feat(lib): surgical doc-commit helper + real-exec scope/exclusion tests

New lib/doc-commit.sh: surgical commit of ONLY the public-doc files doc-sync patched (passed as args), twin of memory-commit.sh with INVERSE scope. Δ1 dynamic pathspec filtered to changed paths (LRN-051); Δ2 fail-closed + LOUD scope guard rejecting .claude/** and CLAUDE.md (BDR-022) with a dedicated exit 4; Δ3 no hash anchoring (LRN-052); Δ4 `docs:` message. Hash-only on stdout for `doc_hash=$(...)` capture.

lib/tests/run-doc-commit.sh: 24 assertions, all REALLY EXECUTED on real git fixtures (no presumed behavior). T1a/b/c prove the guard CATCHES — forbidden path alone → exit 4, mixed legit+forbidden → refuse-all (nothing half-committed, offender named); T2 dynamic pathspec no-match filter; T3/T4 dangling + stale-index safety; T5/T6 idempotent + unsafe-state skip. shellcheck clean (both).

Part of the doc-sync coupled chantier (twin of BDR-034). Include + 2 orchestrator reorders follow.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ho5EQCFTSvYamuRtVZpp2d
This commit is contained in:
Bastien Chanot 2026-06-27 00:38:18 +02:00
parent bb341d3c87
commit ae1f218d3e
2 changed files with 311 additions and 0 deletions

147
lib/doc-commit.sh Executable file
View File

@ -0,0 +1,147 @@
#!/usr/bin/env bash
# doc-commit.sh — surgically commit ONLY the PUBLIC-DOC files doc-sync patched.
#
# Twin of memory-commit.sh, INVERSE scope: memory-commit TARGETS .claude/; this
# one commits public docs and must NEVER touch .claude/ or CLAUDE.md (BDR-022).
# Safety lives in the (dynamic) PATHSPEC + a fail-closed scope guard, never in a
# human diff review — automation removes that review, so the scope must be airtight.
#
# The scope guard is fail-CLOSED and LOUD: a forbidden path (.claude/** or
# CLAUDE.md) in the list is an UPSTREAM bug — doc-syncer must never patch those
# (BDR-022). Seeing one, ABORT THE WHOLE COMMIT and signal; do NOT silently filter
# it and commit the rest. A half-commit with no alert would MASK the violation.
# Caller passes EXACTLY the files doc-sync patched this run.
#
# Usage (CLI):
# doc-commit.sh pending <file>... # exit 0 if any passed file has changes, 1 if clean
# doc-commit.sh commit "<message>" <file>... # surgical commit
#
# Exit codes (commit): 0 ok/no-op · 2 usage · 3 unsafe git state · 4 scope violation.
# Output contract: diagnostics → stderr; on a real commit the short hash of the doc
# commit is the ONLY thing on stdout (empty on no-op/abort), so callers can capture
# it: doc_hash=$(doc-commit.sh commit "msg" README.md USAGE.md).
#
# Sourceable: docs_pending and commit_docs for the v2 hook.
set -uo pipefail
_in_git_repo() { git rev-parse --git-dir >/dev/null 2>&1; }
# True (0) when the repo is in a state where we must NOT auto-commit:
# detached HEAD, or a merge/rebase/cherry-pick in progress.
_unsafe_state() {
local gitdir
gitdir="$(git rev-parse --git-dir 2>/dev/null)" || return 0
if [ -e "$gitdir/MERGE_HEAD" ] || [ -e "$gitdir/rebase-merge" ] ||
[ -e "$gitdir/rebase-apply" ] || [ -e "$gitdir/CHERRY_PICK_HEAD" ]; then
return 0
fi
git symbolic-ref -q HEAD >/dev/null 2>&1 || return 0 # detached HEAD
return 1
}
# True (0) when a path is OUT OF SCOPE for a doc commit: anything under .claude/
# (any depth) or a CLAUDE.md (root or nested). These are doc-syncer's read-only
# context, never sync targets (BDR-022) — their presence is an upstream anomaly.
_forbidden_path() {
case "$1" in
.claude | .claude/* | */.claude/* | CLAUDE.md | */CLAUDE.md) return 0 ;;
*) return 1 ;;
esac
}
# Print every forbidden path in the argument list, one per line (empty = none).
_scope_violations() {
local p
for p in "$@"; do
_forbidden_path "$p" && printf '%s\n' "$p"
done
}
# Of the passed paths, those that EXIST and have real pending changes. A clean or
# absent path is dropped (not fatal): `git commit -- <no-match>` aborts the WHOLE
# commit, while `git add` tolerates it — so scope = only paths git would accept.
_changed_paths() {
local p
for p in "$@"; do
[ -e "$p" ] || continue
[ -n "$(git status --porcelain -- "$p" 2>/dev/null)" ] && printf '%s\n' "$p"
done
}
# 0 if any passed path has pending changes, 1 if all clean / absent.
docs_pending() {
_in_git_repo || return 1
local changed
mapfile -t changed < <(_changed_paths "$@")
[ "${#changed[@]}" -gt 0 ]
}
# Surgical commit of the passed doc paths only. Returns 0 (ok/no-op), 3 (unsafe),
# 4 (scope violation). On a real commit, prints the doc-commit short hash to stdout.
commit_docs() {
local msg="${1:?commit message required}"
shift
_in_git_repo || {
echo "doc-commit: not a git repo — skip" >&2
return 3
}
if _unsafe_state; then
echo "doc-commit: detached HEAD or merge/rebase in progress — skip (no commit)" >&2
return 3
fi
# FAIL-CLOSED scope guard. A forbidden path is an upstream BDR-022 violation
# (doc-syncer must never patch .claude/ or CLAUDE.md). Abort the WHOLE commit and
# name the offenders — never filter-and-commit-the-rest (that masks the bug).
local violations
mapfile -t violations < <(_scope_violations "$@")
if [ "${#violations[@]}" -gt 0 ]; then
{
echo "doc-commit: REFUSED — out-of-scope path(s) in the doc list (upstream BDR-022 violation):"
printf ' - %s\n' "${violations[@]}"
echo "doc-commit: NOTHING committed. doc-syncer must never patch .claude/ or CLAUDE.md —" \
"investigate why these surfaced before retrying."
} >&2
return 4
fi
local changed
mapfile -t changed < <(_changed_paths "$@")
if [ "${#changed[@]}" -eq 0 ]; then
echo "doc-commit: nothing pending — no-op" >&2
return 0
fi
# Re-stage working-tree content over any stale index entry, then commit ONLY
# those paths. The pathspec on `git commit` makes it partial: other staged files
# (dangling code) are not recorded.
git add -- "${changed[@]}"
if git diff --cached --quiet -- "${changed[@]}"; then
echo "doc-commit: only ignored/no-op changes — no-op" >&2
return 0
fi
git commit -q -m "$msg" -- "${changed[@]}"
printf 'doc-commit: committed %d file(s): %s\n' "${#changed[@]}" "${changed[*]}" >&2
git rev-parse --short HEAD
}
main() {
local cmd="${1:-}"
case "$cmd" in
pending)
shift
docs_pending "$@"
;;
commit)
shift
commit_docs "$@"
;;
*)
echo "usage: doc-commit.sh {pending <file>... | commit <message> <file>...}" >&2
return 2
;;
esac
}
# Run main only when executed, not when sourced.
if [ "${BASH_SOURCE[0]}" = "${0}" ]; then
main "$@"
fi

164
lib/tests/run-doc-commit.sh Executable file
View File

@ -0,0 +1,164 @@
#!/usr/bin/env bash
# Deterministic tests for lib/doc-commit.sh.
#
# Proves the contract on REAL git behavior (not assumed). Load-bearing deltas vs
# memory-commit, each tested by what it must REFUSE, not only what it accepts:
# T1 inverse-exclusion scope guard (BDR-022) — fail-CLOSED and LOUD:
# T1a forbidden path ALONE (.claude/ and CLAUDE.md) → exit 4, nothing committed
# T1b legit docs only → commits cleanly
# T1c MIXED legit + forbidden → exit 4, NOTHING committed (the trap)
# T2 dynamic pathspec — a clean passed path is filtered, commit does NOT abort
# T3 dangling code (untracked OR pre-staged) never embarked
# T4 stale-staged doc (version A) → commit carries working-tree version B
# T5 idempotent — empty list / clean tree → no-op exit 0
# T6 unsafe git state (detached HEAD) → exit 3, no commit
#
# No -e: run every test and report, even after a failure.
set -uo pipefail
HERE="$(cd -P "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
HELPER="$HERE/../doc-commit.sh"
ERRFILE="$(mktemp)"
PASS=0
FAIL=0
ok() { printf ' \033[32m✓\033[0m %s\n' "$1"; PASS=$((PASS + 1)); }
ko() { printf ' \033[31m✗\033[0m %s\n' "$1"; FAIL=$((FAIL + 1)); }
# Fresh throwaway repo: public docs + forbidden context + code, all tracked.
new_repo() {
local d
d="$(mktemp -d)"
git -C "$d" init -q
git -C "$d" config user.email t@t.t
git -C "$d" config user.name tester
mkdir -p "$d/.claude/memory" "$d/src"
printf 'readme baseline\n' >"$d/README.md"
printf 'usage baseline\n' >"$d/USAGE.md"
printf 'deploy baseline\n' >"$d/DEPLOY.md"
printf 'claude-md baseline\n' >"$d/CLAUDE.md"
printf 'decisions baseline\n' >"$d/.claude/memory/decisions.md"
printf 'src baseline\n' >"$d/src/app.txt"
git -C "$d" add -A
git -C "$d" commit -qm baseline
printf '%s' "$d"
}
# Files recorded in HEAD, sorted (stable compare).
head_files() { git -C "$1" diff-tree --no-commit-id --name-only -r HEAD | sort | tr '\n' ' '; }
# run <repo> <args...> → sets RC (exit), OUT (stdout = hash), ERR (stderr = diag).
run() {
local r="$1"; shift
OUT="$( (cd "$r" && "$HELPER" "$@") 2>"$ERRFILE" )"; RC=$?
ERR="$(cat "$ERRFILE")"
}
echo "T1a — forbidden path ALONE → REFUSE loud (exit 4), nothing committed"
R="$(new_repo)"
BEFORE="$(git -C "$R" rev-parse HEAD)"
printf 'dirtied\n' >>"$R/.claude/memory/decisions.md"
run "$R" commit "docs: T1a-claude" ".claude/memory/decisions.md"
printf ' rc=%s out=[%s]\n' "$RC" "$OUT"
printf ' err: %s\n' "$(printf '%s' "$ERR" | head -1)"
if [ "$RC" -eq 4 ]; then ok ".claude/ alone → exit 4"; else ko "expected 4, got $RC"; fi
if [ "$(git -C "$R" rev-parse HEAD)" = "$BEFORE" ]; then ok "no commit created"; else ko "a commit was created"; fi
if [ -z "$OUT" ]; then ok "stdout empty (no hash)"; else ko "stdout leaked: [$OUT]"; fi
if printf '%s' "$ERR" | grep -qi 'REFUSED'; then ok "stderr is loud (REFUSED)"; else ko "stderr not loud"; fi
printf 'dirtied\n' >>"$R/CLAUDE.md"
run "$R" commit "docs: T1a-claudemd" "CLAUDE.md"
printf ' [CLAUDE.md] rc=%s out=[%s]\n' "$RC" "$OUT"
if [ "$RC" -eq 4 ]; then ok "CLAUDE.md alone → exit 4"; else ko "expected 4, got $RC"; fi
rm -rf "$R"
echo "T1b — legit docs only → commits cleanly"
R="$(new_repo)"
printf 'feature added\n' >>"$R/README.md"
printf 'cmd changed\n' >>"$R/USAGE.md"
run "$R" commit "docs: T1b update README + USAGE" "README.md" "USAGE.md"
COMMITTED="$(head_files "$R")"
printf ' rc=%s out(hash)=[%s]\n' "$RC" "$OUT"
printf ' committed: [%s]\n' "$COMMITTED"
if [ "$RC" -eq 0 ]; then ok "exit 0"; else ko "expected 0, got $RC"; fi
if [ "$COMMITTED" = "README.md USAGE.md " ]; then ok "committed exactly README + USAGE"; else ko "got [$COMMITTED]"; fi
if [ -n "$OUT" ]; then ok "hash on stdout"; else ko "no hash printed"; fi
if [ -z "$(git -C "$R" status --porcelain -- .claude CLAUDE.md)" ]; then ok ".claude/CLAUDE.md untouched"; else ko "forbidden paths touched"; fi
rm -rf "$R"
echo "T1c — MIXED legit + forbidden → exit 4, NOTHING committed (the trap)"
R="$(new_repo)"
BEFORE="$(git -C "$R" rev-parse HEAD)"
printf 'feature added\n' >>"$R/README.md"
printf 'dirtied\n' >>"$R/.claude/memory/decisions.md"
run "$R" commit "docs: T1c mixed" "README.md" ".claude/memory/decisions.md"
printf ' rc=%s out=[%s]\n' "$RC" "$OUT"
printf ' err: %s\n' "$(printf '%s' "$ERR" | grep -i decisions | head -1)"
if [ "$RC" -eq 4 ]; then ok "mixed → exit 4"; else ko "expected 4, got $RC"; fi
if [ "$(git -C "$R" rev-parse HEAD)" = "$BEFORE" ]; then ok "NOTHING committed (README not half-committed)"; else ko "a commit slipped through"; fi
if printf '%s' "$ERR" | grep -q '.claude/memory/decisions.md'; then ok "stderr names the offender"; else ko "offender not named"; fi
if git -C "$R" status --porcelain | grep -q ' M README.md'; then ok "README left dirty (not embarked)"; else ko "README state wrong"; fi
rm -rf "$R"
echo "T2 — dynamic pathspec: clean passed path filtered, no abort"
R="$(new_repo)"
printf 'feature added\n' >>"$R/README.md"
printf 'cmd changed\n' >>"$R/USAGE.md"
# DEPLOY.md passed but NOT modified → must be filtered, must not abort the commit.
run "$R" commit "docs: T2" "README.md" "USAGE.md" "DEPLOY.md"
COMMITTED="$(head_files "$R")"
printf ' rc=%s committed=[%s]\n' "$RC" "$COMMITTED"
if [ "$RC" -eq 0 ]; then ok "exit 0 (clean DEPLOY.md did not abort)"; else ko "expected 0, got $RC"; fi
if [ "$COMMITTED" = "README.md USAGE.md " ]; then ok "committed README + USAGE only (DEPLOY filtered)"; else ko "got [$COMMITTED]"; fi
rm -rf "$R"
echo "T3 — dangling code (untracked + pre-staged) NOT embarked"
R="$(new_repo)"
printf 'feature added\n' >>"$R/README.md"
printf 'untracked junk\n' >"$R/src/dangling.txt"
printf 'staged junk\n' >"$R/src/staged.txt"; git -C "$R" add src/staged.txt
run "$R" commit "docs: T3" "README.md"
COMMITTED="$(head_files "$R")"
STATUS="$(git -C "$R" status --porcelain)"
printf ' committed=[%s]\n' "$COMMITTED"
if [ "$COMMITTED" = "README.md " ]; then ok "only README committed"; else ko "got [$COMMITTED]"; fi
if printf '%s\n' "$STATUS" | grep -q '^?? src/dangling.txt$'; then ok "untracked code left untracked"; else ko "untracked code embarked"; fi
if printf '%s\n' "$STATUS" | grep -q '^A src/staged.txt$'; then ok "pre-staged code stays staged"; else ko "pre-staged code embarked"; fi
rm -rf "$R"
echo "T4 — stale-staged doc (A) → commit carries working-tree (B)"
R="$(new_repo)"
printf 'VERSION A\n' >>"$R/README.md"; git -C "$R" add README.md # stage A
printf 'VERSION B\n' >>"$R/README.md" # working tree = A+B
run "$R" commit "docs: T4" "README.md"
HEADCONTENT="$(git -C "$R" show HEAD:README.md)"
printf ' HEAD README tail: %s\n' "$(printf '%s' "$HEADCONTENT" | tail -1)"
if printf '%s\n' "$HEADCONTENT" | grep -q 'VERSION B'; then ok "commit contains working-tree B (re-stage neutralized stale index)"; else ko "stale index A leaked"; fi
rm -rf "$R"
echo "T5 — idempotent: empty list / clean tree → no-op exit 0"
R="$(new_repo)"
BEFORE="$(git -C "$R" rev-parse HEAD)"
run "$R" commit "docs: T5 empty" # no files at all
printf ' [no files] rc=%s err=%s\n' "$RC" "$(printf '%s' "$ERR" | head -1)"
if [ "$RC" -eq 0 ]; then ok "empty list → exit 0"; else ko "expected 0, got $RC"; fi
run "$R" commit "docs: T5 clean" "README.md" # passed but clean
printf ' [clean README] rc=%s\n' "$RC"
if [ "$RC" -eq 0 ]; then ok "clean path → exit 0"; else ko "expected 0, got $RC"; fi
if [ "$(git -C "$R" rev-parse HEAD)" = "$BEFORE" ]; then ok "no commit created"; else ko "a commit was created"; fi
rm -rf "$R"
echo "T6 — unsafe state (detached HEAD) → exit 3, no commit"
R="$(new_repo)"
git -C "$R" checkout --detach -q
BEFORE="$(git -C "$R" rev-parse HEAD)"
printf 'feature added\n' >>"$R/README.md"
run "$R" commit "docs: T6" "README.md"
printf ' rc=%s err=%s\n' "$RC" "$(printf '%s' "$ERR" | head -1)"
if [ "$RC" -eq 3 ]; then ok "detached HEAD → exit 3"; else ko "expected 3, got $RC"; fi
if [ "$(git -C "$R" rev-parse HEAD)" = "$BEFORE" ]; then ok "no commit created"; else ko "a commit was created"; fi
rm -rf "$R"
rm -f "$ERRFILE"
echo ""
printf 'RESULT: %d passed, %d failed\n' "$PASS" "$FAIL"
[ "$FAIL" -eq 0 ]