doc lint: fix bugs in, simplify and improve lint script

The lint-gitlink.perl script added in ab81411 (ci: validate
"linkgit:" in documentation, 2016-05-04) was more complex than it
needed to be. It:

 - Was using File::Find to recursively find *.txt files in
   Documentation/, let's instead use the Makefile as a source of truth
   for *.txt files, and pass it down to the script.

 - We now don't lint linkgit:* in RelNotes/* or technical/*, which we
   shouldn't have been doing in the first place anyway.

 - When the doc-diff script was added in beb188e (add a script to
   diff rendered documentation, 2018-08-06) we started sometimes having
   a "git worktree" under Documentation/.

   This tree contains a full checkout of git.git, as a result the
   "lint" script would recurse into that, and lint any *.txt file
   found in that entire repository.

   In practice the only in-tree "linkgit" outside of the
   Documentation/ tree is contrib/contacts/git-contacts.txt and
   contrib/subtree/git-subtree.txt, so this wouldn't emit any errors

Now we instead simply trust the Makefile to give us *.txt files.
Since the Makefile also knows what sections each page should be in we
don't have to open the files ourselves and try to parse that out. As a
bonus this will also catch bugs with the section line in the files
themselves being incorrect.

The structure of the new script is mostly based on
t/check-non-portable-shell.pl. As an added bonus it will also use
pos() to print where the problems it finds are, e.g. given an issue
like:

    diff --git a/Documentation/git-cherry.txt b/Documentation/git-cherry.txt
    [...]
     and line numbers.  git-cherry therefore detects when commits have been
    -"copied" by means of linkgit:git-cherry-pick[1], linkgit:git-am[1] or
    -linkgit:git-rebase[1].
    +"copied" by means of linkgit:git-cherry-pick[2], linkgit:git-am[3] or
    +linkgit:git-rebase[4].

We'll now emit:

    git-cherry.txt:20: error: git-cherry-pick[2]: wrong section (should be 1), shown with 'HERE' below:
    git-cherry.txt:20:      '"copied" by means of linkgit:git-cherry-pick[2]' <-- HERE
    git-cherry.txt:20: error: git-am[3]: wrong section (should be 1), shown with 'HERE' below:
    git-cherry.txt:20:      '"copied" by means of linkgit:git-cherry-pick[2], linkgit:git-am[3]' <-- HERE
    git-cherry.txt:21: error: git-rebase[4]: wrong section (should be 1), shown with 'HERE' below:
    git-cherry.txt:21:      'linkgit:git-rebase[4]' <-- HERE

Signed-off-by: Ævar Arnfjörð Bjarmason <[email protected]>
Signed-off-by: Junio C Hamano <[email protected]>

Author: avar

Documentation/Makefile

@@ -478,7 +478,11 @@ print-man1:
 	@for i in $(MAN1_TXT); do echo $$i; done
 
 lint-docs::
-	$(QUIET_LINT)$(PERL_PATH) lint-gitlink.perl
+	$(QUIET_LINT)$(PERL_PATH) lint-gitlink.perl \
+		$(HOWTO_TXT) $(DOC_DEP_TXT) \
+		--section=1 $(MAN1_TXT) \
+		--section=5 $(MAN5_TXT) \
+		--section=7 $(MAN7_TXT)
 
 ifeq ($(wildcard po/Makefile),po/Makefile)
 doc-l10n install-l10n::

Documentation/lint-gitlink.perl

@@ -2,72 +2,66 @@
 
 use strict;
 use warnings;
-use File::Find;
-use Getopt::Long;
 
-my $basedir = ".";
-GetOptions("basedir=s" => \$basedir)
-	or die("Cannot parse command line arguments\n");
+# Parse arguments, a simple state machine for input like:
+#
+# howto/*.txt config/*.txt --section=1 git.txt git-add.txt [...] --to-lint git-add.txt a-file.txt [...]
+my %TXT;
+my %SECTION;
+my $section;
+my $lint_these = 0;
+for my $arg (@ARGV) {
+	if (my ($sec) = $arg =~ /^--section=(\d+)$/s) {
+		$section = $sec;
+		next;
+	}
 
-my $found_errors = 0;
+	my ($name) = $arg =~ /^(.*?)\.txt$/s;
+	unless (defined $section) {
+		$TXT{$name} = $arg;
+		next;
+	}
 
-sub report {
-	my ($where, $what, $error) = @_;
-	print "$where: $error: $what\n";
-	$found_errors = 1;
+	$SECTION{$name} = $section;
 }
 
-sub grab_section {
-	my ($page) = @_;
-	open my $fh, "<", "$basedir/$page.txt";
-	my $firstline = <$fh>;
-	chomp $firstline;
-	close $fh;
-	my ($section) = ($firstline =~ /.*\((\d)\)$/);
-	return $section;
+my $exit_code = 0;
+sub report {
+	my ($pos, $line, $target, $msg) = @_;
+	substr($line, $pos) = "' <-- HERE";
+	$line =~ s/^\s+//;
+	print "$ARGV:$.: error: $target: $msg, shown with 'HERE' below:\n";
+	print "$ARGV:$.:\t'$line\n";
+	$exit_code = 1;
 }
 
-sub lint {
-	my ($file) = @_;
-	open my $fh, "<", $file
-		or return;
-	while (<$fh>) {
-		my $where = "$file:$.";
-		while (s/linkgit:((.*?)\[(\d)\])//) {
-			my ($target, $page, $section) = ($1, $2, $3);
+@ARGV = sort values %TXT;
+die "BUG: Nothing to process!" unless @ARGV;
+while (<>) {
+	my $line = $_;
+	while ($line =~ m/linkgit:((.*?)\[(\d)\])/g) {
+		my $pos = pos $line;
+		my ($target, $page, $section) = ($1, $2, $3);
 
-			# De-AsciiDoc
-			$page =~ s/{litdd}/--/g;
+		# De-AsciiDoc
+		$page =~ s/{litdd}/--/g;
 
-			if ($page !~ /^git/) {
-				report($where, $target, "nongit link");
-				next;
-			}
-			if (! -f "$basedir/$page.txt") {
-				report($where, $target, "no such source");
-				next;
-			}
-			my $real_section = grab_section($page);
-			if ($real_section != $section) {
-				report($where, $target,
-					"wrong section (should be $real_section)");
-				next;
-			}
+		if (!exists $TXT{$page}) {
+			report($pos, $line, $target, "link outside of our own docs");
+			next;
+		}
+		if (!exists $SECTION{$page}) {
+			report($pos, $line, $target, "link outside of our sectioned docs");
+			next;
+		}
+		my $real_section = $SECTION{$page};
+		if ($section != $SECTION{$page}) {
+			report($pos, $line, $target, "wrong section (should be $real_section)");
+			next;
 		}
 	}
-	close $fh;
-}
-
-sub lint_it {
-	lint($File::Find::name) if -f && /\.txt$/;
-}
-
-if (!@ARGV) {
-	find({ wanted => \&lint_it, no_chdir => 1 }, $basedir);
-} else {
-	for (@ARGV) {
-		lint($_);
-	}
+	# this resets our $. for each file
+	close ARGV if eof;
 }
 
-exit $found_errors;
+exit $exit_code;