#!/bin/bash

# sbodl, download slackbuilds.org source files

# 20230513 bkw: move to the sbo-maintainer-tools repo, clean up for release.
# 20191222 bkw: handle broken symlinks in current dir.
#               change default cache dir.
# 20170306 bkw: add caching

# Don't edit the next line; use "make version" instead.
VERSION=0.9.0

: <<EOF
=pod

=head1 NAME

sbodl - download and/or check md5sums for SlackBuilds.org source files.

=head1 SYNOPSIS

B<sbopkglint> [--help] | [ --version ] | [-a | -f | -af ] [I<wget-options> ...]

=head1 DESCRIPTION

B<sbodl> reads the .info file in the current directory and downloads
the source files listed there, if they don't already exist. Then
it checks the md5sums of the source files (whether they were just
downloaded or not).

To avoid extra network overhead, the downloaded files are stored in a
cache directory (see ENVIRONMENT), and maintained as symlinks in the
current directory. This allows you to remove the symlinks prior to
creating a tarball of your SlackBuild directory for upload, without
losing the actual files. It also allows builds that use the same
source files to share them (no need to re-download).

By default, B<--content-disposition> is used. If you want to disable it, use
the B<--no-content-disposition> wget option.

If you need to download files for a different architecture
(e.g. 32-bit files when running on a 64-bit OS), you can override ARCH
in the environment: "ARCH=i686 sbodl" would download the 32-bit source
file(s). This only matters if there's a DOWNLOAD_x86_64 URL in the
.info file.

=head1 OPTIONS

B<NOTE>: Any arguments other than the options listed below are passed
through to B<wget>(1) verbatim. Also, only the first argument may be
a B<sbodl> option. This means you can't say B<-a -f>; use B<-af> or B<-fa> instead.

=over 4

=item B<-a>

Download from the SBo Source Archive (sbosrcarch) rather than the
original URL(s). Use this if the original site is down or the file has
gone 404. See ENVIRONMENT, if you need to change the sbosrcarch URL.

=item B<-f>

Force downloading the file(s), even if they already exist in the
current directory or in the cache.

=item B<-af>, B<-fa>

Combine the above two options (B<-a> and B<-f>).

=item B<--help>, B<--doc>

Show this documentation in your pager.

=item B<--version>

Show version number and exit.

=item B<--man>

Output this documentation, formatted as a man page, on standard output.

=back

=head1 ENVIRONMENT

=over 4

=item B<SBODL_ARCHIVE>

The URL of the sbosrcarch instance. The only known one at the time
of this writing is B<https://slackware.uk/sbosrcarch>, but that may
change in the future (or you may run your own private instance). This
is the URL of the directory containing the B<by-name> and B<by-md5>
dirs, and should not end with a trailing slash.

=item B<SBODL_CACHEDIR>

Where to cache downloaded files. By default, this is
B<$HOME/sbodl-cache>. This directory will be created if it does not
exist.

=item B<ARCH>

Used to determine which files to download; default is autodetect. Only
matters for builds with a separate 64-bit download URL. Use B<i586>
or B<i686> for 32-bit x86, B<x86_64> for 64-bit. Note that there's
no point setting this to e.g. B<aarch64>, as the .info file format
doesn't have a DOWNLOAD_aarch64 field (yet).

=back

=head1 EXIT STATUS

0 (success) if all files were downloaded or found in the cache, and their
md5sums all match the .info file.

1 (failure) if any download failed, or if at least one md5sum failed to match.

=head1 BUGS

One very rare but nasty one: all the cached files are stored in the
same directory. If two SlackBuilds happen to have the same filename
for their downloads (but they're actually different files), you'll get
md5sum failures. This is most noticeable with I<haskell/> builds, with
filenames like B<1.cabal>, B<2.cabal>, etc. At some point this will be
fixed. For now, if this happens, use the B<-f> option.

Also, since --content-disposition is used, occasionally the downloaded
filename won't match the filename in the URL, and B<sbodl> will warn
that it can't find the downloaded file. Again, this will be fixed
soonish. I left this enabled specifically to detect URLs that have
this problem (mostly, these are github URLs and there's a way to
rewrite them to work either way).

=head1 AUTHOR

B. Watson <urchlay@slackware.uk>, aka Urchlay on Libera IRC.

=head1 SEE ALSO

B<sbolint>(1), B<sbopkglint>(1), B<sbofixinfo>(1), B<sbopkg>(8), B<sboinstall>(1)

=cut
EOF

SELF=$( basename $0 )

DEFCACHEDIR=~/sbodl-cache
CACHEDIR=${SBODL_CACHEDIR:-$DEFCACHEDIR}

DEFARCHIVE=https://slackware.uk/sbosrcarch
ARCHIVE=${SBODL_ARCHIVE:-$DEFARCHIVE}

if [ -t 1 ]; then
	RED="\033[1;31m"
	GREEN="\033[1;32m"
	YELLOW="\033[1;33m"
	COLOR_OFF="\033[0m"
fi

die() {
	echo "$SELF: $*" 2>&1
	exit 1
}

[ -d "$CACHEDIR/" ] || mkdir -p $CACHEDIR || die "Can't create $CACHEDIR"

# check for our one argument
case "$1" in
	-v|--version) echo $VERSION ; exit 0 ;;
	-h|--help|--doc) exec perldoc "$0" ;;
	--man) exec pod2man --stderr -s1 -csbo-maintainer-tools -r$VERSION "$0" ;;
	-f) FORCEDL="yes" ; shift ;;
	-a) USEARCHIVE="yes" ; shift ;;
	-fa|-af) FORCEDL="yes" ; USEARCHIVE="yes" ; shift ;;
esac

source ./$( basename $( pwd ) ).info \
	|| die "No .info file, are you sure this is a SBo directory? Try '$SELF --help'"

# This stanza copied from the SBo template for 14.1:
if [ -z "$ARCH" ]; then
	case "$( uname -m )" in
		i?86) ARCH=i586 ;;
		arm*) ARCH=arm ;;
		*) ARCH=$( uname -m ) ;;
	esac
fi

if [ "$ARCH" = "x86_64" ]; then
	DL=${DOWNLOAD_x86_64:-$DOWNLOAD}
	SUM=${MD5SUM_x86_64:-$MD5SUM}
else
	DL=$DOWNLOAD
	SUM=${MD5SUM}
fi

if [ -z "$DL" ]; then
	die "Bad .info file (no DOWNLOAD= or DOWNLOAD_x86_64=)."
fi

if [ "$DL" = "UNSUPPORTED" ]; then
	die "$PRGNAM is unsupported on $ARCH"
fi

# save passed-in command line args for use with wget
WGETARGS="$@"

EXIT=0

set $SUM

for dl in $DL; do
	if [ "$USEARCHIVE" = "yes" ]; then
		file="$( echo "$dl" | sed 's,.*/,,' )"
		a=$( echo $1 | cut -b1 )
		b=$( echo $1 | cut -b2 )
		dl="$ARCHIVE/by-md5/$a/$b/$1/$file"
	fi

	EXTRAWGETARGS="--content-disposition "
	case "$dl" in
		*sourceforge.net/*|*.sf.net/*) EXTRAWGETARGS="--user-agent wget" ;;
		*) ;;
	esac

	FILE=$( echo "$dl" | sed 's,.*/,,' )

	# ignore (rm) broken symlinks
	[ -L "$FILE" ] && ! [ -e "$FILE" ] && FORCEDL=yes

	[ "$FORCEDL" = "yes" ] && rm -f "$FILE"

	if [ -f "$FILE" -a ! -L "$FILE" ]; then
		# file exists and is a regular file, cache it
		mv -b "$FILE" "$CACHEDIR"
		ln -s "$CACHEDIR/$FILE" "$FILE"
	elif [ -e "$FILE" ]; then
		# don't do anything
		:
	elif [ "$FORCEDL" != "yes" ] && [ -e "$CACHEDIR/$FILE" ]; then
		ln -s "$CACHEDIR/$FILE" "$FILE"
	else
		wget $EXTRAWGETARGS $WGETARGS "$dl" || die "Download failed, try '$SELF -a'."
		if [ -e "$FILE" ]; then
			mv -b "$FILE" "$CACHEDIR"
			ln -s "$CACHEDIR/$FILE" "$FILE"
		fi
	fi

	if [ -e "$FILE" ]; then
		GOTSUM="$( md5sum "$FILE" | cut -d' ' -f1 )"
		if [ "$1" != "$GOTSUM" ]; then
			echo -e "${RED}WARN: md5sum doesn't match${COLOR_OFF}\\n     expected: $1\\n          got: $GOTSUM"
			case "$FILE" in
				?.cabal) echo -e "${YELLOW}HINT:${COLOR_OFF} for ?.cabal files: if the md5sum fails, retry with '$SELF -f'." ;;
			esac
			EXIT=1
		else
			echo -e "${GREEN}md5sum matches OK${COLOR_OFF}: $1"
		fi
	else
		echo -e "${RED}WARN${COLOR_OFF}: can't find downloaded file $FILE, try '$SELF -a'?"
		EXIT=1
	fi
	shift
done

exit $EXIT