#!/bin/bash
# Git export script
# distributed by puppet

# this takes about 1.5mins to run the first time. <45s thereafter

# if the lv is really small (2G), the default inode ratio is too large.
# eg. mkfs.ext2 -i 4096 /dev/vg/var_mastermirror-staging

# Notes 2015/12/14 by robbat2
# This entire script was run on the CVS server directly, as it was MUCH faster
# than remote checkouts. In the era of Git, this script can probably be merged
# with rsync-gen.sh instead.

GITROOT_file="file:///var/gitroot"
GITROOT_http="https://anongit.gentoo.org/git/"
case $HOSTNAME in
	oystercatcher|swan|manakin) export GITROOT=${GITROOT_file} ;;
	*) export GITROOT=${GITROOT_http} ;;
esac

GIT="/usr/bin/git"
GIT_CL="${GIT} clone -q"
GIT_PL="${GIT} pull -q --ff-only --no-stat --no-rebase"

WGET="/usr/bin/wget --timeout=2 --quiet --timestamping"

GIT_RESTORE_MTIME="$(find /usr/lib/python-exec/python*/ -maxdepth 1 -name git-restore-mtime |sort -r |head -n1)"
GIT_RESTORE_MTIME_ARGS="-q -m -c --first-parent"

rc=0
for cmdname in \
	GIT_RESTORE_MTIME \
	WGET \
	GIT \
	; do
	cmd=${!cmdname}
	cmd=${cmd/ *}
	if [[ -z "$cmd" ]] || [[ ! -x "$cmd" ]]; then
		echo "Missing tool ($cmdname): $cmd" 1>&2
		rc=1
	fi
done
[[ $rc -ne 0 ]] && exit 1

STAGING_DIR="${1:-/var/tmp/mastermirror-staging}"
STAGING_DIR_glsa="${STAGING_DIR}/glsa"
STAGING_DIR_dtd="${STAGING_DIR}/dtd"
STAGING_DIR_xmlschema="${STAGING_DIR}/xml-schema"
STAGING_DIR_news="${STAGING_DIR}/gentoo-news"
STAGING_DIR_gentoo="${STAGING_DIR}/gentoo-x86"
STAGING_DIR_projects="${STAGING_DIR}/projects"
REPO_gentoo=${GITROOT}/repo/gentoo.git
REPO_news=${GITROOT}/data/gentoo-news.git
REPO_dtd=${GITROOT}/data/dtd.git
REPO_xmlschema=${GITROOT}/data/xml-schema.git
REPO_glsa=${GITROOT}/data/glsa.git

PROJECTS_XML_URI=https://api.gentoo.org/metastructure/projects.xml

# gitattributes
ATTRIB_gentoo=(
  '* -ident'
)

# Ensure all files are readable
umask 0022

fetch_git() {
	targetdir=$1
	repo=$2
	shift 2
	local timestampfile=${timestampfile:-timestamp.commit}
	gitdir=${targetdir}.git
	timestampfile=${targetdir}/$timestampfile
	[ -f "$timestampfile" ] && rm -f "$timestampfile"
	if [[ ! -d "${targetdir}" ]] ; then
		# not checked out yet, run initial co
		mkdir -p "${targetdir}.git" "$gitdir" && \
			$GIT init -q \
				--separate-git-dir "$gitdir"
				"${targetdir}" && \
			GIT_DIR=${gitdir} $GIT remote add origin "$repo"
	fi
	attribfile=${gitdir}/info/attributes
	for line in "$@" ; do
		fgrep -sq -e "$line" "$attribfile" ||
			echo "$line" >>"$attribfile"
	done
	excludefile=${gitdir}/info/exclude
	for line in timestamp.commit timestamp.chk timestamp.x timestamp ; do
		fgrep -sq -e "$line" "$excludefile" ||
			echo "$line" >>"$excludefile"
	done
	cd "${targetdir}" || return 1
	$GIT fetch -q --force origin master || return 1
	$GIT reset -q --hard origin/master || return 1
	$GIT_RESTORE_MTIME $GIT_RESTORE_MTIME_ARGS || return 1
	TZ=UTC $GIT log -1 --date=iso-strict-local --format='%H %ct %cd' >${timestampfile}
}

fetch_uri() {
	targetfile=$1
	uri=$2
	targetdir=$(dirname $(readlink -m $targetfile))
	[[ ! -d ${targetdir} ]] &&  mkdir -p ${targetdir}
	tmpfile=$(mktemp -p $targetdir)
	${WGET} "${uri}" -O $tmpfile && chmod a+r $tmpfile && mv -f "$tmpfile" "$targetfile"
	rc=$?
	rm -f "$tmpfile"
	return $rc
}

# TODO: we should probably be checking out to a directory and only swapping on
# success of ALL checkouts and fetches.
rc_sum=0

# repo/gentoo (formerly CVS gentoo-x86)
timestampfile=metadata/timestamp.commit fetch_git ${STAGING_DIR_gentoo} ${REPO_gentoo} "${ATTRIB_gentoo[@]}"
rc=$?
rc_sum=$((rc_sum + $rc))

# gentoo-news
fetch_git ${STAGING_DIR_news} ${REPO_news}
rc=$?
rc_sum=$((rc_sum + $rc))

# projects.xml
fetch_uri ${STAGING_DIR_projects}/projects.xml "${PROJECTS_XML_URI}"
rc=$?
rc_sum=$((rc_sum + $rc))

# dtd
fetch_git ${STAGING_DIR_dtd} ${REPO_dtd}
rc=$?
rc_sum=$((rc_sum + $rc))

# xml-schema
fetch_git ${STAGING_DIR_xmlschema} ${REPO_xmlschema}
rc=$?
rc_sum=$((rc_sum + $rc))

# glsa
fetch_git ${STAGING_DIR_glsa} ${REPO_glsa}
rc=$?
rc_sum=$((rc_sum + $rc))

exit $rc_sum