git/Documentation/lint-gitlink.perl

68 lines
1.4 KiB
Plaintext
Raw Normal View History

#!/usr/bin/perl
use strict;
use warnings;
doc lint: fix bugs in, simplify and improve lint script The lint-gitlink.perl script added in ab81411ced (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 beb188e22a (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 <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-09 15:02:47 +00:00
# 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;
}
doc lint: fix bugs in, simplify and improve lint script The lint-gitlink.perl script added in ab81411ced (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 beb188e22a (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 <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-09 15:02:47 +00:00
my ($name) = $arg =~ /^(.*?)\.txt$/s;
unless (defined $section) {
$TXT{$name} = $arg;
next;
}
doc lint: fix bugs in, simplify and improve lint script The lint-gitlink.perl script added in ab81411ced (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 beb188e22a (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 <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-09 15:02:47 +00:00
$SECTION{$name} = $section;
}
doc lint: fix bugs in, simplify and improve lint script The lint-gitlink.perl script added in ab81411ced (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 beb188e22a (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 <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-09 15:02:47 +00:00
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;
}
doc lint: fix bugs in, simplify and improve lint script The lint-gitlink.perl script added in ab81411ced (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 beb188e22a (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 <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-09 15:02:47 +00:00
@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);
doc lint: fix bugs in, simplify and improve lint script The lint-gitlink.perl script added in ab81411ced (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 beb188e22a (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 <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-09 15:02:47 +00:00
# De-AsciiDoc
$page =~ s/{litdd}/--/g;
doc lint: fix bugs in, simplify and improve lint script The lint-gitlink.perl script added in ab81411ced (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 beb188e22a (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 <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-09 15:02:47 +00:00
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;
}
}
doc lint: fix bugs in, simplify and improve lint script The lint-gitlink.perl script added in ab81411ced (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 beb188e22a (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 <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-09 15:02:47 +00:00
# this resets our $. for each file
close ARGV if eof;
}
doc lint: fix bugs in, simplify and improve lint script The lint-gitlink.perl script added in ab81411ced (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 beb188e22a (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 <avarab@gmail.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-04-09 15:02:47 +00:00
exit $exit_code;