#!/bin/sh
# script based on initial implementation
# by Vitaly Gusach (http://gusach.org.ua)
#
# purpose: warn on those entries in given packagelists(s)
# which are definitely absent; the build might still bail out
# but at least 80% of failures can be predicted early now
#
# usage: check-pkg-list [-n pkgnames] pkglist ...
# (pkgnames file should contain `apt-cache pkgnames`)
# NB: -n pkgnames MUST go first, if given

error() { echo `basename $0`: $* >&2; exit 1; }
debug() { [ -n "$GLOBAL_VERBOSE" ] && echo `basename $0`: $* >&2; }

exit_handler() {
	local rc=$?
	trap - EXIT
	rm -f -- "$ftemp" "$fpkgnames" "$fpkgwildcards" "$favaillist"
	exit $rc
}

# find the most recent apt.conf in workdirs
apt_conf() {
	debug "looking for workdirs with aptboxes..."
	find -maxdepth 3 -name .work \
	| while read dir; do
		file="$dir/aptbox/etc/apt/apt.conf"
		[ -f "$file" ] && echo "$file"
	done \
	| xargs -r ls -t -- \
	| head -1
}

# figure out apt.conf from recent aptbox or fallback to system one
# TODO: would be nice to cache the dump inside aptbox
dump_pkgnames() {
	conf="`apt_conf`"
	debug "autodetected apt.conf: ${conf:-<system>}, dumping pkgnames"
	apt-cache ${conf:+-c $conf} pkgnames | sort > "$favaillist"
}

check_pkglist() {
	fprofilelist="$1"
	[ -f "$fprofilelist" ] || error "invalid packagelist filename: $fprofilelist"
	[ -n "$MULTI" ] && echo "-- $fprofilelist"
	debug "checking $fprofilelist against $favaillist"

	# cleaning pkg list from comments, empty lines,
	# splitting several pkgnames on the same line
	sed -e '/^#/d' -e '/^[ 	]*$/d' -e 's/ \+$//' -e 's/[ 	]\+/\n/g' \
	< "$fprofilelist" \
	| sed 's/-$//' \
	| sort -u \
	> "$ftemp" # got list of pkgnames we need

	# split pkgnames without wildcards and with wildcards
	fgrep -v '*' "$ftemp" > "$fpkgnames"
	fgrep 	 '*' "$ftemp" > "$fpkgwildcards"

	# return unavailable packages
	debug "unavailable packagenames, if any:"
	comm -23 "$fpkgnames" "$ftemp"

	# return unavailable wildcards
	debug "unavailable wildcards, if any:"
	while read i; do
		# replacing * with regexp's \.+
		pattern="^`echo ${i#^} | sed -e 's/\*/.\\\\+/'`$"
		grep -q "$pattern" "$favaillist" || echo "$i"
	done < "$fpkgwildcards"
}

[ "$#" -gt 0 ] || error "need at least one argument, a packagelist to check"

# reusable temporary files with self-cleanup at exit
TEMP="${TMP:-/tmp}"
trap exit_handler HUP INT QUIT TERM EXIT
favaillist="`mktemp $TEMP/pkgchecker.avail.XXXXX`"
fpkgnames="`mktemp $TEMP/pkgchecker.names.XXXXX`"
fpkgwildcards="`mktemp $TEMP/pkgchecker.wildcards.XXXXX`"
ftemp="`mktemp $TEMP/pkgchecker.XXXXX`"

# make sure pkgnames dump is handy
case "$1" in
"-n"|"--pkgnames")
	[ -f "$2" ] && {
		sort "$2" > "$favaillist"
		shift; shift
	} || error "-n needs valid pkgnames filename"
	;;
esac

[ -s "$favaillist" ] || dump_pkgnames

# now on to the real job; Q: skip *.in or not?
[ "$#" -eq 1 ] && MULTI= || MULTI=1

for list in "$@"; do
	check_pkglist "$list"
done