#!/usr/bin/env bash
#
# build-baseline.sh — reproducibly assemble the www.archline.hu clean CODE baseline.
#
# Output (out/) = the deployed docroot (code only), which the operator's HIDS
# baseline-diff verifies the live server against.
#
# The repo stores the baseline CONSTITUENTS, not the deployed tree; this script
# performs the deterministic "deployment" (Joomla manifest-mapping for packaged
# extensions, direct placement for no-source ones), per RUNBOOK.md.
#
# Usage: ./build-baseline.sh [OUT_DIR]   (default: ./out)
set -euo pipefail
ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
OUT="${1:-$ROOT/out}"
JMAP="$ROOT/lib/jmap.py"
log(){ printf '[build] %s\n' "$*"; }

rm -rf "$OUT"; mkdir -p "$OUT"

# 1) Joomla core — official 3.8.11 package (deployed form), minus demo assets and
#    runtime data dirs (cache/tmp/logs/images are data, not verified code).
log "core: Joomla 3.8.11 (official package)"
rsync -a \
  --exclude='cache/' --exclude='tmp/' --exclude='logs/' \
  --exclude='images/sampledata/' --exclude='images/banners/' --exclude='images/headers/' \
  --exclude='images/joomla_black.png' --exclude='images/powered_by.png' \
  "$ROOT/core/" "$OUT/"

# 2) Upstream-package extensions — deployed-layout, byte-reconciled to live.
#    In git these are authored as two commits per modified module: (a) the
#    pristine upstream package manifest-mapped by lib/jmap.py, (b) the reconcile
#    to the live bytes (line-endings + install files + Cadline delta). `git diff`
#    between those two commits documents exactly how live differs from upstream.
#    At build time the reconciled (live-byte) deployed files are placed directly.
for d in "$ROOT"/packages/*/; do
  [ -d "$d" ] || continue
  log "package: $(basename "$d")"
  rsync -a --exclude='.provenance' "$d" "$OUT/"
done

# 3) No-source extensions — vetted deployed files (no obtainable package),
#    placed directly (already in deployed layout).
for d in "$ROOT"/deployed/*/; do
  [ -d "$d" ] || continue
  log "deployed: $(basename "$d")"
  rsync -a --exclude='.provenance' "$d" "$OUT/"
done

# 4) Cadline own-code (components/modules/plugins/templates/docroot apps).
for d in "$ROOT"/cadline/*/; do
  [ -d "$d" ] || continue
  log "cadline: $(basename "$d")"
  rsync -a --exclude='.provenance' "$d" "$OUT/"
done

# 5) Cadline core modifications — live versions overlaid on the package core
#    (the ~24 legit customizations); the 3 trojanized core files stay clean.
if [ -d "$ROOT/overrides" ]; then
  log "overrides: Cadline core modifications"
  rsync -a --exclude='.gitkeep' "$ROOT/overrides/" "$OUT/"
fi

# 6) Safety net — malware IOC + scan artifacts must never appear in the baseline
#    (the sources above already exclude them; this is defensive, documented in
#    malware-iocs.md / malware-iocs.paths).
log "exclude: malware IOC + scan artifacts"
find "$OUT" \( -name '*.php.json' -o -name '*.phtml.json' -o -name 'h3x_*' -o -name '_h3x_*' \) -delete 2>/dev/null || true
if [ -f "$ROOT/malware-iocs.paths" ]; then
  while IFS= read -r ioc; do
    [ -z "$ioc" ] && continue
    case "$ioc" in \#*) continue;; esac
    rm -rf "${OUT:?}/$ioc"
  done < "$ROOT/malware-iocs.paths"
fi

log "baseline built at: $OUT ($(find "$OUT" -type f | wc -l) files)"