Add collect_shadows dup detection + self-check

Finds upstream SBo packages shadowed by a personal-tree twin, read-only,
no removal. Personal package = subdir with matching <name>.SlackBuild;
prune array excludes every SBO_PERSONAL_TREES path from the search so a
personal package never lists itself.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

2 files changed, 48 insertions, 0 deletions

diff --git a/sbo-batch-test b/sbo-batch-test
index f2b9c14..c028df1 100755
--- a/sbo-batch-test
+++ b/sbo-batch-test
@@ -343,6 +343,32 @@ rsync_tree() {
     "$SBO_RSYNC_URL/" "$root/"
 }
 
+# Echo (one per line) the <root>-relative dirs of upstream packages that
+# duplicate a package in a personal subtree. Reads only, never removes. A
+# personal package is a subdir with a matching <name>.SlackBuild; an upstream
+# dup is a same-named depth-2 dir anywhere under <root> outside the personal
+# trees.
+collect_shadows() {
+  local root="$1"
+  local -a personal_dirs=() prune=()
+  local dir
+  for dir in "${!SBO_PERSONAL_TREES[@]}"; do
+    [[ -d "$root/$dir" ]] || continue
+    personal_dirs+=("$root/$dir")
+    prune+=(! -path "$root/$dir/*")
+  done
+  [[ ${#personal_dirs[@]} -eq 0 ]] && return
+  local pkgdir pkg match
+  while IFS= read -r pkgdir; do
+    pkg=$(basename "$pkgdir")
+    [[ -f "$pkgdir/$pkg.SlackBuild" ]] || continue
+    while IFS= read -r match; do
+      echo "${match#"$root"/}"
+    done < <(find "$root" -mindepth 2 -maxdepth 2 -type d \
+        -name "$pkg" "${prune[@]}")
+  done < <(find "${personal_dirs[@]}" -mindepth 1 -maxdepth 1 -type d)
+}
+
 # =============================================================================
 # Package cache. Persistent, on-disk store of built packages so unchanged deps
 # are installed from cache instead of rebuilt. Layout mirrors the SBo tree:
diff --git a/test-logic.sh b/test-logic.sh
index 423d2b5..cbdef9d 100755
--- a/test-logic.sh
+++ b/test-logic.sh
@@ -152,6 +152,28 @@ echo "$ex" | grep -qx -- "--exclude=/pentesting/" && ok "rsync_excludes pentesti
 declare -A SBO_PERSONAL_TREES=()
 [[ -z "$(rsync_excludes)" ]] && ok "rsync_excludes empty for no trees" || bad "rsync_excludes should be empty, got [$(rsync_excludes)]"
 
+# --- collect_shadows: upstream dups of personal packages -------------------
+S=$(mktemp -d)
+mkslb() { mkdir -p "$S/$1/$2"; : > "$S/$1/$2/$2.SlackBuild"; }
+# upstream copies
+mkslb academic foo
+mkslb network  bar
+# personal copies (win)
+mkslb personal   foo
+mkslb pentesting bar
+# decoy: personal pkg with no upstream twin
+mkslb personal nomatch
+# decoy: personal subdir without a .SlackBuild (ignored)
+mkdir -p "$S/personal/noslb"
+declare -A SBO_PERSONAL_TREES=([personal]=u1 [pentesting]=u2)
+
+got="$(collect_shadows "$S" | sort)"
+want="$(printf 'academic/foo\nnetwork/bar\n' | sort)"
+[[ "$got" == "$want" ]] && ok "collect_shadows finds upstream dups only" || bad "collect_shadows wrong; got [$got] want [$want]"
+echo "$got" | grep -q "personal/"   && bad "collect_shadows returned a personal path" || ok "collect_shadows excludes personal trees"
+echo "$got" | grep -q "nomatch"     && bad "collect_shadows invented a match for nomatch" || ok "collect_shadows ignores no-twin pkg"
+rm -rf "$S"
+
 # --- result -----------------------------------------------------------------
 echo
 echo "$pass passed, $fail failed"