path: root/sbolint

#!/usr/bin/perl -w

# ChangeLog:

# 0.3 20200420 bkw:
# - Check github URLs for validity.

# 0.2 20200103 bkw:
# - Use "git rev-parse" to decide if we're in a git repo, because
#   "git status" traverses the whole repo looking for untracked files.
#   It does this even if you use -uno (it won't *print* the untracked
#   files, but it still searches for them). Thanks to alienBOB for cluing
#   me in to using rev-parse for this.
# - Skip the junkfiles check when we're in a git repo. It's more
#   annoying than it is useful.
# - Allow possible -e/-u arguments in the shebang.
# - Avoid false positives when the script does a "cd $PKG" and then
#   uses relative paths for install/*.
# - Require VERSION= to appear within the first 10 non-comment/non-blank
#   lines, and don't check it anywhere else in the script.
# - Allow scripts to skip lint checks via ###sbolint on/off comments.

# 0.1 20141114 bkw, Initial release.

$VERSION="0.3";

# generate man page with:
# pod2man --stderr -r0.2 -s1 -c"SBo Maintainer Tools" sbolint > sbolint.1

# This script is meant to be fairly self-contained, prefer not to
# require a huge pile of perl module dependencies. In some cases this
# means using system() or backticks or such (e.g. to run tar, instead of
# using Archive::Tar). Please don't "improve" the script by using a ton
# of modules. The POSIX module ships with perl, not afraid of using that.

# future options:
# -l list packages with errs/warnings, don't give details
# possibly some way to selectively disable the checks (does anyone
# really need this?)

# future ideas for checks:
# - REQUIRES= packages have to exist? annoying if you're working on a batch
#   of stuff to be submitted together.
# - Validate images, e.g. icon.png or .xpm or such. ImageMagick's identify
#   command can tell a non-image or a wrong-format image (a .jpg filename
#   that's actually a PNG image), but it doesn't detect truncated images.
#   Also we have to parse its stdout/stderr, it returns 0.

=pod

=head1 NAME

sbolint - check SlackBuild directories or tarballs for common errors.

=head1 SYNOPSIS

B<sbolint> [-q] [-u] [-n] [build [build ...]]

=head1 DESCRIPTION

sbolint checks for common errors in SlackBuilds.org scripts. It's
intended for slackbuild authors and maintainers, and can cut down on
"There was a problem with your upload" errors from the submission form.

The [build] arguments must be either directories or tarballs, each
containing a SlackBuild script, slack-desc, README, and .info file.
With no [build] arguments, the current directory is checked.

sbolint will flag errors for problems that would prevent the build from
being accepted by the upload form (or by the SBo admins, if if passes
the upload checks). There may also be warnings, which are things that
(probably) won't stop your build from being accepted, but may cause the
SBo admins extra work.

sbolint was not written by the SlackBuilds.org team, and shares no code
with the upload form's submission checker. Lack of errors/warnings from
sbolint does not guarantee that your build will be accepted!

sbolint doesn't check built packages, and never executes the build
script. If you want a lint tool for binary Slackware packages, try
pprkut's B<lintpkg>.

=head1 OPTIONS

=over 4

=item B<-q>

Quiet. Suppresses 'xxx checks out OK' and the total errors/warnings summary.

=item B<-u>

URL check. Uses B<curl> to make HTTP HEAD requests for the B<HOMEPAGE>,
B<DOWNLOAD>, and B<DOWNLOAD_x86_64> links. This won't guarantee that
the links are good, but some kinds of failure (e.g. site down, 404)
means they're definitely bad. Unfortunately a lot of sites have stopped
responding to HEAD requests in the name of "security", so your mileage
man vary.

=item B<-n>

Suppress warnings. Only errors will be listed. This also affects the
exit status (see below).

=back

=head1 CHECKS

For tar files only:

=over 4

=item -

File size must not be bigger than the upload form's limit (currently one
megabyte).

=item -

File must be a tar archive, possibly compressed with gzip, bzip2, or xz,
extractable by the B<tar>(1) command.

=item -

Filename extension must match compression type.

=item -

Archive must contain a directory with the same name as the archive's base name,
e.g. I<foo.tar.gz> must contain I<foo/>. Everything else in the archive must be
inside this directory.

=item -

Archive must contain I<dirname/Idirname.SlackBuild>.

=back

For all builds:

=over 4

=item -

The SlackBuild, .info, README files must have Unix \n line endings (not
DOS \r\n), and the last line of each must have a \n.

=item -

The SlackBuild script must exist, with mode 0755 (or 0644, if in a git repo),
and be a I<#!/bin/bash> script.

=item -

The script must contain the standard variable assignements for PRGNAM,
VERSION, BUILD, and TAG.

=item -

I<PRGNAM> in the script must match I<PRGNAM> in the .info file. Both must
match the script name (I<PRGNAM.SlackBuild>) and the directory name.

=item -

I<VERSION> must match the I<VERSION> in the .info file.

=item -

TAG=${TAG:-_SBo} must occur in the script.

=item -

The I<VERSION> and I<BUILD> variables must respect the environment.

=item -

The script must install the slack-desc in $PKG/install.

=item -

If there is a doinst.sh script, the SlackBuild must install it to I<$PKG/install>.

=item -

Template boilerplate comments should be removed, e.g. I<"REMOVE THIS ENTIRE BLOCK OF TEXT">
or I<"Automatically determine the architecture we're building on">.

=item -

Script must contain a B<makepkg> command.

=item -

README must exist and have mode 0644.

=item -

slack-desc must exist and have mode 0644.

=item -

slack-desc contents must match the SBo template, including the "handy-ruler",
comments, and correct spacing/indentation.

=item -

.info file must exist, have mode 0644, and match the SBo template.

=item -

.info file URLs must be valid URLs (for a very loose definition of "valid": they
must begin with B<ftp://>, B<http://>, or B<https://>).

=item -

Optionally, .info file URLs can be checked for existence with an HTTP HEAD
request (see the B<-u> option).

=back

The following tests are only done when sbolint's starting directory
was NOT in a git repo:

=over 4

=item -

Any files other than the .SlackBuild, .info, slack-desc, and README are
checked for permissions (should be 0644) and excessive size.

=item -

The source archive(s) must not exist. Also sbolint attempts to detect
extracted source trees (but isn't all that good at it).

=item -

Files named 'build.log' or 'strace.out*' must not exist. The B<sbrun>
tool creates these.

=back

The rationale for skipping the above tests when in a git repo is that
maintainers will be using git to track files and push changes, so we
don't need to check them here.

=head1 EXIT STATUS

Exit status from sbolint will normally be 0 (success) if there were no
errors or warnings in any of the builds checked. With the B<-n> option,
exit status will be 0 if there are no errors.

Exit status 1 indicates there was at least one warning or error (or, with
B<-n>, at least one error).

Any other exit status means sbolint itself failed somehow (e.g. called
with nonexistent filename).

=head1 BUGS

Probably quite a few. Watch this space for details.

=head1 AUTHOR

B. Watson (yalhcru at gmail dot com, or Urchlay on Freenode IRC)

=head1 SEE ALSO

B<sbofixinfo>(1), B<sbosearch>(1)

=cut

use POSIX qw/getcwd/;

@boilerplate = (
	qr/#\s*REMOVE THIS ENTIRE BLOCK OF TEXT/,
	qr/#\s*replace with (?:version:name) of program/,
	qr/#\s*the "_SBo" is required/,
	qr/#\s*Automatically determine the architecture we're building on/,
	qr/#\s*Unless \$ARCH is already set,/,
	qr/#\s*For consistency's sake, use this/,
	qr/#\s*Drop the package in \/tmp/,
	qr/#\s*Exit on most errors/,
	qr/#\s*If you prefer to do selective error checking with/,
	qr/#\s*Your application will probably need/,
	qr/#\s*Compile the application and install it into the/,
	qr/#\s*Strip binaries and libraries - this can be done with/,
	qr/#\s*Compress man pages$/,
	qr/#\s*Compress info pages and remove the/,
	qr/#\s*Copy program documentation into the package/,
	qr/#\s*Copy the slack-desc \(and a custom doinst\.sh if necessary\)/,
	qr/#\s*Make the package; be sure to leave it in/,
);

# this was scraped from the HTML source for the upload form:
$MAX_TARBALL_SIZE = 1048576;

($SELF = $0) =~ s,.*/,,;

$buildname = $build = "";
$g_warncount = 0;
$g_errcount = 0;
$warncount = 0;
$errcount = 0;

$tempdir = 0;

our %info = (); # has to be global, check_info sets it, check_script needs it

# main() {
#check_github_url("testing", $_) for @ARGV;
#exit 0;

while(@ARGV && ($ARGV[0] =~ /^-/)) {
	my $opt = shift;
	$opt =~ /^-u/ && do { $url_head = 1; next; };
	$opt =~ /^-d/ && do { $url_download = 1; next; };
	$opt =~ /^--?q(uiet)?/ && do { $quiet = 1; next; };
	$opt =~ /^-$/ && do { $stdin = 1; next; };
	$opt =~ /^--?h(elp)?/ && do { usage(); exit 0; };
	$opt =~ /^-n$/ && do { $nowarn = 1; next; };
	die_usage("Unrecognized option '$opt'");
}

if($url_head && $url_download) {
	die_usage("-u and -d options are mutually exclusive");
}

if($url_head || $url_download) {
	if(system("curl --version > /dev/null") != 0) {
		die "$SELF: -u and -d options require curl, can't find it in your \$PATH.\n";
	}
}

if($stdin) {
	@ARGV = <STDIN>;
	chomp for @ARGV;
}

push @ARGV, "." unless @ARGV;

# are we in a git repo? build scripts are mode 0644 there, plus
# the junkfile check is skipped.
$in_git_repo = system("git rev-parse >/dev/null 2>/dev/null") == 0;

for(@ARGV) {
	run_checks($_);
	$g_errcount += $errcount;
	$g_warncount += $warncount;

	if(!$quiet) {
		if($errcount == 0 and $warncount == 0) {
			print "$buildname checks out OK\n";
		} else {
			print "$buildname: errors $errcount, warnings $warncount\n";
		}
	}
}

# print total errs/warns only if >1 build checked
if(!$quiet && @ARGV > 1) {
	print "Total errors: $g_errcount\n";
	print "Total warnings: $g_warncount\n" unless $nowarn;
}

exit ($g_errcount > 0 || (!$nowarn && $g_warncount > 0));
# }

sub dequote {
	my $a = shift;
	#warn "dequote arg: $a\n";
	$a =~ s/^("|')(\S+)(\1)$/$2/;
	#warn "dequote ret: $a\n";
	return $a;
}

sub logmsg {
	my $severity = shift;
	my $format = shift;
	printf("$buildname: $severity: $format\n", @_);
}

sub log_error {
	logmsg("ERR", @_);
	$errcount++;
}

sub log_warning {
	return if $nowarn;
	logmsg("WARN", @_);
	$warncount++;
}

sub usage {
	if(@_) {
		warn "$SELF: $_\n" for @_;
	}

	warn <<EOF;

$SELF - check SlackBuilds.org scripts for common problems.

Usage: $SELF [-q] [-u] [-n] <build <build ...>>

builds may be directories or tarballs. If no build arguments given,
. (current directory) is assumed. Use - to read a list of tarballs/dirs
from stdin.

Options:

-q  Quiet: only emit errors/warnings, no 'checks out OK' or totals.
-u  URL Check: use HTTP HEAD request to verify download/homepage URLs exist.
-n  Suppress warnings, log only errors.

See the man page for more details. If you don't have the man page, you
can generate it with:

pod2man --stderr -s1 -cSBoStuff -r$VERSION /path/to/sbolint > sbolint.1
EOF
# not yet:
#-d  URL Download: as -u, plus download & check md5sums of download URLs.
}

sub die_usage {
	usage(@_);
	exit 1;
}

sub chdir_or_die {
	chdir($_[0]) or die "$SELF: chdir($_[0]): $!\n";
}

sub make_temp_dir {
	return if $tempdir;
	my $tmp = $ENV{TMP} || "/tmp";
	$tempdir = "$tmp/$SELF." . int(rand(2**32-1));
	system("rm -rf $tempdir");
	system("mkdir -p $tempdir");
	if(! -d $tempdir) {
		die "$SELF: can't create temp dir $tempdir\n";
	}
}

sub rm_temp_dir {
	if($tempdir && (-d $tempdir)) {
		system("rm -rf $tempdir");
		$tempdir = 0;
	}
}

sub check_tarball_mime {
	my $file = shift;

	### This stuff is a little pedantic. It also relies on having a recent-ish
	### version of GNU file (the one in Slack 14.1 works fine).
	my %types = (
			'tar' => 'application/x-tar',
			'tar.gz' => 'application/x-gzip',
			'tar.bz2' => 'application/x-bzip2',
			'tar.xz' => 'application/x-xz',
	);

	(my $basename = $file) =~ s,.*/,,;
	my (undef, $ext) = split /\./, $basename, 2;
	my $mime = `file --brief --mime-type $file`;
	chomp $mime;

	if(!grep { $_ eq $mime } values %types) {
		log_error("$file is not a tarball (mime type is '$mime')");
	} elsif(!$ext) {
		log_error("$file: filename has no extension (will be rejected by upload form)");
	} elsif($types{$ext} ne $mime) {
		log_error("$file mime type '$mime' doesn't match filename (should be $types{$ext})");
	} elsif($ext ne 'tar') {
		my $realmime = `file -z --brief --mime-type $file`;
		chomp $realmime;
		if($realmime ne 'application/x-tar') {
			log_error("$file doesn't contain a tar archive (content mime type is $realmime, should be application/x-tar)");
		}
	}
}

sub check_tarball {
	my $file = shift;

	### First, mime type checks. None of this will be fatal (no return 0 on error).
	check_tarball_mime($file);

	### one more pre-extraction check:
	if(-s "$file" > $MAX_TARBALL_SIZE) {
		log_warning("$file is larger than $MAX_TARBALL_SIZE bytes, upload may be rejected");
	}

	### now call tar to list the contents, and start returning 0 on failure.
	my @list = split "\n", `tar tf $file`;
	if($?) {
		log_error("$file: tar failed to list contents");
		return 0;
	}

	if(!@list) {
		log_error("$file is empty archive?");
		return 0;
	}

	if($list[0] ne "$buildname/") {
		log_error("$file not a SBo-compliant tarball, first element should be '$buildname/', not '$list[0]'");
		return 0;
	}

	my $foundsb = 0;
	shift @list; # 1st element is dirname/, we already checked it
	for(@list) {
		my $bn = quotemeta($buildname); # some builds have + in the name
		if(not /^$bn\//) {
			log_error("$file not a SBo-compliant tarball, contains extra junk '$_'");
			return 0;
		}

		if(/^$bn\/$bn.SlackBuild$/) {
			$foundsb = 1;
		}
	}

	if(not $foundsb) {
		log_error("$file not a SBo-compliant tarball, doesn't contain '$buildname/$buildname.SlackBuild'");
		return 0;
	}

	return 1;
}

sub extract_tarball {
	my $file = shift;
	$file = `readlink -n -e $file`;
	make_temp_dir();
	chdir_or_die($tempdir);
	system("tar xf $file");
	return "$tempdir/$buildname";
}

# run_checks will extract its argument (then cd to it) if it's a tarball,
# otherwise cd to its argument if it's a dir, otherwise error.
sub run_checks {
	$build = shift;
	my $oldcwd = getcwd();

	$errcount = $warncount = 0;

	if(-f $build || -l $build) {
		($buildname = $build) =~ s,\.tar(\..*)?$,,;
		$buildname =~ s,.*/,,;
		if(check_tarball($build)) {
			chdir_or_die(extract_tarball($build));
		} else {
			return 0;
		}
	} elsif(-d $build) {
		chdir_or_die($build);
	} else {
		die_usage "'$build' not a file or a directory.";
	}

	# last component of directory is the build name
	$buildname = `readlink -n -e .`;
	$buildname =~ s,.*/,,;

	my @checks = (
		\&check_readme,
		\&check_slackdesc,
		\&check_info,
		\&check_script,
		\&check_images,
	);

	# if we're in a git repo, it's assumed we're going to track extra
	# files with git, and use git to update the build, not tar it up
	# and use the web form.
	push @checks, \&check_junkfiles unless $in_git_repo;
	for(@checks) {
		$_->($build);
	}

	chdir_or_die($oldcwd);
	rm_temp_dir();
}

sub check_mode {
	my ($file, $wantmode) = @_;
	if(! -e $file) {
		log_error("$file does not exist");
		return 0;
	}

	my $gotmode = 07777 & ((stat($file))[2]);
	if($wantmode != $gotmode) {
		log_error("$file should be mode %04o, not %04o", $wantmode, $gotmode);
		return 0;
	}

	return 1;
}

sub check_crlf {
	my $file = shift;
	for(@_) {
		if(/\r/) {
			log_error("$file has DOS-style CRLF line endings");
			return 0;
		}
	}
	return 1;
}

sub check_and_read {
	my ($file, $mode) = @_;

	my $crlf_err;
	my @lines;
	my $lastline_nonl;

	check_mode($file, $mode);

	if(open my $fh, "<$file") {
		while(<$fh>) {
			$lastline_nonl = 1 unless /\n$/;
			chomp;
			$crlf_err = 1 if s/\r$//;
			push @lines, $_;
		}
		if(scalar @lines == 0) {
			log_error("$file exists but is empty");
		}
	}

	log_error("$file has DOS-style CRLF line endings") if $crlf_err;
	log_error("$file has no newline at EOF") if $lastline_nonl;
	return @lines;
}

sub check_readme {
	my @lines = check_and_read("README", 0644);
	if(grep { length > 72 } @lines) {
		log_warning("README has lines >72 characters");
	}
	# TODO: check encoding (must be ASCII or UTF-8)
}

# the slack-desc checking code offends me (the author), on the one hand it's
# overly complex, and on the other hand it assumes the slack-desc is at
# least close to being right...
sub check_slackdesc {
	my @lines = check_and_read("slack-desc", 0644);
	return unless scalar @lines;

	my $lineno = 1;

	if($lines[0] =~ /^# HOW TO EDIT THIS FILE:$/) {
		shift @lines;
		$lineno++;
	} else {
		log_warning("slack-desc doesn't start with how-to-edit comment");
	}

	my $count = 0;
	while($lines[0] =~ /^#/) {
		$count++;
		$lineno++;
		shift @lines;
	}

	if($count != 5) {
		log_warning("slack-desc doesn't have standard how-to-edit stanza");
	}

	$count = 0;
	while($lines[0] eq "") {
		$count++;
		$lineno++;
		shift @lines;
	}

	if($count == 0) {
		log_warning("slack-desc missing blank line before handy-ruler");
	} elsif($count > 1) {
		log_warning("slack-desc has extra blank lines before handy-ruler");
	}

	if($lines[0] =~ /handy-ruler/) {
		my $ruler = shift @lines;
		$lineno++;
		my ($spaces, $prefix, $hr, $suffix, $junk) = ($ruler =~ /^( *)(\|-+)(handy-ruler)(-+\|)(.*)$/);

		if(length($spaces) != length($buildname)) {
			log_error("slack-desc:$lineno: handy-ruler has wrong number of indent spaces (%d, should be %d)",
					length($spaces),
					length($buildname));
		}

		if(length($junk) > 0) {
			log_error("slack-desc:$lineno: handy-ruler has %d characters of trailing junk after last |", length($junk));
		}

		my $rlen = length($prefix . $hr . $suffix);
		if($rlen != 72) {
			log_error("slack-desc:$lineno: handy-ruler must be 72 characters, not %d", $rlen);
		} elsif(length($prefix) != 6) {
			log_error("slack-desc:$lineno: handy-ruler malformed, has '$prefix' instead of '|-----'");
		}
	} else {
		log_error("slack-desc missing handy-ruler");
	}

	$count = 0;
	for(@lines) {
		$count++;
		if(my ($prefix, $text) = /^([^\s]+:)(.*)/) {
			if($prefix ne "$buildname:") {
				log_error("slack-desc:$lineno: wrong prefix '$prefix', should be '$buildname:'");
			} elsif($text =~ /^\s+$/) {
				log_error("slack-desc:$lineno: trailing whitespace after colon, on otherwise-blank line");
			} elsif(length($text) > 72) {
				log_error("slack-desc:$lineno: text too long, %d characters, should be <= 72", length($text));
			} elsif(length($text) && $text !~ /^ /) {
				log_error("slack-desc:$lineno: missing whitespace after colon, on non-blank line");
			}

			my $bn = quotemeta($buildname); # some builds have + in the name
			if(($count == 1) && ($text !~ /^ $bn \(.+\)$/)) {
				log_warning("slack-desc:$lineno: first description line should be '$buildname: $buildname (short desc)'");
			}
		} else {
			log_error("slack-desc:$lineno: malformed line in description section");
		}

		$lineno++;
	}

	if($count < 11) {
		log_error("slack-desc only has $count description lines, should be 11 (add some empties)");
	} elsif($count > 11) {
		log_error("slack-desc has too many description lines ($count, should be 11)");
	}
}

# This is a damn mess. Needs refactoring badly.
sub check_info {
	my $file = $buildname . ".info";
	my @lines = check_and_read($file, 0644);
	return unless scalar @lines;

	my $lineno = 0;
	my $file_lineno = 0;
	my @expected = qw/PRGNAM VERSION HOMEPAGE
	                  DOWNLOAD MD5SUM
	                  DOWNLOAD_x86_64 MD5SUM_x86_64
	                  REQUIRES MAINTAINER EMAIL/;
	my $next_exp = 0;
	my @keys;
	my $continuation = 0;

	# parse and bitch about bad syntax...
	for(@lines) {
		$file_lineno++;
		if($continuation) {
			s/^\s*//;
			$_ = "$continuation $_";
			$continuation = 0;
			$lineno = $file_lineno - 1;
		} else {
			$lineno = $file_lineno;
		}

		if(s/\s*\\$//) {
			$continuation = $_;
			next;
		}

		if(/^\s*$/) {
			log_error("$file:$lineno: blank line (get rid of it)");
			next;
		}

		unless(/=/) {
			log_error("$file:$lineno: malformed line (no = sign, missing \\ on prev line?)");
			next;
		}

		if(s/^\s+//) {
			log_error("$file:$lineno: leading whitespace before key");
		}

		if(s/\s+$//) {
			log_error("$file:$lineno: trailing whitespace at EOL");
		}

		if(my ($k, $s1, $s2, $q1, $val, $q2) = /^(\w+)(\s*)=(\s*)("?)(.*?)("?)$/) {
			if(!grep { $k eq $_ } @expected) {
				log_error("$file:$lineno: invalid key '$k'");
			} else {
				if($k ne $expected[$next_exp]) {
					log_warning("$file:$lineno: out of order, expected $expected[$next_exp], got $k");
				}
				$next_exp++;
			}

			if(not $q1) {
				log_error("$file:$lineno: missing opening double-quote");
			}

			if(not $q2) {
				log_error("$file:$lineno: missing closing double-quote");
			}

			if(length($s1) || length($s2)) {
				log_error("$file:$lineno: no spaces allowed before/after = sign");
			}

			my $oldval = $val;
			if($val =~ s/^\s+//) {
				log_error("$file:$lineno: leading space in value: \"$oldval\"");
			}

			if($val =~ s/\s+$//) {
				log_error("$file:$lineno: trailing space in value: \"$oldval\"");
			}

			$info{$k} = $val;
		} else {
			log_error("$file:$lineno: malformed line");
		}
	}

	# parsing done, now for semantic checks

	my @missing;
	for(@expected) {
		if(not exists($info{$_})) {
			push @missing, $_;
		}
	}

	log_error("$file: missing required key(s): " . (join ", ", @missing)) if @missing;

	# init this to avoid checking undef values below
	$info{$_} ||= "" for @expected;

	if($info{PRGNAM} && ($info{PRGNAM} ne $buildname)) {
		log_error("$file: PRGNAM is '$info{PRGNAM}', should be '$buildname'");
	}

	if($info{VERSION} =~ /-/) {
		log_error("$file: VERSION may not contain - (dash) characters");
	}

	if(!check_url($info{HOMEPAGE})) {
		log_error("$file: HOMEPAGE=\"$info{HOMEPAGE}\" doesn't look like a valid URL (http, https, or ftp)");
	}

	# use a HEAD request for homepage, even if downloading other files
	if($url_head || $url_download) {
		curl_head_request($file, $info{HOMEPAGE}) || do {
			log_warning("$file: HOMEPAGE URL broken?");
		};
	}

	if($info{MD5SUM} =~ /^\s*$/) {
		log_error("$file: MD5SUM is missing or blank") unless $info{DOWNLOAD} eq 'UNSUPPORTED';
	} else {
		check_dl_and_md5($file, "");
	}

	my $dl64 = $info{DOWNLOAD_x86_64};
	if($dl64 =~ /^(?:|UNSUPPORTED|UNTESTED)$/) {
		if($info{MD5SUM_x86_64} ne "") {
			log_error("$file: MD5SUM_x86_64 must be blank if DOWNLOAD_x86_64 is not set");
		}
	} elsif($info{MD5SUM_x86_64} eq "") {
		log_error("$file: MD5SUM_x86_64 may not be blank if DOWNLOAD_x86_64 is set");
	} else {
		check_dl_and_md5($file, "_x86_64");
	}
}

sub check_dl_and_md5 {
	my($file, $suffix) = @_;
	my $md5key = "MD5SUM" . $suffix;
	my $dlkey = "DOWNLOAD" . $suffix;

	my @dlurls = split /\s+/, $info{$dlkey};
	my @md5s = split /\s+/, $info{$md5key};

	if(@md5s != @dlurls) {
		log_error("$file: we have " . @dlurls . " $dlkey URLs but " . @md5s . " $md5key" . " values");
	}

	for my $u (@dlurls) {
		if(!check_url($u)) {
			log_error("$file: $dlkey URL '$u' doesn't look like a valid URL (http, https, or ftp)");
			next;
		}

		#check_github_url($file, $u);

		if($url_head) {
			curl_head_request($file, $u) || do {
				warn '$u is '. $u;
				log_warning("$file: $dlkey URL '$u' broken?");
			};
		} elsif($url_download) {
			warn "$SELF: -d option not yet implemented\n";
		}
	}

	for(@md5s) {
		unless(/^[0-9a-f]{32}$/) {
			log_error("$file: $md5key '$_' is invalid (must be 32 hex digits)");
		}
	}

	# TODO: maybe actually download and check md5sums.
}

sub check_url {
	# url is bad if:
	return 0 if $_[0] =~ /\s/;  #  ...it contains a space,
	return 0 if $_[0] !~ /\./;  #  ...it has no dots, or
	return 0 if $_[0] !~ /\//;  #  ...it has no slashes, or
	return ($_[0] =~ /^(?:ftp|https?):\/\//); # ...it doesn't have a known protocol,
	# ...which doesn't necessarily mean it's a good URL either.
}

sub curl_head_request {
	#return !system("curl --head --location --silent --fail $_[0] >/dev/null");
	#warn $_[1];
	my $file = $_[0];
	my $client_filename = $_[1];
  	$client_filename =~ s,.*/,,;
	my $curlcmd = "curl -m20 --head --location --silent --fail $_[1]";
	open my $pipe, "$curlcmd|";
	#warn "$curlcmd";
	while(<$pipe>) {
		chomp;
		s/\r//;
		if(/^content-disposition:\s+attachment;\s+filename=["']?(.*?)["']?$/i) {
			#warn $1;
			if(defined($client_filename) && ($client_filename ne $1)) {
				log_warning("$file: download filename varies based on content disposition: '$1' vs. '$client_filename'");
			}
		}
	}
	return close($pipe);
}

# WIP, maybe no longer needed
## sub check_github_url {
## 	my $file = shift;
## 	my $url = shift;
## 	return unless $url =~ m{(https?:)//github\.com};
## 
## 	if($1 eq "http:") {
## 		log_warning("$file: github URL $url should be https");
## 	}
## 
## 	(my $expect_filename = $url) =~ s,.*/,,;
## 	my(undef, undef, undef, $user, $prog, $archive, $ver, $filename) = split /\//, $url;
## 	warn "user $user, prog $prog, archive $archive, ver $ver, filename $filename, expect_filename $expect_filename\n";
## 
## 	# assume these are correct, for now
## 	return if $user eq 'downloads';
## 	return if $archive eq 'releases';
## 
## 	# TODO: work out what to do about /raw/
## 	return if $archive eq 'raw';
## 
## 	if($archive ne 'archive') {
## 		log_warning("$file: unknown github URL type: $url");
## 		return;
## 	}
## 
## 	# OK, good URLs look like this:
## 	# https://github.com/jeetsukumaran/DendroPy/archive/v4.4.0/DendroPy-4.4.0.tar.gz
## 	# ...and bad ones look like this:
## 	# https://github.com/haiwen/seafile-client/archive/v4.4.2.tar.gz
## 	# Corrected version of the bad one would be:
## 	# https://github.com/haiwen/seafile-client/archive/v4.4.2/seafile-client-4.4.2.tar.gz
## 	# Notice the "v" isn't part of the version number. It's not always there,
## 	# and sometimes it's a different letter (r, or g, or capital V, etc).
## }

# NOT going to police the script too much. Would end up rewriting most of
# the shell, in perl. Plus, it'd become a straitjacket. Here's what I'll
# implement:
# - #!/bin/bash on line 1
# - PRGNAM must match $buildname
# - VERSION must match the .info VERSION
# - BUILD line must be present
# - TAG line must be present
# - If VERSION, BUILD, TAG don't respect the env, it's a warning
# - Check for strings like slack-desc, $PKG/install, makepkg, stuff
#   that's standard for SBo. Don't be too specific here.
# - If there's a doinst.sh, it must mentioned in the script. If not,
#   it better not be mentioned.
# - Check for leftover boilerplate
# - cp -a <documentation> is an error

sub check_script {
	my $file = $buildname . ".SlackBuild";
	my $wantmode = $in_git_repo ? 0644 : 0755;

	my @lines = check_and_read($file, $wantmode);
	return unless scalar @lines;

	if($lines[0] !~ /^#!/) {
		log_error("$file:1: missing or invalid shebang line (should be '#!/bin/bash')");
	} elsif($lines[0] !~ m,#!/bin/bash(?: (?:-e|-eu|-ue|-e -u|-u -e))?$,) {
		log_warning("$file:1: shebang line should be #!/bin/bash (possibly with -e/-u arg(s)), not '$lines[0]'");
	}

	my $lineno = 0;
	my ($prgnam, $version, $build, $tag, $need_doinst, $slackdesc, $makepkg, $install);
	my ($cdpkg, $codestart, $lint_enabled);
	$lint_enabled = 1;

	for(@lines) {
		$lineno++;

		if(/^\s*[^#]/ && !defined($codestart)) {
			$codestart = $lineno;
		}

		if(/^###sbolint\s*(\S+)/) {
			my $arg = $1;
			if(lc($arg) eq "on") {
				$lint_enabled = 1;
			} elsif(lc($arg) eq "off") {
				$lint_enabled = 0;
			} else {
				log_warning("$file:$lineno: unknown ###sbolint argument '$arg' (should be 'on' or 'off')");
			}
		}

		next unless $lint_enabled;

		# TODO: cp without -a (or -p, or a couple other flags) is OK.
##		if(/^[^#]*cp\s+(?:-\w+\s+)*[\"\$\{]*CWD/) {
##			log_error("$file:$lineno: copying files from CWD with cp (use cat instead)");
##		}

		if(/^PRGNAM=(\S+)/) {
			if($prgnam) {
				log_error("$file:$lineno: PRGNAM redefined");
			}
			$prgnam = dequote($1);
			if($prgnam ne $buildname) {
				log_error("$file:$lineno: PRGNAM doesn't match dir name ($prgnam != $buildname)");
			}
		} elsif(/^VERSION=(\S+)/ && ($lineno <= $codestart + 10)) {
			$version = dequote($1);
			if(not ($version =~ s/\$\{VERSION:-([^}]+)\}/$1/)) {
				log_warning("$file:$lineno: VERSION ignores environment, try VERSION=\${VERSION:-$version}");
			}
			$version = dequote($1);
			if($version ne $info{VERSION}) {
				log_error("$file:$lineno: VERSION ($version) doesn't match VERSION in the .info file ($info{VERSION})");
			}
		} elsif(/^BUILD=(\S+)/) {
			$build = dequote($1);
			if(not ($build =~ /\d/)) {
				log_error("$file:$lineno: BUILD is non-numeric");
			} elsif(not ($build =~ /\$\{BUILD:-\d+}/)) {
				log_warning("$file:$lineno: BUILD ignores environment (try BUILD=\${BUILD:-$build}");
			}
		} elsif(/^TAG=(\S+)/) {
			$tag = dequote($1);
			if($tag !~ /\$\{TAG:-(?:_SBo|("|')_SBo(\1))\}/) {
				log_error("$file:$lineno: TAG=\${TAG:-_SBo} is required");
			}
		} elsif(/^[^#]*\$\S*CWD\S*\/doinst.sh/) {
			$need_doinst = $lineno;
		} elsif(/^[^#]*slack-desc/) {
			$slackdesc = $lineno;
			$install = $lineno if m,install/,; # assume OK
		} elsif(/^[^#]*?cd\s+[{\$"]*PKG[}"]*/) {
			$cdpkg = $lineno;
		} elsif(/^[^#]*?["{\$]+PKG[}"]*\/install/) {
			$install = $lineno;
		} elsif($cdpkg && /^[^#]*mkdir[^#]*install/) {
			$install = $lineno;
		} elsif(/^[^#]*makepkg/) {
			if($makepkg) {
				log_error("$file:$lineno: makepkg called twice (here and line $makepkg");
			}
			$makepkg = $lineno;
		}

		if(/^[^#]*<documentation>/) {
			log_error("$file:$lineno: copy actual documentation, not <documentation>");
		}

		my $line = $_;
		if(grep { $line =~ /$_/ } @boilerplate) {
			log_error("$file:$lineno: template comment should be removed");
		}

		# special case here: don't complain about this comment if it's a perl-* build
		if($file !~ /^perl-/) {
			if($line =~ /#\s*Remove perllocal.pod and other special files/) {
				log_error("$file:$lineno: template comment should be removed");
			}
		}

	}

	if(not defined($prgnam)) {
		log_error("$file: no PRGNAM= line");
	}

	if(not defined($version)) {
		log_error("$file: no VERSION= line");
	}

	if(not defined($build)) {
		log_error("$file: no BUILD= line");
	}

	if(not defined($tag)) {
		log_error("$file: no TAG= line");
	}

	if(not defined($slackdesc)) {
		log_error("$file: doesn't seem to install slack-desc in \$PKG/install");
	}

	if(not defined($makepkg)) {
		log_error("$file: no makepkg command found");
	}

	if(not defined($install)) {
		log_error("$file: nothing gets installed in \$PKG/install");
	}

	my $have_doinst = (-f "doinst.sh");
	if($have_doinst) {
		check_and_read("doinst.sh", 0644);
	}
	if($need_doinst && !$have_doinst) {
		log_error("$file:$need_doinst: script installs doinst.sh, but it doesn't exist");
	} elsif($have_doinst && !$need_doinst) {
		log_error("$file: doinst.sh exists, but the script doesn't install it");
	}
}

# stuff like editor backups and dangling symlinks.
# maybe *any* symlinks?
# ELF objects are bad, too.
# Big-ass files...
# directories are OK, but hidden dirs are not.
sub check_junkfiles {
	my @sources = split(/\s+/, $info{DOWNLOAD} . " " . $info{DOWNLOAD_x86_64});
	s,.*/,, for @sources;
	@sources = grep { $_ !~ /^(?:\s*|UNTESTED|UNSUPPORTED)$/ } @sources;
	if(!grep { $_ =~ /^v$info{VERSION}\./ } @sources) {
		push @sources, "v$info{VERSION}.$_" for qw /zip tar.gz tar.bz2 tar.xz/;
	}

	open my $fh, "-|", "find . ! -type d -print0 | xargs -0 file --mime-type";
	FILE: while(<$fh>) {
		chomp;
		my ($file, $type) = split /: */, $_, 2;
		$file =~ s,\./,,;

		# skip the files caught by other checks
		next if $file eq "$buildname.SlackBuild";
		next if $file eq "$buildname.info";
		next if $file eq "README";
		next if $file eq "slack-desc";
		next if $file =~ /(?:diff|patch)$/;

		check_mode($file, 0644);

		if(grep { $_ eq $file } @sources) {
			log_error("source archive found: $file");
			next FILE;
		}

		for($file) {
			(/\.swp\w*$/ || /#/ || /~/ ) && do {
				log_error("editor backup found: $file");
				next FILE;
			};
			/^\./ && do {
				log_error("hidden file found: $file");
				next FILE;
			};
			/\.(?:orig|bak|old)[^.]*$/ && do {
				log_warning("$file looks like sort some of backup file");
				next FILE;
			};
			/^(?:build.log|strace.out)/ && do {
				log_warning("$file is a build log");
				next FILE;
			};
			/\.desktop$/ && do {
				system("desktop-file-validate $file");
				if($? != 0) {
					log_warning("$file fails desktop-file-validate");
					next FILE;
				}
			}
		}

		for($type) {
			($_ eq "inode/x-empty") && do {
				log_error("$file is empty (0 bytes long)");
				next FILE;
			};
			($_ =~ /^inode/) && do {
				log_error("$file is $type, not a regular file or directory");
				next FILE;
			};
			($_ =~ m,application/x-(?:executable|dosexec|object|coredump),) && do {
				log_error("$file is object code ($type)");
				next FILE;
			};
		}

		my $size = -s $file;
		if($size > 1024 * 100) {
			log_warning("$file is large ($size bytes), may be rejected by submission form");
		}
	}
	close $fh;

	open $fh, "-|", "find . -type d -mindepth 1";
	while(<$fh>) {
		chomp;
		s,\./,,;

		if(/^\./) {
			log_error("found hidden directory: $_");
			next;
		}

		if(glob("$_/*.o")) {
			log_error("$_ contains compiled object files (leftover source tree?)");
			next;
		}

		for my $badfile (qw/Makefile configure CmakeLists.txt makefile.pl SConstruct/) {
			if(-f "$_/$badfile") {
				log_error("$_ looks like extracted source tree (contains $badfile)");
			}
		}
	}
	close $fh;

#	# this won't always catch everything (e.g. PRGNAM=foo VERSION=1, but the
#	# extracted dir is foo1 or foo_1 or foo-source-1).
#	if(-d "$buildname-$version") {
#		log_warning("$buildname-$version/ looks like extracted source dir");
#	}
}

# if anything *.diff or *.patch contains \r, warn the
# user about git stripping the \r's (better gzip it).
sub check_patches {
	for(<*.diff>,<*.patch>) {
		check_and_read($_, 0644);
	}
}

# checking an image is a bit of a PITA. "file" can tell us if it's
# not an image, or has the wrong extension.
# ImageMagick's "identify" command won't detect truncated images.
# "convert" will, but it always returns 0/success, so we have to
# parse its output.
sub im_check_img {
	our %ext2mime;
	my $mime;
	my $ok = 1;

	%ext2mime = (
			png => 'image/png',
			jpg => 'image/jpeg',
			xpm => 'image/x-xpm',
			gif => 'image/gif',
	) unless %ext2mime;

	my $img = shift;
	my $ext = $img;
	$ext =~ s,.*\.,,;
	$ext = lc $ext;

	chomp($mime = `file --brief --mime "$img"`);
	if($mime !~ /$ext2mime{$ext}/) {
		log_error("$img has wrong extension $ext (MIME type is $mime)");
		return;
	}

	open my $im, "convert \"$img\" png:/dev/null 2>&1 |";
	while(<$im>) {
		$ok = 0 if /premature|corrupt/i;
	}
	close $im;

	log_error("$img appears to be corrupt") unless $ok;
}

sub check_images {
	my $images = `find . \\( -iname '*.jpg' -o -iname '*.png' -o -iname '*.xpm' -o -iname '*.gif' \\) -print0`;
	for(split /\x00/, $images) {
		check_mode($_, 0644);
		im_check_img($_);
	}
}