По поводу польских файлов предлагаю следующее.  Алгоритм никакой,
но на польских файлах срабатывает хорошо за счет non-ascii.  Было бы
интересно сделать обучающую машину, но на это может уйти много времени.


--- /usr/lib/rpm/find-provides~	2003-09-27 22:27:57 +0400
+++ /usr/lib/rpm/find-provides	2003-10-06 19:35:16 +0400
@@ -146,7 +146,7 @@
 	# Ignore symlinks for non-PAM scripts.
 	[ ! -L "$f" ] || return 0
 
-	if [ -z "${t##perl script text*}" -o -z "${t##Perl5 module source text}" ] || [ -z "${f%%*.p[lmh]}" -a -z "${t##*text*}" ]; then
+	if [ -z "${t##perl script text*}" -o -z "${f%%*.p[lmh]}" ]; then
 		if [ -n "$FIND_PERL" ]; then
 			[ -z "$LIST_PERL" ] && LIST_PERL="$f" || LIST_PERL="$LIST_PERL
 $f"

--- /usr/lib/rpm/perl.req	2003-09-30 14:42:01 +0000
+++ RPM/SOURCES/rpm-perl.req	2003-10-06 15:28:05 +0000
@@ -276,9 +276,18 @@ sub process_file {
 		process_line($_);
 	}
 	unless (close PIPE) {
-		$method ne 'relaxed'
-			and die "$fname: deparse failed.\n"
-			or warn "$fname: deparse failed.\n";
+		if ($method eq 'relaxed') {
+			warn "$fname\: deparse failed, but I don't care\n";
+		} elsif ($method eq 'strict') {
+			die "$fname\: deparse failed.\n";
+		} else {
+			my $v = isPerl($fname);
+			if ($v > 0) {
+				die "$fname\: deparse failed. isPerl=$v\n";
+			} else {
+				warn "$fname\: deparse failed, isPerl=$v, ok\n";
+			}
+		}
 	}
 }
 
@@ -390,3 +399,56 @@ foreach my $k (keys %req) {
 }
 # nothing special?
 print "perl-base\n" unless %req;
+
+# auxiliary stuff
+sub count($$) {
+	debug "$_[0] $_[1]";
+}
+
+sub isPerl {
+	my $fname = shift;
+	chomp $fname;
+	open(FILE, $fname) || die;
+	local $_ = join "" => <FILE>;
+	close FILE;
+	debug "processing $fname";
+	my ($n, @n);
+
+# POSITIVE
+# variables
+	@n = /[$%@]\w+/g;
+	count @n, "variables";
+	$n += @n;
+# comments
+	@n = /^\s*#/gm;
+	count @n, "comments";
+	$n += @n;
+# blocks
+	@n = /{$|^\s*}/gm;
+	count @n, "blocks";
+	$n += @n;
+# keywords
+	@n = /\b(unless|foreach|package|sub|use|strict)\b/gm;
+	count @n, "keywords";
+	$n += @n;
+# pod
+	@n = /^=(?:back|begin|cut|end|for|head|item|over|pod)/gm;
+	count @n, "pod sections";
+	$n += @n;
+# modules
+	@n = /^1;$/gm;
+	count @n, "`1;'";
+	$n += @n;
+
+# NEGATIVE
+# non-ascii characters
+	@n = /[^[:ascii:]]/g;
+	count @n, "non-ascii characters";
+	$n -= @n;
+# prolog
+	@n = /:-/g;
+	count @n, "prolog :- operators";
+	$n -= @n;
+# density
+	$n /= -s $fname;
+}