A bash script to backup system only with modified files

Question

Hi,
I've made a simple bash script to exam which files are modified and needed to be backed up.
It will use:
1. the hash in Pacman's database (/var/lib/pacman/local/<pkg_ver>/files
2. if no hash in the database, calculate it by our self from cache (/var/cache/pacman/pkg/<pkg_ver>.pkg.tar.gz
3. if no cache found, compare build date and last modified date
4. otherwise, this file better be backed up.
Currently it will only print which files are needed to be backed up, but no backup actually happens. (dry run)
And it is only in early development stage, please be careful.
USAGE:
<the script name> <where to backup files, a directory> <the files, or directories, separated by space, that are going to be examined>...
Here is the code.
#!/bin/bash
# usage:
# $1: file to backup
function do_backup() {
echo
function smart_bak() {
pkg=$(pacman -Qo "$1" 2>/dev/null)
if [ 1 -eq $? ] ; then
# No package owns $1 (locally created)
if [ "$1" != "${1%.pacsave}" ] ; then
echo "skip $1 (left by removed package)"
else
echo "backup $1 (local file)"
fi
else
pkg=${pkg#$1 is owned by }
# evaluate
# by hash
cur=$(md5sum $1)
cur=${cur%% *}
pkg_ver=${pkg/ /-}
org=$(grep "^${1:1}" "/var/lib/pacman/local/$pkg_ver/files")
org_hash=${org##* }
if [ "$org" != "$org_hash" ] ; then
# the org hash is in Pacman's database
if [ "$org_hash" = "$cur" ] ; then
echo "skip $1 (original config)"
else
echo "backup $1 (modified config)"
fi
else
# no hash
# find out hash myself?
ARCH=$(uname -m)
if [ -r "/var/cache/pacman/pkg/$pkg_ver-$ARCH.pkg.tar.gz" ] ; then
org=$(tar -Oxzf "/var/cache/pacman/pkg/$pkg_ver-$ARCH.pkg.tar.gz" "${1:1}" | md5sum -)
org_hash=${org%% *}
if [ "$cur" = "$org_hash" ] ; then
echo "skip $1 (original)"
else
echo "backup $1 (modified)"
fi
else
# no cache, may be a AUR package
# fall back to built date?
date=$(pacman -Qi ${pkg% *} | grep "^Build Date")
date=${date#*: }
date=$(date -d "$date" +%s)
mod=$(ls -l $1)
mod=${mod% $1}
mod=${mod#* * * * * }
mod=$(date -d "$mod" +%s)
tmp1=$(expr $mod "+" 60)
tmp2=$(expr $mod "-" 60)
if [ "$date" -le "$tmp1" -a "$date" -ge "$tmp2" ] ; then
echo "skip $1 (the same date)"
else
echo "backup $1 (unknown)"
fi
fi
fi
fi
function smart_bak_dir() {
for i in $(ls -A "$1") ; do
tmp="${1%/}/$i"
if [ -f "$tmp" ] ; then
smart_bak "$tmp"
elif [ -d "$tmp" ] ; then
smart_bak_dir "$tmp"
else
echo "skip $tmp (not a regular file nor a directory)"
fi
done
# usage:
# $1: the directory to store this backup
# $2: directory to evalualte for backup
# function main()
# init
target="$1"
shift
# check
if [ ! -d "$target" -o ! -x "$target" -o ! -w "$target" ] ; then
exit 4
fi
for i in $* ; do
if [ -f "$i" ] ; then
smart_bak "$i"
elif [ -d "$i" ] ; then
smart_bak_dir "$i"
else
echo "skip $i (unknown argument)"
fi
done
Good luck,
bsdson
Last edited by bsdson.tw (2008-12-30 07:53:05)

Answer 1

Thanks for this script. Nice work.
Can we traverse the pacman.log backwards up and rollback each operation (including "installed" in this). Something like:
history=$(tail -n +$(grep -m 1 -n "$1" "$LOG" | cut -d ":" -f 1) "$LOG" | tac | grep -E "(upgraded|installed)" | cut -d " " -f 3-)
Last edited by g33k (2008-11-04 15:08:07)