#!/bin/bash
# audit-packages.sh — verify every package archsetup installs still exists
# at its declared source, and flag packages that moved between the official
# repos and the AUR.
#
# Extraction covers direct `pacman_install pkg` / `aur_install pkg` calls
# and `for software in a b c; do` loop lists (backslash continuations
# included). Variable arguments ("$software") are the loop plumbing, not
# packages, and are skipped.
#
# Checks:
#   official package missing from sync dbs  -> MOVED TO AUR (if AUR has it)
#                                              or MISSING EVERYWHERE
#   AUR package gone from the AUR           -> NOW IN OFFICIAL (if repos
#                                              have it) or MISSING EVERYWHERE
#
# Usage: audit-packages.sh [--list] [installer-script]
#   --list   print the extracted package lists and exit (no network)
#   default  audit; exit 1 when anything is missing or moved
#
# Env overrides (tests): AUDIT_PACMAN, AUDIT_CURL.

set -u

PACMAN="${AUDIT_PACMAN:-pacman}"
CURL="${AUDIT_CURL:-curl}"

mode=audit
case "${1:-}" in
    --list) mode=list; shift ;;
esac

script="${1:-$(dirname "$0")/../archsetup}"
if [ ! -r "$script" ]; then
    echo "audit-packages: cannot read installer script: $script" >&2
    exit 2
fi

# Extract package names per source. Loop lists are attributed to the
# install command used inside the loop body.
extract() {
    awk '
        # Collect a `for software in ...` list (may span backslash-continued
        # lines); attribute it when the loop body shows which installer runs.
        /for[ \t]+[A-Za-z_]+[ \t]+in[ \t]/ {
            line = $0
            sub(/^.*[ \t]in[ \t]+/, "", line)
            listpkgs = ""
            while (1) {
                stop = (line ~ /;[ \t]*do([ \t]|$)/)
                sub(/;[ \t]*do.*$/, "", line)
                gsub(/\\/, "", line)
                listpkgs = listpkgs " " line
                if (stop) break
                if ((getline line) <= 0) break
            }
            pending = listpkgs
            next
        }
        /^[ \t]*(pacman_install|aur_install)[ \t]/ {
            cmd = ($0 ~ /pacman_install/) ? "official" : "aur"
            arg = $2
            gsub(/["'\'']/, "", arg)
            if (arg ~ /^\$/) {
                # loop plumbing — emit the pending for-list under this source
                n = split(pending, w, /[ \t]+/)
                for (i = 1; i <= n; i++) if (w[i] != "") print cmd, w[i]
                pending = ""
            } else if (arg != "") {
                print cmd, arg
            }
        }
    ' "$script" | sort -u
}

pairs=$(extract)
official=$(echo "$pairs" | awk '$1 == "official" { print $2 }')
aur=$(echo "$pairs" | awk '$1 == "aur" { print $2 }')

if [ "$mode" = list ]; then
    echo "OFFICIAL ($(echo "$official" | grep -c .)):"
    echo "$official" | sed 's/^/  /'
    echo "AUR ($(echo "$aur" | grep -c .)):"
    echo "$aur" | sed 's/^/  /'
    exit 0
fi

# One batched AUR RPC query answers existence for every AUR-relevant name.
aur_query() {
    # args: package names; prints the names the AUR knows
    local args="" p
    for p in "$@"; do args="$args&arg[]=$p"; done
    "$CURL" -sm 20 "https://aur.archlinux.org/rpc/v5/info?${args#&}" 2>/dev/null \
        | tr ',' '\n' | sed -n 's/.*"Name":[[:space:]]*"\([^"]*\)".*/\1/p'
}

# shellcheck disable=SC2086
aur_known=$(aur_query $official $aur)

in_repos()  { "$PACMAN" -Si "$1" >/dev/null 2>&1; }
in_aur()    { echo "$aur_known" | grep -qx "$1"; }

moved_to_aur="" now_official="" missing=""
for p in $official; do
    in_repos "$p" && continue
    if in_aur "$p"; then moved_to_aur="$moved_to_aur $p"; else missing="$missing $p"; fi
done
for p in $aur; do
    if in_aur "$p"; then
        in_repos "$p" && now_official="$now_official $p (also in repos)"
        continue
    fi
    if in_repos "$p"; then now_official="$now_official $p"; else missing="$missing $p"; fi
done

rc=0
if [ -n "$missing" ]; then
    echo "MISSING EVERYWHERE (not in repos, not in AUR):"
    for p in $missing; do echo "  $p"; done
    rc=1
fi
if [ -n "$moved_to_aur" ]; then
    echo "MOVED TO AUR (pacman_install will fail; switch to aur_install):"
    for p in $moved_to_aur; do echo "  $p"; done
    rc=1
fi
if [ -n "$now_official" ]; then
    echo "NOW IN OFFICIAL (aur_install works but pacman_install is cleaner):"
    echo "$now_official" | tr ' ' '\n' | grep -v '^$' | sed 's/^/  /'
    rc=1
fi
total=$(( $(echo "$official" | grep -c .) + $(echo "$aur" | grep -c .) ))
echo "---"
echo "$total packages audited"
exit $rc