospective 05.2: one-time phantom cleanup; no reusable tooling warranted.”
- Run /sync-claude on THIS subsection — check whether code changes invalidated any CLAUDE.md, .claude/rules/*.md, or canon.md claims. If no API/command/phase changes, document briefly. Fix any drift NOW.

05.3 Delete rebuild-playground.sh

File(s): rebuild-playground.sh

Re-verify zero references (Pass 1 already did this, but we re-verify at execution time per §01.3 discipline):
```
cd /home/eric/projects/ori_lang
rg -n 'rebuild-playground' . ../ori-lang-website/ 2>&1 \
  | grep -v 'plans/project-reorganization/' \
  | grep -v 'rebuild-playground\.sh:'
# Expected: empty output (no external references)
```
- No matches outside this plan directory and the script itself
- If matches appear: STOP. A reference has been added since Pass 1. Reclassify and DO NOT delete.
Quick-read the script to confirm it’s the obsolete one (references ./website/playground-wasm/ that no longer exists):
```
head -20 rebuild-playground.sh
```
- Content confirms the ./website/ assumption
Delete:
```
git rm rebuild-playground.sh
```

Commit §05 as a single atomic chore commit:

cd /home/eric/projects/ori_lang
git commit -m "chore(plan): delete orphan directories and obsolete rebuild-playground.sh

Three orphans removed:

samples/ (empty, gitignored, leftover placeholder)
compiler/ori_lsp/.gitkeep (phantom — directory reservation lives in plans/roadmap/section-22-tooling.md:150-156, not in the filesystem)
rebuild-playground.sh (obsolete — 0 references, dead ./website/ path; ori-lang-website has scripts/rebuild-wasm.sh as canonical)

Pass 1 research confirmed zero active uses for each entry.

tools/ori-lsp/ is NOT deleted here — it has its own atomic-deletion section at §06 (touches 20+ referencing files across Cargo.toml, release workflows, design docs, version sync).

Refs: plans/project-reorganization/section-05-orphan-cleanup.md”


- [ ] Verify the commit:
```bash
git log --oneline -1
git show --stat HEAD
# Expected stat: ~2 deletions (rebuild-playground.sh + .gitkeep); samples/ was already gitignored-empty so no git op

Run ./test-all.sh to prove no regression (rebuild-playground.sh had no references, so test-all cannot know about the deletion — this is a sanity check):
```
timeout 150 ./test-all.sh 2>&1 | tail -10
```
- Exit 0, all phases PASS
Subsection close-out (05.3) — MANDATORY before 05.R:
- rebuild-playground.sh absent from git and filesystem
- Commit landed atomically
- ./test-all.sh green
- Update this subsection’s status in section frontmatter to complete
- Run /improve-tooling retrospectively on THIS subsection — reflect on the “verify zero refs → delete” pattern. Is this the same pattern §02.3 used (global sweep verification)? If so, and if the plan-ref-sweep.sh helper from §02.3’s retrospective was built, is it applicable here? Unify the approach if possible. If no helper was built: “Retrospective 05.3: the grep-filter-investigate pattern was already documented in 02.3 retrospective; no additional tooling needed.”
- Run /sync-claude on THIS subsection — check whether code changes invalidated any CLAUDE.md, .claude/rules/*.md, or canon.md claims. If no API/command/phase changes, document briefly. Fix any drift NOW.

05.R Third Party Review Findings

[TPR-XX-004-codex][medium] plans/project-reorganization/section-05-orphan-cleanup.md:6 — Reconcile the DRIFT in §05’s compiler/ori_lsp reservation contract. Evidence: §05 gave two incompatible contracts for the compiler/ori_lsp/.gitkeep cleanup. The frontmatter goal said “keeping the directory as a reserved canonical future home” (line 6), and the section-level success criterion said “directory itself stays reserved as future canonical LSP home” (line 46). But §05.2’s subsection correctly re-evaluated the mechanics (lines 114-149) and concluded that the LSP reservation is a plan-level claim at plans/roadmap/section-22-tooling.md:150-156, NOT a filesystem artifact — and that deleting .gitkeep AND letting the empty directory disappear is the correct architectural choice. Impact: The section was internally contradictory. An executor following the subsection’s corrected decision would still fail the section’s own frontmatter and top-level success criteria. Required plan update: Align §05’s frontmatter goal, success_criteria, and section-level success criteria with the §05.2 decision (delete .gitkeep, allow the empty directory to disappear, preserve the canonical LSP reservation via the roadmap commitment). Basis: fresh_verification. Confidence: high. Resolved: Fixed on 2026-04-11 iteration 2. Updated §05 frontmatter goal to explicitly say “let the now-empty directory disappear from the filesystem since git doesn’t track empty dirs; the canonical LSP reservation is preserved by plans/roadmap/section-22-tooling.md:150-156, not by an empty filesystem directory”. Updated frontmatter success_criteria row to match. Updated section-level success criterion to say “directory ITSELF may or may not exist on the filesystem after the deletion — both states are acceptable”. All three places (frontmatter goal, frontmatter success_criteria, body success_criteria) now tell the same story as §05.2’s implementation.

05.N Completion Checklist

Exit Criteria: Three orphan entries are gone from the repository. ls /home/eric/projects/ori_lang/ no longer shows samples/ or rebuild-playground.sh. git ls-files compiler/ori_lsp/ is empty. Zero references to any deleted entry survive in the repo (verified by post-commit rg). A single chore(plan): commit documents the cleanup. ./test-all.sh green. §06 is unblocked — it can now focus exclusively on tools/ori-lsp/ without also handling the phantom .gitkeep.