#!/bin/bash

. "$(dirname $0)/../db-functions"
. "$(dirname $0)/../config"

# setup paths
SPATH="/srv/http/archweb"
ENVPATH="/srv/http/archweb-env/bin/activate"

# having "more important repos" last should make [core] trickle to the top of
# the updates list each hour rather than being overwhelmed by big [extra] and
# [community] updates
REPOS=('community-testing' 'multilib-testing' 'multilib' 'community' 'extra' 'testing' 'core')
LOGOUT="/tmp/archweb_update.log"

# figure out what operation to perform
cmd="$(basename $0)"
if [[ $cmd != "update-web-db" && $cmd != "update-web-files-db" ]]; then
	die "Invalid command name '$cmd' specified!"
fi

script_lock

# run at nice 5. it can churn quite a bit of cpu after all.
renice +5 -p $$ > /dev/null

echo "$cmd: Updating DB at $(date)" >> "${LOGOUT}"

# source our virtualenv if it exists
if [ -f "$ENVPATH" ]; then
	. "$ENVPATH"
fi

case "$cmd" in
	update-web-db)
		dbfileext="${DBEXT}"
		flags=""
	;;
	update-web-files-db)
		dbfileext="${FILESEXT}"
		flags="--filesonly"
	;;
esac

# Lock the repos and get a copy of the db files to work on
for repo in ${REPOS[@]}; do
	for arch in ${ARCHES[@]}; do
		repo_lock ${repo} ${arch} || exit 1
		dbfile="/srv/ftp/${repo}/os/${arch}/${repo}${dbfileext}"
		if [ -f "${dbfile}" ]; then
			mkdir -p "${WORKDIR}/${repo}/${arch}"
			cp "${dbfile}" "${WORKDIR}/${repo}/${arch}/${repo}${dbfileext}"
		fi
		repo_unlock ${repo} ${arch}
	done
done

# Run reporead on our db copy
pushd $SPATH >/dev/null
for repo in ${REPOS[@]}; do
	for arch in ${ARCHES[@]}; do
		dbcopy="${WORKDIR}/${repo}/${arch}/${repo}${dbfileext}"
		if [ -f "${dbcopy}" ]; then
			echo "Updating ${repo}-${arch}" >> "${LOGOUT}"
			./manage.py reporead ${flags} ${arch} "${dbcopy}" >> "${LOGOUT}" 2>&1
			echo "" >> "${LOGOUT}"
		fi
	done
done
popd >/dev/null
echo "" >> "${LOGOUT}"

# rotate the file if it is getting big (> 10M), overwriting any old backup
if [[ $(stat -c%s "${LOGOUT}") -gt 10485760 ]]; then
	mv "${LOGOUT}" "${LOGOUT}.old"
fi

script_unlock